Front cover IBM System Storage DS8000 Easy Tier Server Unified storage caching and tiering solution Leverage AIX direct-attached flash devices Cache management and workload monitoring Bertrand Dufrasne Bruno Anderson Barbosa Peter Cronauer Delmar Demarchi Hans-Paul Drumm Ronny Eliahu Xin Liu Michael Stenson ibm.com/redbooks Redpaper International Technical Support Organization IBM System Storage DS8000 Easy Tier Server August 2013 REDP-5013-00 Note: Before using this information and the product it supports, read the information in “Notices” on page v. First Edition (August 2013) This edition applies to the IBM System Storage DS8870 with Licensed Machine Code (LMC) 7.7.10.xx.xx (bundle version 87.10.xxx.xx). This document was created or updated on August 7, 2013. © Copyright International Business Machines Corporation 2013. All rights reserved. Note to U.S. Government Users Restricted Rights -- Use, duplication or disclosure restricted by GSA ADP Schedule Contract with IBM Corp. Contents Notices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .v Trademarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Now you can become a published author, too! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii Comments welcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Stay connected to IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Chapter 1. Easy Tier Server overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Introduction to Easy Tier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 General Easy Tier functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.2 Easy Tier evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.3 Easy Tier fifth generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Easy Tier Server overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Business motivation for Easy Tier Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2 Easy Tier Server for performance improvements . . . . . . . . . . . . . . . . . . . . . . . . . 13 14 14 15 16 17 17 18 Chapter 2. Easy Tier Server concepts and architecture . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Easy Tier Server Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Easy Tier Server architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Easy Tier Server design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Easy Tier Server coherency client to coherency server communication. . . . . . . . 2.4 Easy Tier Server caching details. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 Caching advice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Easy Tier Server and Easy Tier data placement integration . . . . . . . . . . . . . . . . . . . . . 2.6 Direct-attached storage considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 22 22 24 25 26 27 29 30 Chapter 3. Planning for Easy Tier Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Planning and requirements guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1 Easy Tier Server coherency server requirements. . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2 Easy Tier Server coherency client requirements . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3 Supported DAS enclosures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.4 Easy Tier Server coherency client and server connectivity. . . . . . . . . . . . . . . . . . 3.2 Validating Easy Tier Server requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Easy Tier Server coherency server validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Easy Tier Server coherency client validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Other considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 DS CLI and DS GUI support to Easy Tier Server . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Easy Tier and Easy Tier Server integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3 Easy Tier Server interaction with other DS8870 advanced features. . . . . . . . . . . 33 34 34 35 36 38 38 38 42 49 49 49 49 Chapter 4. Easy Tier Server implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Implementing Easy Tier Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Setting up DS8870 for Easy Tier Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Setting up an AIX host for Easy Tier Server client . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Uninstalling Easy Tier server coherency client driver . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Upgrading Easy Tier Server coherency client driver . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 54 54 58 64 65 © Copyright IBM Corp. 2013. All rights reserved. iii iv Chapter 5. Managing Easy Tier Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Managing Easy Tier Server. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 Easy Tier Server coherency client management tasks . . . . . . . . . . . . . . . . . . . . . 5.1.2 Managing and configuring direct-attached storage . . . . . . . . . . . . . . . . . . . . . . . . 69 70 70 77 Chapter 6. Easy Tier Server monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Monitoring Easy Tier Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.1 DS8870 Storage Tier Advisor Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.2 AIX operating system IOSTAT Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.3 Monitoring with etcadmin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 84 84 87 89 Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Help from IBM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 91 91 91 92 IBM System Storage DS8000 Easy Tier Server Notices This information was developed for products and services offered in the U.S.A. IBM may not offer the products, services, or features discussed in this document in other countries. Consult your local IBM representative for information on the products and services currently available in your area. Any reference to an IBM product, program, or service is not intended to state or imply that only that IBM product, program, or service may be used. Any functionally equivalent product, program, or service that does not infringe any IBM intellectual property right may be used instead. However, it is the user's responsibility to evaluate and verify the operation of any non-IBM product, program, or service. IBM may have patents or pending patent applications covering subject matter described in this document. The furnishing of this document does not grant you any license to these patents. You can send license inquiries, in writing, to: IBM Director of Licensing, IBM Corporation, North Castle Drive, Armonk, NY 10504-1785 U.S.A. The following paragraph does not apply to the United Kingdom or any other country where such provisions are inconsistent with local law: INTERNATIONAL BUSINESS MACHINES CORPORATION PROVIDES THIS PUBLICATION "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Some states do not allow disclaimer of express or implied warranties in certain transactions, therefore, this statement may not apply to you. This information could include technical inaccuracies or typographical errors. Changes are periodically made to the information herein; these changes will be incorporated in new editions of the publication. IBM may make improvements and/or changes in the product(s) and/or the program(s) described in this publication at any time without notice. Any references in this information to non-IBM websites are provided for convenience only and do not in any manner serve as an endorsement of those websites. The materials at those websites are not part of the materials for this IBM product and use of those websites is at your own risk. IBM may use or distribute any of the information you supply in any way it believes appropriate without incurring any obligation to you. Any performance data contained herein was determined in a controlled environment. Therefore, the results obtained in other operating environments may vary significantly. Some measurements may have been made on development-level systems and there is no guarantee that these measurements will be the same on generally available systems. Furthermore, some measurements may have been estimated through extrapolation. Actual results may vary. Users of this document should verify the applicable data for their specific environment. Information concerning non-IBM products was obtained from the suppliers of those products, their published announcements or other publicly available sources. IBM has not tested those products and cannot confirm the accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on the capabilities of non-IBM products should be addressed to the suppliers of those products. This information contains examples of data and reports used in daily business operations. To illustrate them as completely as possible, the examples include the names of individuals, companies, brands, and products. All of these names are fictitious and any similarity to the names and addresses used by an actual business enterprise is entirely coincidental. COPYRIGHT LICENSE: This information contains sample application programs in source language, which illustrate programming techniques on various operating platforms. You may copy, modify, and distribute these sample programs in any form without payment to IBM, for the purposes of developing, using, marketing or distributing application programs conforming to the application programming interface for the operating platform for which the sample programs are written. These examples have not been thoroughly tested under all conditions. IBM, therefore, cannot guarantee or imply reliability, serviceability, or function of these programs. © Copyright IBM Corp. 2013. All rights reserved. v Trademarks IBM, the IBM logo, and ibm.com are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. These and other IBM trademarked terms are marked on their first occurrence in this information with the appropriate symbol (® or ™), indicating US registered or common law trademarks owned by IBM at the time this information was published. Such trademarks may also be registered or common law trademarks in other countries. A current list of IBM trademarks is available on the Web at http://www.ibm.com/legal/copytrade.shtml The following terms are trademarks of the International Business Machines Corporation in the United States, other countries, or both: AIX® DB2® DS8000® Easy Tier® Enterprise Storage Server® FlashCopy® Global Technology Services® HyperSwap® IBM® Power Systems™ POWER6® POWER7+™ POWER7® Redbooks® Redpaper™ Redbooks (logo) System p® System Storage® Tivoli® z/OS® ® The following terms are trademarks of other companies: Windows, and the Windows logo are trademarks of Microsoft Corporation in the United States, other countries, or both. UNIX is a registered trademark of The Open Group in the United States and other countries. Other company, product, or service names may be trademarks or service marks of others. vi IBM System Storage DS8000 Easy Tier Server Preface IBM® Easy Tier® Server is one of several Easy Tier enhancements introduced with the IBM DS8000® Licensed Machine Code 7.7.10.xx.xx. Easy Tier Server is a unified storage caching and tiering solution across IBM AIX® servers and supported direct-attached storage (DAS) flash drives. Easy Tier Server manages placing a copy of the “hottest” volume extents on flash drives attached to an AIX server. Data can be read directly from flash drives local to the application host rather than from cache or disk drives in the DS8870, while maintaining other advanced feature functions. This IBM Redpaper™ publication explains the Easy Tier Server concept and explores key aspects of its architecture, design, and implementation. From a more practical standpoint, this publication also contains numerous illustrations and examples that help you set up, manage, and monitor Easy Tier Server. Authors This paper was produced by a team of specialists from around the world working at the International Technical Support Organization (ITSO), San Jose Center. Bertrand Dufrasne is an IBM Certified IT Specialist and Project Leader for IBM System Storage® disk products at the ITSO, San Jose Center. He has worked at IBM in various IT areas. He has written many IBM Redbooks® publications and has developed and taught technical workshops. Before joining the ITSO, he worked for IBM Global Services as an Application Architect. He holds a Master’s degree in Electrical Engineering. Bruno Anderson Barbosa is a Senior Software Support Specialist for Global and Technology Services (GTS) in IBM Brazil. He has seven years of experience working with IBM Power Systems™, storage area network (SAN) and IBM Storage Systems. He holds a degree in IT Systems Analysis, a Business Administration postgraduation degree, and an MBA diploma in Information Technology Management from Fundação Getúlio Vargas (FGV) in Brazil. His areas of expertise include Implementation, Software Support, Problem Determination, and Performance Analysis on IBM Storage Products for Open Systems. Peter Cronauer is a certified SAP consultant who works for IBM in the European Storage Competence Center (ESCC) in Germany. He joined the IBM Advanced Technical Skills (ATS) department implementing and managing client projects for SAP solutions on IBM storage. He managed the STG Lab Services storage for Europe and the ATS development support department. Peter is the speaker of the Solution Advisory Board (SAB) and has led Redbooks residencies on various storage topics. He holds a diploma in Computer Science and wrote white papers on SAP and IBM storage solutions. Delmar Demarchi is an IT Storage Specialist at IBM Lab Services Brazil team with more than 20 years of experience. He has expertise in IBM System p® and Pure Systems, UNIX and High Availability solutions, IBM Storage Systems, and SAN products. Delmar participated in various projects involving those technologies. Delmar holds an MBA Diploma in Business and Information Technology from the Fundação Getulio Vargas in Brazil. © Copyright IBM Corp. 2013. All rights reserved. vii Hans-Paul Drumm is an IT Specialist at IBM Germany. He has 28 years of experience in the IT industry. He has worked at IBM for twelve years. He holds a degree in Computer Science from the University of Kaiserslautern. His areas of expertise include Solaris, HP-UX, Veritas Storage Foundation, and IBM z/OS®, with a special focus on Disk Solutions Attachment. Ronny Eliahu is an IBM Senior SAN and Storage Architect with over 10 years of experience. Ronny has worked for several clients with large SAN deployment. He is a member of the Storage Networking Industry Association (SNIA) and participated in a few panels for developing standards and procedures. Ronny also participated in a project in Southern France, involving IBM DB2®, SAP, and multi-partition database on IBM System p (Big Data). Xin Liu is a Senior IT Specialist in IBM China. He has seven years of experience in the technical support area for IBM system hardware and software products. He joined IBM Global Technology Services® in 2006 as a Field Software Engineer supporting Power Systems and Storage. In 2010, he became a member of the storage ATS team, and since then he worked on Storage Pre-Sales Support focusing on High-End Storage Solutions. Xin holds a Master’s degree in Electronic Engineering from Tsinghua University, China. Michael Stenson joined IBM DS8000 Product Engineering in 2005 and is currently the team lead in Tucson, Arizona. He has worked with all generations of the DS8000, as well as the IBM Enterprise Storage Server® (ESS). He has over 15 years of experience of administration and engineering in storage, server, and network environments. His current focus is process, tools, and document development. Special thanks to the Enterprise Disk team manager, Bernd Müller; ESCC Pre-Sales and Service Delivery Manager, Friedrich Gerken; and the ESCC Director, Klaus-Jürgen Rünger; for their continuous interest and support regarding the ITSO Redbooks projects. Many thanks to the following people who helped with equipment provisioning and preparation: Roland Beisele, Uwe Heinrich Müller, Hans-Joachim Sachs, Günter Schmitt, Mike Schneider, Dietmar Schniering, Stephan Schorn, Uwe Schweikhard, Edwin Weinheimer, Jörg Zahn. IBM Systems Lab Europe, Mainz, Germany Thanks to the following people for their contributions to this project: Dale H Anderson, Chiahong Chen, Lawrence Chiu, John Elliott, Yong YG Guo (Vincent), Yang Liu (Loren), Michael Lopez, Thomas Luther, Stephen Manthorpe, Allen Marin, Mei Mei, Paul Muench, Andreas Reinhardt, Brian Rinaldi, Rick Ripberger, David Sacks, Louise Schillig, Falk Schneider, Cheng-Chung Song, Jeff Steffan, Allen Wright. IBM Now you can become a published author, too! Here’s an opportunity to spotlight your skills, grow your career, and become a published author—all at the same time! Join an ITSO residency project and help write a book in your area of expertise, while honing your experience using leading-edge technologies. Your efforts will help to increase product acceptance and customer satisfaction, as you expand your network of technical contacts and relationships. Residencies run from two to six weeks in length, and you can participate either in person or as a remote resident working from your home base. Find out more about the residency program, browse the residency index, and apply online at: ibm.com/redbooks/residencies.html viii IBM System Storage DS8000 Easy Tier Server Comments welcome Your comments are important to us! We want our papers to be as helpful as possible. Send us your comments about this paper or other IBM Redbooks publications in one of the following ways: Use the online Contact us review Redbooks form found at: ibm.com/redbooks Send your comments in an email to: redbooks@us.ibm.com Mail your comments to: IBM Corporation, International Technical Support Organization Dept. HYTD Mail Station P099 2455 South Road Poughkeepsie, NY 12601-5400 Stay connected to IBM Redbooks Find us on Facebook: http://www.facebook.com/IBMRedbooks Follow us on Twitter: http://twitter.com/ibmredbooks Look for us on LinkedIn: http://www.linkedin.com/groups?home=&gid=2130806 Explore new Redbooks publications, residencies, and workshops with the IBM Redbooks weekly newsletter: https://www.redbooks.ibm.com/Redbooks.nsf/subscribe?OpenForm Stay current on recent Redbooks publications with RSS Feeds: http://www.redbooks.ibm.com/rss.html Preface ix x IBM System Storage DS8000 Easy Tier Server 1 Chapter 1. Easy Tier Server overview IBM Easy Tier Server is a unified storage caching and tiering solution across AIX servers and supported direct-attached storage (DAS) flash drives. Easy Tier Server is one of several Easy Tier enhancements, introduced with the DS8000 Licensed Machine Code 7.7.10.xx.xx. Easy Tier is now in its fifth generation. This chapter starts with a summary of the Easy Tier functions in general and briefly describes how Easy Tier has evolved over five generations. The remaining sections provide an overview of Easy Tier Server and discuss its business case and potential performance improvements it can use in a DS8870 environment. © Copyright IBM Corp. 2013. All rights reserved. 13 1.1 Introduction to Easy Tier This section reviews the various functions and evolution of Easy Tier. 1.1.1 General Easy Tier functions At the core of its functionality, Easy Tier is an optional and no-charge feature of the IBM DS8700, DS8800, and DS8870 Storage Systems that offers enhanced capabilities through automated hot spot management and data relocation, auto-rebalancing, manual volume rebalancing and volume migration, rank depopulation, merging of extent pools, and thin provisioning support. Easy Tier determines the appropriate tier of storage based on data access requirements and then automatically and non-disruptively moves data to the appropriate tier on the DS8000. The basic IBM Easy Tier features can be summarized in two operating modes: Easy Tier Automatic Mode: Easy Tier Automatic Mode is designed to automatically optimize storage performance and storage economics management across different drive tiers through data placement on a subvolume level in multitier or hybrid extent pools. Multitier or hybrid extent pools are storage pools that contain a mix of different disk drive technologies or storage tiers. It can automatically and non-disruptively relocate data at the subvolume level (extent level) across different drive tiers or even within the same drive tier according to its data temperature (I/O activity) to optimize performance and resource utilization. This feature significantly improves the overall storage cost performance ratio and simplifies storage performance tuning and management. Easy Tier Automatic Mode manages any combination of the three disk drive technology tiers available for the DS8000 series. In the DS8870, the following three disk technologies are supported: – Solid-state drives (SSDs) – Serial-attached SCSI (SAS) Enterprise disks – SAS Nearline disks Easy Tier allows cold demotion and warm demotion. Cold demotion aims to optimize the placement of extents across tiers, moving inactive extents, or extents with low activity patterns, from a high performance tier, to a lower tier. Warm demotion is designed to prevent the activity overload of a higher performance tier by demoting extents to a lower-cost tier. In Automatic Mode, Easy Tier also provides an auto-rebalance capability that adjusts the system to continuously provide excellent performance by balancing the load on the ranks within a given tier in an extent pool. Easy Tier Manual Mode: Easy Tier Manual Mode allows a set of manually initiated actions to relocate data among the storage system resources in a dynamic fashion (without any disruption of the host operations). The Manual Mode capabilities include dynamic volume relocation, dynamic extent pool merge, and rank depopulation. Dynamic volume relocation allows a DS8000 volume to be migrated to the same or another extent pool. This capability also provides the means to manually rebalance the extents of a volume across ranks when additional capacity is added to the pool. Dynamic extent pool merge allows an extent pool to be merged to another extent pool. Rank depopulation allows you to remove an allocated rank from an extent pool and relocate the allocated extents to the other ranks in the pool. 14 IBM System Storage DS8000 Easy Tier Server Combining these different capabilities greatly improves the configuration flexibility of the DS8000 providing ease of use. Even though both modes have data relocation capabilities, Manual Mode and Automatic Mode do not have the same goals: Easy Tier Manual Mode enables operations such as dynamic volume relocation and dynamic extent pool merge that simplify manual DS8000 storage management regarding capacity and performance needs. Easy Tier Automatic Mode enables automated storage performance and storage economics management through automated data placement across or even within storage tiers on extent level. Automatic Mode provides automated tiering capabilities on subvolume (extent) level across different physical resources with various performance and cost characteristics. IBM System Storage DS8000 Easy Tier, REDP-4667 covers all the Easy Tier concepts and usage, in relevant details. For more information, refer to this Redpaper publication. The general functions that we just described have been improved over time and new functions have been added. A chronological summary of those functions is presented in the section that follows: 1.1.2, “Easy Tier evolution” on page 15. 1.1.2 Easy Tier evolution The evolution of Easy Tier advanced functions throughout its five generations is summarized in Figure 1-1. Easy Tier Functions by Release Easy Tier Release/ DS8000 Model Introduced in Microcode Relea se Tier Support Auto Mode (Sub Volume) Manual Mode (Full Volume) Easy Tier 1 DS8700 R5.1 Two ti er SSD + ENT SSD + NL • Promote • Demote • Swap • Dynamic extent pool merge • Dynamic volume relocation Easy Tier 2 DS8700 DS8800 R6.1 R6.1 Any two tiers SSD + ENT SSD + NL ENT + NL • • • • • Rank depopulation • Manual vo lume rebala nce Easy Tier 3 DS8700 DS8800 R6.2 R6.2 Any th ree tiers SSD + ENT + NL • Auto Rebalance (Homog eneous Pool) • ESE Volume supp ort Easy Tier 4 DS8700 DS8800 DS8870 R6.3 R6.3 R7.0 Ful l support for FDE (encryption) d ri ves • Automatic data relocati on capabi lities for all FDE di sk environme nts Easy Tier 5 DS8870 Easy Tier App licati on R7.1 • Storage admin istrato rs can control data placement vi a CLI • Provides directive data placement API to enable software integrati on • Learning data capture and apply for heat map transfer for remote copy environme nts • Unified storage caching and tiering capability for AIX servers Easy Tier Heat Map Transfer Easy Tier Server 1 Promote Demote Swap Auto Rebalance (Hybrid pool only) • Support for all manual mod e command for FDE environments © 2013 IB M Corporat ion Figure 1-1 Easy Tier functions by release The first generation of Easy Tier introduced automated storage performance management by efficiently boosting Enterprise-class performance with SSDs and automating storage tiering from Enterprise-class (ENT) or Nearline (NL) drives to SSDs, thus optimizing SSD Chapter 1. Easy Tier Server overview 15 deployments with minimal costs. It also introduced dynamic volume relocation and dynamic extent pool merge. The second generation of Easy Tier added automated storage economics management by combining Enterprise-class drives with Nearline drives with the objective to maintain Enterprise-tier performance while shrinking the footprint and reducing costs with large capacity Nearline drives. The second generation also introduced intra-tier performance management (auto-rebalance) for hybrid pools as well as manual volume rebalance and rank depopulation. The third generation of Easy Tier introduced further enhancements that provide automated storage performance and storage economics management across all three drive tiers, which allows you to consolidate and efficiently manage more workloads on a single DS8000 system. It also introduced support for auto-rebalance in homogeneous pools and support for thin provisioned (extent space-efficient (ESE)) volumes. The fourth generation of Easy Tier added the support for Full Disk Encryption (FDE). The FDE can protect business-sensitive data by providing disk-based hardware encryption that is combined with a sophisticated key management software (IBM Tivoli® Key Lifecycle Manager).This level of tiering offers an advanced level of security and efficiency in data protection. For more information about this and other Easy Tier functions, refer to IBM System Storage DS8000 Easy Tier, REDP-4667. The fifth generation of Easy Tier starts with DS8000 Licensed Machine Code (LMC) 7.7.10.xx.xx, on bundle version 87.10.xxx.xx. The main approach of this Easy Tier generation is to move the hottest data closer to the host and overcome latency of the storage area network (SAN). Also, the Easy Tier design is evolving towards enabling applications to guide Easy Tier data placement functions in order to increase performance even more (Easy Tier Application) or maintain performance at a secondary site (Easy Tier Heat Map Transfer). These advanced features offer a self-tuning mechanism and reduce administrative costs. 1.1.3 Easy Tier fifth generation One of the Easy Tier fifth generation improvements is to move the hottest data closer to the hosts (AIX hosts only in this release). This functionality is implemented by the integrated Easy Tier Server feature, and is the focus of this IBM Redpaper publication. Easy Tier fifth generation also brings additional enhancements. It implements applicationand user-assisted data placement and optimization through another new feature, called Easy Tier Application. Moreover, it introduces the Easy Tier Heat Map Transfer Utility to manage data placement optimization across Copy Services-paired DS8870 Storage Systems. Easy Tier Application Easy Tier Application is an application-aware storage application programming interface (API) to help clients to deploy storage more efficiently. It enables applications and middleware to direct more optimal placement of data by communicating important information about current workload activity and application performance requirements. In this release, the new Easy Tier Application feature also enables clients to assign distinct application volumes to a particular tier in the Easy Tier pool, disregarding Easy Tier's advanced data migration function. This provides a flexible option for clients that want to ensure that certain applications remain on a particular tier to meet performance or cost requirements. 16 IBM System Storage DS8000 Easy Tier Server For more information about this feature, refer to IBM System Storage DS8000 Easy Tier Application, REDP-5014. Easy Tier Heat Map Transfer Easy Tier Heat Map Transfer enables a DS8870 Easy Tier optimized data placement on the primary site of either Metro Mirror, Global Copy, or Global Mirror to be applied on a DS8870 Storage System at the secondary site. With this capability, DS8000 systems can maintain application-level performance at the secondary site following a failover from the primary to secondary site. Upon receiving a heat map, also known as learning data, Easy Tier on the DS8000 at the secondary site follows the heat map to relocate data to the most optimized storage tiers, regularly. See IBM System Storage DS8000 Easy Tier Heat Map Transfer, REDP-5014 for detailed information about this feature. 1.2 Easy Tier Server overview Easy Tier Server allows the most frequently accessed or “hottest” data to be placed (cached) closer to the hosts, thus overcoming the SAN latency. This feature, introduced with the IBM DS8000 DS8870 LMC 7.7.10.xx.xx, leverages the integration between AIX hosts and DS8870 by implementing advanced caching coherency algorithms. Essentially, Easy Tier Server copies the DS8870 “hottest” data to direct-attached storage (DAS) solid-state drives (SSDs), also known as flash devices, on Easy Tier Server-enabled hosts. Easy Tier Server caching coherency algorithms ensure that the most valuable data is available on the host local flash devices and also guarantees data integrity across DS8870 internal tiers and AIX hosts DAS cache. The Easy Tier Server core relies on DS8870 cooperating with heterogeneous hosts to make a global decision on which data to copy to the hosts’ local SSDs, for improved application response time. Therefore, DAS SSD devices play an important role in an Easy Tier Server implementation. Solid-state storage direct-attached to the host means using a memory-type device for mass storage, rather than a spinning disk. IBM is making solid-state storage affordable, with innovative architectures, system and application integration, and management tools that enable effective use of solid-state storage. By eliminating the rotational delay of a spinning platter and of waiting for an arm to move to the correct position, solid-state drive technology makes data available nearly immediately. Thus, it can result in a great performance improvement when integrated with Easy Tier Server. 1.2.1 Business motivation for Easy Tier Server Traditional usage of SSD technology within storage systems can certainly enhance the system’s overall performance at many levels, but it comes at a cost. Easy Tier Server architecture might not only break through I/O performance barriers, but also helps to minimize costs, by leveraging SSD usage on the hosts, cooperatively with DS8870 Easy Tier data placement functions. DAS SSD cache on the hosts that are coherently integrated with DS8870 can enable analytical applications that were not cost- or time-effective previously. Cooperative caching can also drastically reduce I/O-bound batch windows, while improving response time and Chapter 1. Easy Tier Server overview 17 throughput. Easy Tier Server can speed up critical response time applications to allow people and systems to react more quickly and provide a higher service level to the business. Specializing in high I/O performance, SSD cache has the upper hand in cost per input/output operations per second (IOPS). To this effect, you can look at Easy Tier Server benefits from two different, yet complementary, perspectives: performance improvements and cost-efficiency ratio. With regard to performance gains, Easy Tier Server allows response time improvements by overcoming SAN latency when its algorithm fetches data from DS8870 and keeps it in the hosts’ local SSD cache, accordingly. Still, DAS SSD cache on the hosts allows scaling up and out while maintaining good performance levels. On the other hand, a good cost-efficiency ratio can be achieved by dedicating SSD resources, as cache devices, just for hosts or applications that would benefit the most from them or for the mission critical ones. Therefore, the resources would be spent based on the importance of the application to the business. Because Easy Tier Server implements a read-only local DAS cache on the hosts, there are some particular scenarios that can take the best advantage of this feature. In general terms, read-intensive environments tend to most benefit from the Easy Tier Server cooperative caching implementation. The workload types listed below are just a few examples that are a good fit for Easy Tier Server: Real-time analytics workload Large content data Online transaction processing (OLTP) workload Virtual machine (VM) consolidation Big Data Whether it is a latency sensitive environment, high read/write ratio applications, or a highly parallel processing system, there is an increasing need to process data quickly and Easy Tier Server can be considered for these situations. In some cases, the high volume of clients accessing a database can result in the need for a high IOPS rate. Nevertheless, some applications simply cannot be run fast enough to satisfy their business need. OLTP systems are the classic example of such applications. Many of these use-cases create a need to operate at high speed. In cases where the read performance of the storage can lead to a major bottleneck to the environment, there is a high value in faster storage, and therefore, a good fit for the Easy Tier Server. This feature is available at no charge with the DS8870. 1.2.2 Easy Tier Server for performance improvements The Easy Tier Server feature has been extensively tested by IBM and the test results have shown that it can improve overall response time up to five times, as explained in the demo video Planned IBM Easy Tier integration with server-based SSDs. Although this video was released in June 2012, a year before the official announcement of the feature by IBM, it presents a good visual demonstration of the Easy Tier Server concept, not mentioning the real use-case performance improvement. The video is available at the following website: http://www.youtube.com/watch?v=SLQfNoidG3I 18 IBM System Storage DS8000 Easy Tier Server In preparing this IBM Redpaper publication, we set up a similar environment to the one described in the video, with Easy Tier Server enabled on a DS8870 Storage System and a Power 770 host, attached to an EXP30 Expansion Enclosure with SSD devices. Within this environment, we were able to constantly generate random OLTP workload to a set of DS8870 logical unit numbers (LUNs) on the Power 770 host, periodically changing the I/O’s pattern, though. The results are shown in the graph presented in Figure 1-2. Disclaimer: Easy Tier Server workload tests conducted for this publication only demonstrate Easy Tier Server feature behavior and concept, under a random workload. These tests are not meant for benchmarking purposes and the results might vary depending on the workload used and the systems in the environment. Consult your IBM sales representative for benchmarking and performance-related information about Easy Tier Server. Figure 1-2 Easy Tier Server test with random OLTP workload In this scenario, the workload was shifting and changing from time to time. The several points on the horizontal axis that are indicated by letters represent a workload variation, whether it was in terms of read/write ratio, increased load, transfer sizes shifts, and so on. The overall response time, in the vertical axis, was measured by the application that generated the workload. The time is an average of both read and write response times. The tests’ results turned out to demonstrate a very good performance improvement achieved by the application after Easy Tier Server implementation, with an impressive four times improvement on the overall response time to the application. Chapter 1. Easy Tier Server overview 19 Figure 1-3 compares the highest response time that is presented in the different scenarios: with Easy Tier Server enabled, and without Easy Tier Server. Figure 1-3 Highest response time comparison in Easy Tier Server test with random OLTP workload 20 IBM System Storage DS8000 Easy Tier Server 2 Chapter 2. Easy Tier Server concepts and architecture This chapter describes the IBM Easy Tier Server concepts, architecture, and design characteristics. It provides an insight into Easy Tier Server advanced caching functions. © Copyright IBM Corp. 2013. All rights reserved. 21 2.1 Easy Tier Server Concepts Easy Tier Server is a unified storage caching and tiering implementation across AIX hosts and IBM System Storage DS8870. Easy Tier Server manages cache data placement across direct-attached storage (DAS) solid-state drives (SSDs) within IBM Power Systems hosts and DS8870 storage tiers by caching the “hottest” data on hosts’ local DAS flash disks. Easy Tier Server copies the frequently accessed data from the DS8870 storage tier to SSD drawers, directly attached to either an IBM POWER7® or IBM POWER7+™ supported host, as listed in 3.1.2, “Easy Tier Server coherency client requirements” on page 35. Thus, a host that is optimized for Easy Tier Server can read data directly from flash memory that is attached to the host cache locally, rather than from disk drives or cache in the DS8870 storage system. This data-retrieval optimization results in improved performance, with I/O requests that are generally satisfied in microseconds. Essentially, the Easy Tier Server feature provides the following capabilities: Coordination of data placement across hosts’ flash cache, the DS8870 cache, and the DS8870 internal tiers. Managed consistency of the data across the set of hosts that access the data (it is assumed that only one host will access the data at a time). Caching of the “hottest” data to the direct-attached flash disks. Significant reduction of file-retrieval times. Important: Easy Tier Server is an optional feature of the DS8870 and is available with the feature code 0715 along with the corresponding Function Authorization, 7084. No-Charge: Both Easy Tier and Easy Tier Server licenses, although required, are available at no cost. 2.2 Easy Tier Server architecture The Easy Tier Server feature consists of two major components, as represented in Figure 2-1 on page 23: The Easy Tier Server coherency server, which runs on the DS8870 system. The Easy Tier Server coherency server runs in the DS8870 and manages how data is placed onto the internal flash caches on the attached hosts. Also, it integrates with Easy Tier data placement functions for the best optimization on DS8870 internal tiers (SSD, Enterprise, and Nearline). The coherency server asynchronously communicates with the hosts system (the coherency clients) and generates caching advice for each coherency client, which is based on Easy Tier placement and statistics. The Easy Tier Server coherency client, which runs on the host system. The Easy Tier Server coherency client keeps local caches on DAS solid-state drives. The coherency client uses the Easy Tier Server protocol to establish system-aware caching that interfaces with the coherency server. An Easy Tier Server coherency client driver cooperates with the operating system to direct I/Os either to local DAS cache or to DS8870, in a transparent way to the applications. 22 IBM System Storage DS8000 Easy Tier Server Figure 2-1 shows the Easy Tier Server architecture. IBM Power System (AIX with coherency client) DAS DAS ta Da stics ati St Data Statistics DAS IBM Power System (AIX with coherency client) D Sta ata tist ics IBM Power System (AIX with coherency client) DS8870 Easy Tier Server coherency server (in the DS8870 firmware) Figure 2-1 Easy Tier Server architecture Note: In this IBM Redpaper publication, we might refer to the Easy Tier Server coherency server (DS8870), just as server, and to the Easy Tier Server coherency client (host), as client. Easy Tier Server can run in existing hosts and DS8870 environments when the Easy Tier Server software (client) is deployed on the hosts and if the direct-attached flash drawers are installed on these hosts as well. The Easy Tier Server coherency client software consists of an Easy Tier Server I/O driver and a user level daemon. On the DS8870, the Easy Tier Server coherency server is implemented natively by the Licensed Machine Code, starting in version 7.7.10.xx.xx. It must be enabled by particular licensing and configuration, as described in Chapter 3.1.1, “Easy Tier Server coherency server requirements” on page 34. The clients cooperate with the server to determine which data to cache in the DAS flash drives, while the server is monitoring I/O statistics of DS8870 internal extents to determine which ones are likely to benefit from DAS caching. When the extents are selected, the server sends a list to each Easy Tier Server coherency client via storage area network (SAN). Chapter 2. Easy Tier Server concepts and architecture 23 Attention: DAS on the client is used only as read cache. Write I/Os from the client to DS8870 disks (logical unit numbers (LUNs)) are directly transmitted to DS8870. 2.3 Easy Tier Server design Easy Tier Server coherency clients are designed to route I/O read hits to the application host DAS, while sending read misses directly to DS8870. In the same way, the write I/Os are routed to DS8870 and cache pages related to the I/O address spaces are invalidated on the client’s local cache to keep cache coherency and data integrity. Table 2-1 shows the expected client behavior upon different I/O request types. Table 2-1 Easy Tier Server coherency client behavior upon I/O Requests Host I/O request and caching Write DS8870 Normal write command processing Easy Tier Server coherency client 1. Invalidate the DAS cache contents for the addresses written 2. Send write request to the DS8870 Read Miss Read Hit Populate DAS cache based on Easy Tier Server coherency client and server heat advices Normal read command processing N/A 1. Send read request to the DS8870 2. Store data read from DS8870 in the DAS cache 1. Return data in the DAS cache to the user request handler 1. Move hot data from the DS8870 into the DAS cache Normal read command processing 2. Limit the effective size of the DAS cache so that DAS bandwidth and IOPS are not overloaded The coherency client and coherency server share statistics to ensure that the best caching decisions are made. Local DAS caches on clients have the most current information about local data access patterns. The coherency driver in the client Small Computer System Interface (SCSI) stack determines what to cache based on access patterns, server-generated advice (frequency-based advice), and local statistics (recency- and frequency-based statistics). The cache algorithm makes decisions per-I/O for what to keep in cache and what to evict. Selection algorithms achieve high efficiency by remembering the latest evicted tracks when selecting what to promote into the cache. The DS8870 receives through its host adapters, incoming I/O requests sent from hosts over the SAN. Each internal DS8870 server (Server 0 and Server 1) owns the access of half the logical subsystems’ (LSSs) worth of volumes in the DS8870. This means that the host 24 IBM System Storage DS8000 Easy Tier Server adapters in the DS8870 forward requests from the SAN to the internal server that owns access to the appropriate volume. Easy Tier Server coherency server is designed to intercept I/Os directed to Easy Tier Server-enabled LUNs. Its design implements mechanisms and algorithms to support the movement of statistics data from hosts to DS8870 and advices from DS8870 to hosts. 2.3.1 Easy Tier Server coherency client to coherency server communication The Easy Tier Server coherency clients communicate with the Easy Tier Server coherency server using SCSI commands sent over the SAN. The SCSI commands used in this context are IBM proprietary. Most of them transfer small amounts of data, in the kilobytes range, which ensures a small communication overhead of the coherency protocol. The proprietary protocol enables the server to invalidate cache on the client. This capability means that the server can maintain global cache coherency for data integrity and robustness. The Easy Tier Server I/O flow is depicted in Figure 2-2. Figure 2-2 Easy Tier Server I/O flow Chapter 2. Easy Tier Server concepts and architecture 25 2.4 Easy Tier Server caching details After receiving the list of hot extents from the server, the client then measures the heat of subextent ranges called fragments to decide which ones to copy to local DAS cache. These fragments are 1 MB in size. Important: Although the DS8870 extent size is 1 GB, the Easy Tier Server coherency client heat measurement is based on 1 MB fragments, implementing fine-granularity on heat analysis and precise performance improvements by moving to DAS only the “hottest” 1 MB fragments within the 1 GB extent. The client filters the “hottest” fragments to check which are already cached and those eligible to be cached. After a period of measuring fragment heat, the coherency client can decide to demote (evict) fragments currently in the DAS cache and replace (populate) them with hotter fragments. Client adaptive thresholds guarantee that only fragments being hotter than those evicted are populated. Moreover, the algorithm checks that the same fragment is accessed frequently enough, which filters out any sequential data from being cached. As already indicated, the Easy Tier Server coherency client manages 1 MB cache fragments. Furthermore, each fragment is in a contiguous space of the DAS pool. Although a 1 MB fragment is the unit of cache population and eviction on a client DAS pool, which is based on the server’s advice, the invalidation is done at a sector level of 512 bytes. Cache invalidation is a process whereby entries in a cache are deleted. A sector is invalidated when its data is written and the cached data on the host is not coherent anymore. Figure 2-3 represents the physical SSD 1 MB fragments and its 512 bytes valid and invalid sectors. Figure 2-3 Easy Tier Server coherency client cache granularity Upon population, the entire fragment is valid. Sectors that are later invalidated result in 512 bytes “holes” in the 1 MB SSD cache space. The population and eviction logic eliminates these holes. 26 IBM System Storage DS8000 Easy Tier Server 2.4.1 Caching advice Both the Easy Tier Server coherency client and server exchange messages to cooperatively cache and place data. The client sends I/O statistics to the server and the server sends extent placement advice to the client. To optimize short-term performance with more frequently fine-grained data movement, each client sends I/O statistics to the server every 15 minutes. The advice and feedback from the clients are used by the server in its advice generation process. Refer to Figure 2-4. To respond to client workload quickly, the server is generating the data placement advice that is based on a DS8870 one-hour average performance statistic data, always considering the latest 15-minute statistic from the server. Easy Tier Server coherency client Easy Tier Server coherency server DS8870 Figure 2-4 Easy Tier Server coherency server and client caching advices exchange As the coordinator among all heterogeneous clients, Easy Tier Server coherency server on the DS8870 has a consolidated view of all performance statistic data from all Easy Tier Server coherency clients, as depicted in Figure 2-5 on page 28. Attention: Currently, the DS8870 supports a limited number of Easy Tier Server coherency clients. For bandwidth considerations of promoting data to DAS, the DS8870 maximum number of coherency clients is 16. Chapter 2. Easy Tier Server concepts and architecture 27 Figure 2-5 shows the Easy Tier Server coherency server coordinated view among clients. Figure 2-5 Easy Tier Server coherency server coordinated view among clients The server does a good coordination of both the short-term performance optimization (DAS SSD) and the longer term performance optimization (DS8870 internal tiers). 28 IBM System Storage DS8000 Easy Tier Server 2.5 Easy Tier Server and Easy Tier data placement integration Although Easy Tier Server is responsible for promoting the “hottest” extents’ fragments to the hosts’ local DAS flash cache, Easy Tier still manages the data placement optimization within and across internal DS8870 tiers. Easy Tier operations in the DS8870 are based on the following migration types, as represented by Figure 2-6: Promote and swap for moving hot data to higher performing tiers. Warm demote which prevents performance overload of a tier by demoting warm extents to the lower tier and being triggered when bandwidth or IOPS thresholds are exceeded. Cold demote on hard disk drive (HDD) tiers, in which the coldest data is identified and moved into the Nearline tier. Expanded cold demote for HDD tiers to demote some of the sequential workload to better use the bandwidth in the Nearline tier. Auto-rebalance, which redistributes the extents within a tier to balance utilization across all ranks of the same tier for maximum performance. Figure 2-6 Easy Tier migration types The Easy Tier Server coherency server integrates with Easy Tier internal data placement functions to ensure that the home tier for given data is not adversely affected by DAS caching. For example, DAS caching can make hot data appear cold to Easy Tier because little I/O is reaching the DS8870. Typically, Easy Tier demotes cold data to Nearline class storage. If hot data cached in DAS becomes cold, an Easy Tier Server coherency client can demote that data from DAS, invalidating the selection of Nearline class home storage. So, an Easy Tier Server coherency Chapter 2. Easy Tier Server concepts and architecture 29 server informs Easy Tier not to cold demote data that is made cold by DAS caching. The server relies on the cached extent information from the client to determine which extents might be hotter than the server measured. Consequently, Easy Tier Server implies subtle changes to some of the Easy Tier migration types, as discussed below. Promote and swap migration type The promote and swap migration type moves the most frequently accessed extents in the long term to the higher performance tier in the DS8870 to improve performance. Extents cached in DAS have no read access in the storage layer, but write access is tracked. Promoting extents that are cached in DAS, if they are really hot, still have benefit for performance improvement. Therefore, extents that are cached in DAS are allowed to be promoted. Warm demote Because warm demote aims to relieve the bandwidth overload on higher performance tiers, and high-bandwidth extents are the target candidates, there is no implication on the extents cached in DAS. Cold demote and expanded cold demote Easy Tier monitors the DS8870 statistics and selects the extents with no I/O for a long time to cold demote them from the Enterprise to Nearline tier. If extents have been cached in DAS for a long time, there would have been no hits for these extents on DS8870 in this period. Thus, the I/O statistics for those extents in DS8870 will be down to zero, but they are likely frequently accessed in DAS. Easy Tier and Easy Tier Server algorithms will handle this situation in a way that extents on DAS are not cold-demoted on DS8870. Auto-rebalance The auto-rebalance migration type moves extents to balance the IOPS among ranks on the same tier. In the same way as in promote and swap, extents cached in DAS moved by auto-rebalance still have benefit for performance improvement. Therefore, extents cached in DAS can be moved by auto-rebalance. 2.6 Direct-attached storage considerations The Easy Tier Server coherency client requires at least one direct-attached storage expansion drawer with SSDs to be the host local cache for Easy Tier Server implementation. Currently, three different expansion drawer models are supported for this role: EXP30, EXP24S, and 12X I/O Drawer PCIe (#5802), as described in Chapter 3.1.1, “Easy Tier Server coherency server requirements” on page 34. Tip: Although one expansion drawer is the minimum, if you know that you will need multiple drawers, it is better to attach them from the very beginning rather than adding drawers over time. By allocating all the drawers up front, Easy Tier avoids any rebalancing when some cache devices are full. 30 IBM System Storage DS8000 Easy Tier Server When the drawer is attached to the Power Systems host, the newly added SSD devices become available to the AIX operating system as pdisks, as explained in detail in “Direct-attached storage expansion enclosure and SSD devices” on page 44. Then, you can either use each one as a different cache device for Easy Tier Server or create Redundant Array of Independent Disks (RAID) array devices for caching. In summary, a DAS Cache Directory is created on top of all physical cache devices defined on the Easy Tier Server coherency client. This approach creates a layer of abstraction between the hardware devices and the usable cache space available to the client driver, disregarding whether it consists of single SSD devices or SSD RAID arrays, as represented on Figure 2-7. If the cache device is not created on top of a RAID array, failure of one of the SSDs means that the entire cache needs to be destroyed and re-created. The SSDs may have different sizes and different performance characteristics, but the performance characteristics are not taken into account by the cache. The load is distributed evenly across the cache virtual address space. If the devices have the same size, this effectively means that the load is distributed evenly across devices, too. Figure 2-7 Easy Tier Server coherency client DAS architecture Easy Tier Server coherency client’s cache implementation, as well as DAS SSDs formatting, are discussed in “Configuring direct-attached storage flash (SSD) devices” on page 61. Chapter 2. Easy Tier Server concepts and architecture 31 32 IBM System Storage DS8000 Easy Tier Server 3 Chapter 3. Planning for Easy Tier Server Before deploying IBM Easy Tier Server, you must check the minimum software and hardware requirements for both the Easy Tier Server coherency server and the Easy Tier Server coherency client. This chapter covers these requirements and also discusses some general relevant information to be validated when deploying Easy Tier Server. © Copyright IBM Corp. 2013. All rights reserved. 33 3.1 Planning and requirements guidelines As a baseline, the Easy Tier Server implementation requires licensing and enabling both Easy Tier and Easy Tier Server on the IBM System Storage DS8870. For the client-side of this solution, an AIX host must have particular direct-attached storage (DAS) expansion drawers (with solid-state drives (SSDs)) locally attached. In addition, this host must have the Easy Tier Server coherency client driver installed and configured. Important: In order for a host to support DS8870 Easy Tier Server, you need to install and configure supported flash (SSD) devices on the host system, as indicated in 3.1.2, “Easy Tier Server coherency client requirements” on page 35. The requirements that must be met and validated during the planning phase for Easy Tier Server deployment are summarized into the following items: 1. Easy Tier Server coherency server: DS8870 a. b. c. d. DS8870 Licensed Machine Code Easy Tier Licensing Easy Tier Server Licensing Easy Tier Monitor Setting 2. Easy Tier Server coherency clients: Power Systems a. b. c. d. AIX operating system level Supported DAS Easy Tier Server coherency client driver Configuration requirements on the host 3.1.1 Easy Tier Server coherency server requirements Easy Tier Server is supported starting with DS8000 Licensed Machine Code 7.7.10.xx.xx. In addition, both Easy Tier and Easy Tier Server licenses (both, available at no charge) must be enabled on the DS8870. All of the base requirements are listed in Table 3-1. Table 3-1 DS8870 requirements for Easy Tier Server 34 Requirement Description Model DS8870 Licensed Machine Code (LMC) 7.7.10.xx.xx or higher Hardware Configuration All supported Volume Support Open System Volumes only (fixed block) Licenses Required Easy Tier (0713) and Easy Tier Server (0715), along with the correspondent Function Authorizations, 7083 and 7084, respectively. Maximum number of Easy Tier Server coherency clients attached to the DS8870 16 hosts IBM System Storage DS8000 Easy Tier Server 3.1.2 Easy Tier Server coherency client requirements Easy Tier Server is supported on AIX hosts starting with operating system levels 6100-08 and 7100-02 on selected IBM Power Systems. The fileset bos.etcacheDD.rte is the driver required for the Easy Tier Server coherency client to work with Easy Tier Server coherency server. Important: Currently, an IBM Power System running AIX is the only supported platform for Easy Tier Server coherency clients. All the base requirements for the Easy Tier Server coherency clients are listed in Table 3-2. Table 3-2 Host requirements for Easy Tier Server Requirement Description Host Platform IBM Power Systems Hardware Model Power 720 (8202-E4B/E4C/E4D), Power 740 (8205-E6B/E6C/E6D), Power 750 (8233-E8B), Power 750 (8408-E8D), Power 760 (9109-RMD), Power 770 (9117-MMB/MMC/MMD), Power 780 (9179-MHB/MHC/MHD), Power 795 (9119-FHB) Direct-Attached Storage Expansion Enclosures with SSDs EXP30 Ultra SSD I/O Drawer (FC-EDR1) with GX++ 2-port PCIe2 x8 Adapters (FC-1914) EXP24S SFF Gen2-bay Drawer (FC-5887) with PCIe2 1.8 GB Cache RAID SAS Adapters Tri-port (FC-5913) FC 5802/5877 Expansion Drawer, SSD in expansion drawer via RAID SAS adapter Operating System Native AIX No support for Virtual I/O Servers (VIOSs) OS Levels AIX 6100-08 and higher Technology Levels AIX 7100-02 and higher Technology Levels Multipathing Driver Native AIX MPIO (AIXPCM), Subsystem Device Driver (SDD), or Subsystem Device Driver Path Control Module (SDDPCM) Maximum number of SSD Cache Devices Maximum values supported by the Power Systems host SAN All Power Systems and DS8870 supported switches and host bus adapters (HBAs) Attention: Refer to the IBM System Storage Interoperation Center (SSIC) website for the most up-to-date list of supported host devices on the DS8870 at the following website: http://www.ibm.com/systems/support/storage/config/ssic Besides the hardware and software requirements for the Easy Tier Server coherency client, there are some mandatory disk and specific host adapter configurations that must be Chapter 3. Planning for Easy Tier Server 35 checked and validated during the planning phase of your Easy Tier Server implementation. The required configuration is shown in Table 3-3. Table 3-3 Host configuration requirements for Easy Tier Server Requirements Description Easy Tier Server coherency client Fileset bos.etcacheDD.rte Host bus adapter (HBA) settings fc_err_recov = fast_fail dyntrk = yes Disk (LUN) settings reserve_policy = no_reserve Therefore, all the host’s HBAs used for DS8870 connectivity must have its respective fscsiX device’s parameters checked. In the same way, all disks (DS8870 LUNs) to be configured for Easy Tier Server reserve_policy attribute must not be set to single_path. 3.1.3 Supported DAS enclosures Currently (with DS8000 Licensed Machine Code 7.7.10.xx.xx), only specific SSD expansion enclosures can be connected to an AIX host in support of the Easy Tier Server functionality. As indicated in Table 3-2 on page 35, the supported SSD enclosures are: EXP30, EXP24S, and 12X I/O Drawer PCIe (#5802). Note: For setting up any of these enclosures on the Power Systems hosts, refer to the IBM Power Systems Hardware documentation available at the following website or contact your IBM service representative: http://pic.dhe.ibm.com/infocenter/powersys/v3r1m5/index.jsp We provide below a brief description of the supported expansion drawers models. EXP30 Ultra SSD I/O Drawer (#5888 and #EDR1) The IBM EXP30 Ultra SSD I/O Drawer (feature code 5888 and EDR1) is a 1.8-inch solid-state drive (SSD) optimized PCIe storage enclosure that is mounted in a 19-inch rack. It attaches to supported POWER7 processor-based systems by using PCI Express generation 2 (PCIe2) GX++ adapters and PCIe2 cables. Figure 3-1 shows the EXP30. Figure 3-1 EXP30 Ultra SSD I/O Drawer The 5888 and EDR1 PCIe storage enclosures feature two redundant enclosure RAID modules (ERMs), which contain an SAS RAID controller, an SAS port expander, and a replaceable fan assembly. The SAS RAID controllers are always configured in a dual controller (dual storage I/O adapter) configuration. This configuration provides redundant 36 IBM System Storage DS8000 Easy Tier Server access paths to the SAS devices and mirrored copies of cache data and parity update footprints. The dual controller configuration also allows for increased performance capability when multiple RAID arrays are used in an Active/Active configuration mode. The SAS RAID controllers support RAID 0, 10, 5, and 6 and the hot-spare function. EXP24S SFF Gen2-bay Drawer (#5887): The EXP24S SFF Gen2-bay Drawer is an expansion drawer with twenty-four 2.5-inch form factor (small form factor (SFF)) SAS bays. It supports up to 24 SAS-bay-based SSD on IBM POWER6® or POWER7 servers in 2U of 19-inch rack space. The EXP24S bays are controlled by SAS adapters attached to the I/O drawer by SAS X or Y cables. Figure 3-2 shows EXP24S. Figure 3-2 EXP24S SFF Gen2-bay Drawer The EXP24S (#5887) has many high reliability design points. The SAS bays support hot swap. It has redundant and hot-swap power/fan assemblies. It has dual line cords. It has redundant and hot-swap ESMs (Electronic Service Modules). It has redundant data paths to all drives. LED indicators on drives, bays, ESMs, and power supplies support problem identification. Plus, through the SAS adapters/controllers, drives can be protected with RAID and mirroring and hot spare capability. 12X I/O Drawer PCIe (#5802) The Feature Code 5802 12X I/O Drawer PCIe is a 19-inch I/O and storage drawer. It provides a 4U-tall drawer containing 10 PCIe-based I/O adapter slots and 18 SAS hot-swap SFF disk bays. Figure 3-3 shows #5802 Expansion Drawer. Figure 3-3 12X I/O Drawer PCIe Chapter 3. Planning for Easy Tier Server 37 3.1.4 Easy Tier Server coherency client and server connectivity The same Fibre Channel connections between DS8870 and the hosts used for data access are the ones used by the Easy Tier Server coherency server to communicate with the Easy Tier Server coherency client. Important: There is no need for additional connectivity between hosts and DS8870 for Easy Tier Server implementation. Fibre Channel (FC) connectivity already used by hosts for data access on DS8870 is all that is required. 3.2 Validating Easy Tier Server requirements All Easy Tier Server implementation requirements for both the client and server are described in Table 3-1 on page 34, Table 3-2 on page 35, and Table 3-3 on page 36. In this section, we go through the validation of all those requirements. 3.2.1 Easy Tier Server coherency server validation We divided the validation into three phases: First, validating the DS8870 Licensed Machine Code; next, ensuring that required licenses on the DS8870 have been enabled; and finally, validating the Easy Tier settings. DS8870 Licensed Machine Code Before deploying Easy Tier Server on DS8870, you must ensure that its Licensed Machine Code (LMC) level is 7.7.10.xx.xx or higher. You can use the data storage graphical user interface (DS GUI) to check your LMC version. From the DS GUI main window, click Home System Status Right-Click the Storage Image Storage Image Properties Advanced tab. The resulting panel is shown in Figure 3-4. Figure 3-4 Checking DS8870 LMC with the DS GUI 38 IBM System Storage DS8000 Easy Tier Server Alternatively, you can use data storage command-line interface (DS CLI) ver command, as shown in Example 3-1. Example 3-1 Checking DS8870 LMC using DS CLI dscli> ver -lmc Storage Image LMC =========================== IBM.2107-75ZA571 7.7.10.287 DS8870 required licenses As previously mentioned, both Easy Tier and Easy Tier Server licenses are required for Easy Tier Server to be deployed. They are feature codes 0713 (Function Authorization 7083) and 0715 (Function Authorization 7084). You can obtain the authorization codes from the IBM data storage feature activation (DFSA) website at http://www.ibm.com/storage/dsfa. The codes provided are then applied using either DS CLI commands or through the DS GUI. See Chapter 10, “IBM System Storage DS8000 features and license keys” from IBM System Storage DS8870 Architecture and Implementation, SG24-8085 for more assistance on entering licenses keys on DS8870. Attention: Easy Tier and Easy Tier Server are no charge features of the IBM System Storage DS8870. However, as with any other DS8870 licensed function, they must first be ordered from IBM. Consult with your IBM sales representative if these licenses are not available for your DS8870 Storage System on the IBM DFSA website. After enabling the licenses, check the status using DS CLI or DS GUI. Example 3-2 shows how to check it using DS CLI. Example 3-2 Checking DS8870 entered license keys using DS CLI dscli> lskey -l IBM.2107-75ZA571 Activation Key Authorization Level (TB) Scope ========================================================================== Easy Tier Server on All IBM System Storage Easy Tier on All Operating environment (OEL) 170,4 All Chapter 3. Planning for Easy Tier Server 39 For checking the DS8870’s currently entered License Keys using DS GUI, click Home System Status Right-Click the appropriate Storage Image Storage Image Add Activation Key, as shown in Figure 3-5. Figure 3-5 Checking DS8870 licenses via DS GUI Easy Tier settings for Easy Tier Server Both Easy Tier Server and Easy Tier internal data placement functions depend upon DS8870 array I/O statistics and their statistics monitoring configurations are integrated on the storage system. For this reason, Easy Tier Server requires Easy Tier Monitor Mode (ETMonitor) to be enabled. Easy Tier Monitor Mode The Easy Tier monitoring capability monitors the workload on the DS8870 back-end storage, at the extent level. Monitoring statistics are gathered and analyzed every 24 hours. In an Easy Tier managed extent pool, the analysis is used to create an extent relocation or migration plan for the extents to be relocated to the most appropriate storage tier and storage resource within the pool. The Easy Tier Monitor control can be set to automode, all, or none, referring to the volumes to be monitored. The default setting is automode, which means that only volumes under control of Easy Tier Automatic Mode in managed extent pools are monitored. The settings are described as follows: In automode, only extent statistics for logical volumes in managed extent pools under control of Easy Tier Automatic Mode are updated to reflect current workload activity. When set to all, extent statistics for all logical volumes in managed and non-managed extent pools are updated to reflect current workload activity. When set to none, extent statistics collection is suspended. The Easy Tier learning data in memory is reset and all migration plans are cleared. However, the last Easy Tier summary report for the Storage Tier Advisor Tool (STAT) remains available for download and is not automatically deleted. 40 IBM System Storage DS8000 Easy Tier Server Requirement: Easy Tier Server requires Easy Tier Monitor Mode to be set to either automode or all. If the volume to be used by an Easy Tier Server-enabled host is within a DS8870 Easy Tier managed pool, automode is enough. Otherwise, all must be selected for the Easy Tier Monitor Mode. The lsextpool -l command shows whether an extent pool is managed or not by Easy Tier. Although Easy Tier Monitor Mode has to be enabled, it does not mean that Easy Tier Automatic Mode has to be on (tiered or all) for Easy Tier Server to work. Still, Easy Tier Server volumes do not need to reside in hybrid pools on DS8870, if ETMonitor is set to all. Refer to Chapter 2, “IBM System Storage DS8000 Easy Tier concepts, design, and implementation” in IBM System Storage DS8000 Easy Tier, REDP-4667 for further information about Easy Tier Monitor and operation modes. You can use either DS CLI or DS GUI for checking or changing Easy Tier Monitor Mode. If using the DS GUI, click Home System Status Right-Click the Storage Image Storage Image Properties, as shown in Figure 3-4 on page 38. For DS CLI, you can use the showsi command to display the current Easy Tier Monitor Mode setting, as illustrated in Example 3-3. Besides using DS GUI to change the Easy Tier Monitor Mode setting, the DS CLI chsi command can be used as well, with the following syntax: chsi -ETMonitor automode|all|none storage_image_ID Example 3-3 Displaying the current Easy Tier Monitor Mode setting dscli> showsi IBM.2107-75ZA571 Name DS8870_ATS02 desc Mako ID IBM.2107-75ZA571 Storage Unit IBM.2107-75ZA570 Model 961 WWNN 5005076303FFD5AA Signature XXXX-XXXX-XXXX-XXXX State Online ESSNet Enabled Volume Group V0 os400Serial 5AA NVS Memory 8.0 GB Cache Memory 233.7 GB Processor Memory 253.7 GB MTS IBM.2421-75ZA570 numegsupported 1 ETAutoMode all ETMonitor all IOPMmode Managed ETCCMode Enabled ETHMTMode Enabled Chapter 3. Planning for Easy Tier Server 41 Attention: Do not turn off Easy Tier monitoring if Easy Tier Automatic Mode (ETAutoMode) is enabled. When Easy Tier monitoring is turned off, no new performance statistics are collected and Easy Tier Automatic Mode cannot create migration plans. As a result, Easy Tier Automatic Mode stops managing volumes in all managed extent pools. Changing the Easy Tier monitoring mode affects the statistics collection and can lead to a reset (reinitialization) of the gathered monitoring data. This situation means that it might take up to 24 hours of collecting new performance statistics after Easy Tier monitoring has been enabled again until new migration plans are created. 3.2.2 Easy Tier Server coherency client validation Moving to the client-side of the implementation, the host requirements are validated. Hardware, operating system, and software First, the Power Systems model is queried by the prtconf command and AIX version is obtained by the oslevel command, as shown by Example 3-4. Example 3-4 Client’s hardware model and operating system level validation # prtconf | grep "System Model" System Model: IBM,9117-MMD # oslevel -s 7100-02-02-1316 Next, validate the multipathing software requirements by using the lslpp command, as demonstrated in Example 3-5. In our environment, we used SDDPCM as the multipathing software: devices.sddpcm.71.rte. Fileset devices.fcp.disk.ibm.mpio.rte is the required host attachment for SDDPCM in a DS8000 environment. The host attachment allows AIX Multipath Input/Output (MPIO) device driver configuration methods to properly identify and configure the DS8870 logical unit numbers (LUNs). Fileset devices.common.IBM.mpio.rte is natively installed along with AIX Base Operating System. It is the default operating system Path Control Module (PCM) used by SDDPCM. This fileset is always installed in the base operating system installation. Example 3-5 Client’s multipathing software validation # lslpp -l devices.common.IBM.mpio.rte Fileset Level State Description ---------------------------------------------------------------------------Path: /usr/lib/objrepos devices.common.IBM.mpio.rte 7.1.2.15 COMMITTED MPIO Disk Path Control Module Path: /etc/objrepos devices.common.IBM.mpio.rte 7.1.2.15 COMMITTED MPIO Disk Path Control Module # lslpp -l devices.fcp.disk.ibm.mpio.rte Fileset Level State Description ---------------------------------------------------------------------------Path: /usr/lib/objrepos devices.fcp.disk.ibm.mpio.rte 42 IBM System Storage DS8000 Easy Tier Server 1.0.0.24 COMMITTED IBM MPIO FCP Disk Device # lslpp -l devices.sddpcm.71.rte Fileset Level State Description ---------------------------------------------------------------------------Path: /usr/lib/objrepos devices.sddpcm.71.rte 2.6.3.2 COMMITTED IBM SDD PCM for AIX V71 Path: /etc/objrepos devices.sddpcm.71.rte 2.6.3.2 COMMITTED IBM SDD PCM for AIX V71 If you do not have the multipathing software installed in the AIX host system, see “Installing multipathing software” on page 58. You can also refer to the latest Multipath Subsystem Device Driver User's Guide for more information about SDD and SDDPCM installation and usage. It is available at the following website: http://www-01.ibm.com/support/docview.wss?rs=540&context=ST52G7&q=ssg1*&uid=ssg1S7 000303&loc=en_US&cs Important: The AIX operating system is discontinuing support to SDD. Currently, SDD is supported in AIX 6.1, but it is not in AIX 7.1 releases. Therefore, we strongly encourage you to use SDDPCM in your implementation, whether it is AIX 6.1 or 7.1. In Example 3-6, the manage_disk_drivers command is used to list the device driver used by the operating system to manage particular storage model devices. For DS8000 LUNs, the NO_OVERRIDE option (the default option) and the AIX_AAPCM option are supported. By using NO_OVERRIDE, you are selecting SDDPCM to manage the DS8000 disks. Instead, the AIX_AAPCM option makes the operating system use AIX native MPIO. It would be selected if the IBM HyperSwap® function were used, which is not the case with Easy Tier Server. Example 3-6 Client’s multipathing software selection validation # manage_disk_drivers -l Device Present Driver 2107DS8K NO_OVERRIDE Driver Options NO_OVERRIDE,AIX_AAPCM,NO_OVERRIDE,NO_OVERRIDE Refer to the IBM AIX 7.1 Information Center website for more information about this command and the AIX default PCM: http://pic.dhe.ibm.com/infocenter/aix/v7r1/index.jsp After checking the minimum software versions and the required Power Systems hardware model, we validate the SSD device’s availability on the AIX system. Note: At this point, the DAS enclosure with SSD devices is supposed to be already connected to the Power System. For any assistance in this matter, refer to IBM Power Systems Hardware documentation at the following website: http://pic.dhe.ibm.com/infocenter/powersys/v3r1m5/index.jsp Or, consult with your IBM service representative. Chapter 3. Planning for Easy Tier Server 43 Direct-attached storage expansion enclosure and SSD devices After the expansion enclosure is attached and properly installed onto the Power Systems host, each SSD device within the expansion enclosure is recognized by the operating system as an AIX physical disk, or pdisk. Every single pdisk is formatted into a single-device RAID 0 array automatically. Therefore, for each SSD pdisk, there is a corresponding hdisk, which is the logical representation of an SSD array. The lsdev AIX command can be used to list both pdisk and hdisk devices. At this point, you might notice that the amount of pdisks and hdisks matches, as illustrated by Example 3-7. Also, the lsdev command is used to list the expansion drawers attached to the host; in this case, an EXP30. Example 3-7 Client’s SDD devices listing # lsdev -C | grep pdisk pdisk0 Available 0M-00-00 pdisk1 Available 0M-00-00 pdisk2 Available 0M-00-00 pdisk3 Available 0M-00-00 pdisk4 Available 0M-00-00 pdisk5 Available 0M-00-00 pdisk6 Available 0M-00-00 pdisk7 Available 0M-00-00 # lsdev hdisk0 hdisk1 hdisk2 hdisk3 hdisk4 hdisk5 hdisk6 hdisk7 hdisk8 hdisk9 hdisk10 hdisk11 hdisk12 hdisk13 hdisk14 hdisk15 -Cc disk Available Available Available Available Available Available Available Available Available Available Available Available Available Available Available Available 0H-00-00 0H-00-00 0H-00-00 0H-00-00 0N-00-00 0N-00-00 0N-00-00 0N-00-00 0N-00-00 0N-00-00 0N-00-00 0N-00-00 0V-00-00 0V-00-00 0V-00-00 0V-00-00 SAS SAS SAS SAS SAS SAS SAS SAS SAS SAS SAS SAS SAS SAS SAS SAS Physical Physical Physical Physical Physical Physical Physical Physical Disk Disk Disk Disk RAID RAID RAID RAID RAID RAID RAID RAID Disk Disk Disk Disk Drive Drive Drive Drive 0 SSD 0 SSD 0 SSD 0 SSD 0 SSD 0 SSD 0 SSD 0 SSD Drive Drive Drive Drive SAS SAS SAS SAS SAS SAS SAS SAS Solid Solid Solid Solid Solid Solid Solid Solid State State State State State State State State Drive Drive Drive Drive Drive Drive Drive Drive Array Array Array Array Array Array Array Array # lsdev -Cc drawer sasdrawer0 Available 0N-00-00 EXP30 Ultra SSD I/O Drawer Now that the SSD devices were on the operating system, you can check which SAS (sissas) adapters are connecting the expansion enclosure to the Power System. The lscfg command can be used for this cross-check by the devices’ Hardware Location Codes. 44 IBM System Storage DS8000 Easy Tier Server As shown in Example 3-8, we list SSD devices with the lscfg command to get their Hardware Location Codes. Following that, we filter lscfg output with this code. Example 3-8 Client’s SSD to SAS adapter matching # lscfg -v | grep SSD sasdrawer0 UEDR1.001.G2BG00D EXP30 Ultra hdisk4 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-L204B858940-L0 hdisk5 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-L404B858940-L0 hdisk6 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-L604B858940-L0 hdisk7 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-L804B858940-L0 hdisk8 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-LA04B858940-L0 hdisk9 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-LC04B858940-L0 hdisk10 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-LE04B858940-L0 hdisk11 U2C4E.001.DBJG299-P1-C3-T1-L1-T1-L1004B858940-L0 SSD SAS SAS SAS SAS SAS SAS SAS SAS I/O Drawer RAID 0 SSD RAID 0 SSD RAID 0 SSD RAID 0 SSD RAID 0 SSD RAID 0 SSD RAID 0 SSD RAID 0 SSD Array Array Array Array Array Array Array Array # lscfg -v | grep U2C4E.001.DBJG299-P1-C3 | grep sissas sissas7 U2C4E.001.DBJG299-P1-C3-T1-L1-T1 PCIe2 3.1GB Cache RAID SAS Enclosure sissas3 U2C4E.001.DBJG299-P1-C3-T2-L1-T1 PCIe2 3.1GB Cache RAID SAS Enclosure # lsdev sissas0 sissas1 sissas2 sissas3 sissas4 sissas5 sissas6 sissas7 -Cc adapter | grep sas Available 0A-00 PCI Express Available 0G-00 PCI Express Available 0H-00 PCI Express Available 0M-00 PCIe2 3.1GB Available 0O-00 PCI Express Available 0U-00 PCI Express Available 0V-00 PCI Express Available 0N-00 PCIe2 3.1GB x1 Planar 3Gb SAS Adapter x8 Planar 3Gb SAS RAID Adapter x8 Planar 3Gb SAS RAID Adapter Cache RAID SAS Enclosure 6Gb x8 x1 Planar 3Gb SAS Adapter x8 Planar 3Gb SAS RAID Adapter x8 Planar 3Gb SAS RAID Adapter Cache RAID SAS Enclosure 6Gb x8 After the adapters connecting SSD devices to the host are known, the sissasraidmgr command can be used to show the relation between pdisks and hdisks, as demonstrated in Example 3-9. Example 3-9 Client’s SSD pdisks to SSD hdisks arrays relation # sissasraidmgr -L -j1 -l sissas7 -----------------------------------------------------------------------Name Resource State Description Size -----------------------------------------------------------------------sissas7 FEFFFFFF Primary PCIe2 3.1GB Cache RAID SAS Enclosure 6Gb x8 sissas3 FEFFFFFF HA Linked Remote adapter SN 002B9004 hdisk4 pdisk0 FC0000FF 000000FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk5 pdisk2 FC0100FF 000002FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk6 pdisk3 FC0200FF 000003FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk7 pdisk4 FC0300FF 000004FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk8 FC0400FF Optimal RAID 0 Array 387.9GB Chapter 3. Planning for Easy Tier Server 45 pdisk5 000005FF Active SSD Array Member 387.9GB hdisk9 pdisk6 FC0500FF 000006FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk10 pdisk7 FC0600FF 000007FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk11 pdisk1 FC0700FF 000001FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB Note: Despite the initial one-to-one RAID 0 array configuration, SSD devices (pdisks) can be formatted with different RAID levels and capacity (hdisks). For example, you can create one single hdisk array with all pdisks. Such configuration and best practices that are related to SSD cache devices for Easy Tier Server coherency client are discussed in “Configuring direct-attached storage flash (SSD) devices” on page 61. Adapters and disks configuration Fibre Channel adapter configuration requirements on AIX are validated, as shown in Example 3-10. The AIX lsdev command and pcmpath SDDPCM command are used to list the host bus adapters (HBAs) fcs devices on the client. Then, the lsattr command shows the attributes for the fcs correspondent fscsi devices. You must check the settings on all HBAs used for connecting the AIX host to DS8870. Example 3-10 Client devices’ settings requirement validation # lsdev fcs18 fcs19 fcs20 fcs21 -Cc adapter | grep fc Available 0E-00 8Gb PCI Express Dual Port FC Adapter (df1000f114108a03) Available 0E-01 8Gb PCI Express Dual Port FC Adapter (df1000f114108a03) Available 0k-00 PCIe2 4-Port 8Gb FC Adapter (df1000f114100104) Available 0k-01 PCIe2 4-Port 8Gb FC Adapter (df1000f114100104) # pcmpath query adapter Total Dual Active and Active/Asymmetric Adapters : 4 Adpt# Name State Mode Select 0 fscsi19 NORMAL ACTIVE 18 1 fscsi20 NORMAL ACTIVE 17 2 fscsi21 NORMAL ACTIVE 17 3 fscsi18 NORMAL ACTIVE 17 # lsattr -El attach dyntrk fc_err_recov scsi_id sw_fc_class fscsi18 switch yes fast_fail 0xd1900 3 Errors 0 0 0 0 Paths 8 8 8 8 Active 0 0 0 0 How this adapter is CONNECTED False Dynamic Tracking of FC Devices True FC Fabric Event Error RECOVERY Policy True Adapter SCSI ID False FC Class for Fabric True If the DS8870 LUNs are already mapped and recognized by the Easy Tier Server coherency client at this point, you must also validate that the hdisks’ reserve_policy is set to no_reserve, as demonstrated in Example 3-11 on page 47. 46 IBM System Storage DS8000 Easy Tier Server Example 3-11 Client DS8870 LUN disk devices’ settings validation # lsdev hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 -Cc disk | grep 2107 Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-00-02 IBM MPIO MPIO MPIO MPIO MPIO MPIO MPIO MPIO FC FC FC FC FC FC FC FC 2107 2107 2107 2107 2107 2107 2107 2107 # lsattr -El hdisk16 PCM PCM/friend/sddpcm PR_key_value none algorithm load_balance clr_q no dist_err_pcnt 0 dist_tw_width 50 flashcpy_tgtvol no hcheck_interval 60 hcheck_mode nonactive location lun_id 0x404c400000000000 lun_reset_spt yes max_coalesce 0x100000 max_transfer 0x100000 node_name 0x5005076303ffd5aa pvid 00c24dbad90815280000000000000000 q_err yes q_type simple qfull_dly 2 queue_depth 20 recoverDEDpath no reserve_policy no_reserve retry_timeout 120 rw_timeout 60 scbsy_dly 20 scsi_id 0xc2500 start_timeout 180 timeout_policy fail_path unique_id 200B75ZA5714C0007210790003IBMfcp ww_name 0x50050763031015aa PCM Reserve Key Algorithm Device CLEARS its Queue on error Distributed Error Percentage Distributed Error Sample Time Flashcopy Target Lun Health Check Interval Health Check Mode Location Label Logical Unit Number ID Support SCSI LUN reset Maximum COALESCE size Maximum TRANSFER Size FC Node Name Physical volume identifier Use QERR bit Queuing TYPE delay secs for SCSI TASK SET FULL Queue DEPTH Recover DED Failed Path Reserve Policy Retry Timeout READ/WRITE time out value delay in seconds for SCSI BUSY SCSI ID START unit time out value Timeout Policy Device Unique Identification FC World Wide Name Although the preceding example covers just one hdisk, you must check the reserve_policy setting for all DS8870 LUNs on the AIX host. Chapter 3. Planning for Easy Tier Server 47 Changing adapters and disks’ settings on AIX If the adapters and disks’ settings do not match the Easy Tier Server implementation requirements, you can use the chdev command, as shown in Example 3-12, to make them compliant. Attention: Changing these settings on hdisk and fscsi devices on AIX requires either a reboot of the host or that the device be removed and “rescanned” by the operating system. Example 3-12 Changing adapter and disk settings on AIX for Easy Tier Server compliance # pcmpath query adapter Total Dual Active and Active/Asymmetric Adapters : 4 Adpt# Name State Mode Select 0 fscsi19 NORMAL ACTIVE 63261815 1 fscsi20 NORMAL ACTIVE 63078617 2 fscsi21 NORMAL ACTIVE 62873180 3 fscsi18 NORMAL ACTIVE 59490 Errors 6 4 350 0 Paths 8 8 8 8 Active 8 8 8 8 Errors 0 Paths 8 Active 0 Paths 8 8 8 8 Active 8 8 8 8 # pcmpath set adapter 3 offline aa Success: set adapter 3 to offline Adpt# Name 3 fscsi18 State FAILED Mode OFFLINE Select 150265 # chdev -l fscsi18 -a 'fc_err_recov=fast_fail dyntrk=yes' -P fscsi18 changed # rmdev -Rl fcs18 fcnet14 Defined sfwcomm18 Defined fscsi18 Defined fcs18 Defined # cfgmgr -l fcs18 # pcmpath query adapter Total Dual Active and Active/Asymmetric Adapters : 4 Adpt# Name State Mode Select 0 fscsi19 NORMAL ACTIVE 63675002 1 fscsi20 NORMAL ACTIVE 63495052 2 fscsi21 NORMAL ACTIVE 63283670 3 fscsi18 NORMAL ACTIVE 52246 Errors 6 4 352 0 # chdev -l hdisk16 -a reserve_policy=no_reserve -P hdisk16 changed After changing the reserve_policy attribute of the hdisks to no_reserve, either reboot the system or rmdev and cfgmgr the devices. To be able to remove the devices, any structure on top of them (VGs, filesystems, and so on) can be varied off or unmounted. This includes the applications that run on top of that structure. 48 IBM System Storage DS8000 Easy Tier Server 3.3 Other considerations This section presents additional relevant planning information for Easy Tier Server deployment. This information covers the integration of Easy Tier Server with other DS8870 advanced functions and some best practices that you should be aware of. 3.3.1 DS CLI and DS GUI support to Easy Tier Server Although Easy Tier monitor and operation modes can be configured using both DS GUI and DS CLI, Easy Tier Server configuration in its current implementation is only possible with the DS CLI. 3.3.2 Easy Tier and Easy Tier Server integration In order for you to take best advantage of all the benefits of Easy Tier and Easy Tier Server integration, you need to consider using Easy Tier Automatic Mode for extents placement optimization on DS8870 internal tiers. The IBM Redpaper publication, IBM System Storage DS8000 Easy Tier, REDP-4667, presents Easy Tier general concepts and configuration guidelines based on different scenarios. 3.3.3 Easy Tier Server interaction with other DS8870 advanced features In general terms, there is no limitation or impact when it comes to Easy Tier Server integration with other DS8870 advanced features. Still, it is worthy to discuss the DS8870 major features interaction with the Easy Tier Server feature. Easy Tier Server and Easy Tier Application Easy Tier Application enables users and applications to assign distinct volumes to a particular tier in an Easy Tier pool, disregarding Easy Tier's advanced data migration function. Easy Tier Application pinning volumes up in given DS8870 internal tiers, is transparent to Easy Tier Server. If a particular volume manipulated by Easy Tier Application is also Easy Tier Server-enabled and used by an Easy Tier Server coherency client, selected fragments of the volume’s hottest extents will be selected for DAS cache population with no intervention to Easy Tier Application setting for that volume. For more information about Easy Tier Application, see IBM System Storage DS8000 Easy Tier Application, REDP-5014. Easy Tier Server and Easy Tier Heat Map Transfer Easy Tier Heat Map Transfer enables a DS8870 Easy Tier-optimized data placement on the primary site of either Metro Mirror, Global Copy, or Global Mirror to be applied on a System Storage DS8000 at the secondary site. In a scenario with Heat Map Transfer Utility (HMTU), Easy Tier Server works transparently as well. Easy Tier Heat Map Transfer only cares about DS8870 internal tiers data placement. Therefore, data populated to DAS cache in existing Easy Tier Server coherency clients on the primary site are not taken into account. Chapter 3. Planning for Easy Tier Server 49 Refer to the IBM Redpaper publication, IBM System Storage DS8000 Easy Tier Heat Map Transfer, REDP-5014, for detailed information about this feature. Easy Tier Server and DS8870 Copy Services IBM Copy Services are supported by Easy Tier and Easy Tier Server functions enabled. Metro Mirror, Global Mirror, and IBM FlashCopy® do not impact the operations of Easy Tier (in Automatic Mode or Manual Mode). Note: As well as host operations, Copy Services are unaware of the extent or volume level optimizations being performed by Easy Tier and Easy Tier Server. All back-end I/Os, except the extent migration encountered with Easy Tier Manual Mode (dynamic volume relocation) and Automatic Mode (automatic data relocation), are counted in the Easy Tier I/O statistics, including Copy Services back-end I/Os. However, most of the Copy Services background I/O activity has sequential access patterns and is not contributing to the cross-tier heat calculation of Easy Tier on a bigger scale, although it is taken into account for the bandwidth and rank utilization calculations. In a Metro Mirror environment, there is an additional time delay due to the required data transfer of a write I/O to the secondary site. This additional latency or service time is not included in the performance data considered by Easy Tier because this I/O activity is not an activity that is occurring in the disk back-end on the rank level. FlashCopy copy-on-write activity results in full track writes, which are considered large rather than small operations. The FlashCopy target is not hot in terms of small I/Os unless it is being written to. Remember that a FlashCopy track space-efficient repository is not considered for extent relocation. Easy Tier Server and DS8870 thin provisioned volumes DS8870 offers two different types of space-efficient or thin-provisioned volumes: extent space-efficient (ESE) and track space-efficient (TSE) volumes. In essence, TSE volumes are used as target volumes of a FlashCopy SE operation (with the nocopy option enabled). Instead, ESE volumes are designated for standard host access. ESE volumes are fully supported by Easy Tier Server. On the other hand, although TSE volumes can be in an extent pool managed by Easy Tier, they themselves are not managed by Easy Tier. Therefore, TSE volumes are not supported by Easy Tier Server either. Attention: Both ESE and TSE volumes can be in any extent pool managed by Easy Tier Automatic Mode, but only ESE volumes are managed and fully supported by Easy Tier and Easy Tier Server. For more information about thin provisioning on DS8000, see the IBM Redpaper publication DS8000 Thin Provisioning, REDP-4554. Easy Tier Server and I/O Priority Manager DS8870 I/O Priority Manager provides more effective storage consolidation and performance management, which is combined with the ability to align quality of service (QoS) levels to separate workloads in the system. I/O Priority Manager prioritizes access to system resources to achieve the desired QoS, based on defined performance goals (high, medium, or low) for either the volume or single I/O request. The Priority Manager constantly monitors and balances system resources to help 50 IBM System Storage DS8000 Easy Tier Server applications meet their performance targets automatically, without operator intervention, as described in DS8000 I/O Priority Manager, REDP-4760. If I/O Priority Manager, Easy Tier, and Easy Tier Server features are all enabled, they would provide independent benefits. Although I/O Priority Manager attempts to ensure that the most important I/O operations get serviced when a given rank is overloaded by delaying less important I/Os, it does not move any extents. Then, Easy Tier Server moves the “hottest” extents to DAS flash disks on the host for the fastest response time. Cooperatively, Easy Tier optimizes the extents placement moving them to the storage tier that is most appropriate for the frequency and recency of host access. Easy Tier also relocates extents between ranks within a storage tier in an attempt to distribute the workload evenly across available ranks to avoid rank overloading. Tip: Easy Tier Server, Easy Tier, and I/O Priority cooperation in a single DS8870 storage system is supported. Together, these functions can help the various applications running on DS8000 systems to meet their respective service levels in a simple and cost-effective manner. The DS8000 can help address storage consolidation requirements, which in turn helps to manage increasing amounts of data with less effort and lower infrastructure costs. Chapter 3. Planning for Easy Tier Server 51 52 IBM System Storage DS8000 Easy Tier Server 4 Chapter 4. Easy Tier Server implementation This chapter covers the details of an Easy Tier Server deployment, starting with enabling the feature on the DS8870 up to the attached AIX host configuration. The chapter also covers upgrade and uninstallation procedures of the Easy Tier Server coherency client. © Copyright IBM Corp. 2013. All rights reserved. 53 4.1 Implementing Easy Tier Server We assume that all hardware and software requirements discussed in Chapter 3.1, “Planning and requirements guidelines” on page 34 have been validated, for both the Easy Tier Server coherency client and server. Deploying Easy Tier Server starts by enabling the feature on the DS8870 storage system and then installing and configuring Easy Tier Server coherency client driver. 4.1.1 Setting up DS8870 for Easy Tier Server The Easy Tier Server feature is activated by setting the ETCCMode parameter on the DS8870 system. Another important parameter for this implementation is ETMonitor, as stated in “Easy Tier settings for Easy Tier Server” on page 40. Start by checking the status of both parameters and then enabling them if required, as shown in Example 4-1. Example 4-1 Enabling Easy Tier Server ETCCMode parameter status on DS8870 storage image dscli> showsi IBM.2107-75ZA571 Name DS8870_ATS02 desc Mako ID IBM.2107-75ZA571 Storage Unit IBM.2107-75ZA570 Model 961 WWNN 5005076303FFD5AA Signature XXXX-XXXX-XXXX-XXXX State Online ESSNet Enabled Volume Group V0 os400Serial 5AA NVS Memory 8.0 GB Cache Memory 233.7 GB Processor Memory 253.7 GB MTS IBM.2421-75ZA570 numegsupported 1 ETAutoMode all ETMonitor all IOPMmode Managed ETCCMode Disabled ETHMTMode Enabled dscli> chsi -ETCCMode enable IBM.2107-75ZA571 CMUC00042I chsi: Storage image IBM.2107-75ZA571 successfully modified. dscli> showsi IBM.2107-75ZA571 Name DS8870_ATS02 desc Mako ID IBM.2107-75ZA571 Storage Unit IBM.2107-75ZA570 Model 961 WWNN 5005076303FFD5AA Signature XXXX-XXXX-XXXX-XXXX State Online ESSNet Enabled Volume Group V0 54 IBM System Storage DS8000 Easy Tier Server os400Serial NVS Memory Cache Memory Processor Memory MTS numegsupported ETAutoMode ETMonitor IOPMmode ETCCMode ETHMTMode 5AA 8.0 GB 233.7 GB 253.7 GB IBM.2421-75ZA570 1 all all Managed Enabled Enabled At this point, the required Easy Tier Server configuration on the DS8870 is accomplished. No other setting needs to be changed. DS8870 LUNs for an Easy Tier Server coherency client No particular setting has to be defined for logical unit numbers (LUNs) on DS8870 to have them read-cache enabled on the AIX host for Easy Tier Server. As soon as regular DS8870 LUNs are mapped to Easy Tier Server coherency clients, with active and properly configured cache devices, the read cache function is enabled on the client for those LUNs, as discussed later on in “Enabling the read cache function for DS8870 hdisks” on page 62. Thus, if you already have DS8870 LUNs created and mapped to an AIX host that is going to have Easy Tier Server coherency client driver deployed, you are ready to begin. In the client configuration, you choose the LUNs for which you want to have the read cache function enabled. Important: Easy Tier Server does not require any particular setting in the LUN level on DS8870. It uses regular LUNs. The per-LUN read-cache enabled setting on the Easy Tier Server coherency client is what defines whether the LUN is Easy Tier Server-managed or not. If the AIX host is being deployed and does not have any DS8870 LUNs defined yet, proceed as usual, using the mkfbvol command to create LUNs like we did for our experimentation scenario, as shown in Example 4-2. In this scenario, the extent pool that is used for the sample volumes is managed by Easy Tier Automatic Mode, with regard to Easy Tier internal data placement optimization, and it has three different internal tiers (solid-state drive (SSD), Enterprise, and Nearline). Example 4-2 Creating regular volumes on DS8870 for an Easy Tier Server coherency client dscli> showextpool P4 Name EasyTierPool_P4 ID P4 stgtype fb totlstor (2^30B) 18027 availstor (2^30B) 14745 resvdstor (2^30B) 0 rankgrp 0 numranks 6 numvols 129 status below %allocated 18 %available 81 Chapter 4. Easy Tier Server implementation 55 configured allowed available allocated reserved %limit %threshold virextstatus %virallocated %viravailable virconfigured virallowed viravailable virallocated virreserved %virextlimit %virextthreshold encryptgrp %allocated(ese) %allocated(rep) %allocated(std) %allocated(over) %virallocated(ese) %virallocated(tse) %virallocated(init) %migrating(in) %migrating(out) numtiers etmanaged 18027 18027 14745 3282 0 100 15 below 7 92 8944 8916 8238 678 0 2 0 16 0 8 0 0 0 0 3 yes dscli> mkfbvol -dev IBM.2107-75ZA571 -extpool P4 -cap 100 -name itso_ETS00 4C00 CMUC00025I mkfbvol: FB volume 4C00 successfully created. dscli> mkfbvol -dev IBM.2107-75ZA571 -extpool P4 -cap 200 -name itso_ETS01 4C01 CMUC00025I mkfbvol: FB volume 4C01 successfully created. dscli> mkfbvol -dev IBM.2107-75ZA571 -extpool P4 -cap 50 -name itso_ETS02 4C02 CMUC00025I mkfbvol: FB volume 4C02 successfully created. dscli> mkfbvol -dev IBM.2107-75ZA571 -extpool P4 -cap 100 -name itso_ETS03 4C03 CMUC00025I mkfbvol: FB volume 4C03 successfully created. Note: Although the DS8870 extent pool used for these examples is hybrid (mixed tiers) and managed by Easy Tier, these are not requirements. For Easy Tier Server, it is only mandatory that Easy Tier Monitor Mode is either set to all or, if the volumes happen to be in pools already managed by Easy Tier, automode. DS8870 LUN mapping for an Easy Tier Server coherency client If your DS8870 storage system already has hostconnects created for the Easy Tier Server coherency client’s worldwide port name (WWPN), there is no other special setting that is required for Easy Tier Server. 56 IBM System Storage DS8000 Easy Tier Server Example 4-3 demonstrates the regular DS8870 LUN mapping process to a given host. A new volume group is created to be the container and integrator of the client’s Fibre Channel (FC) ports. The client’s FC ports are defined on DS8870 as hostconnects. The DS8870 LUNs that this client is supposed to access are added to the newly created DS8870 volume group. Example 4-3 Creating a volume group and hostconnects on DS8870 dscli> mkvolgrp -dev IBM.2107-75ZA571 -hosttype pSeries -volume 4C00-4C03 ITSO_p7_ETS_vg CMUC00030I mkvolgrp: Volume group V0 successfully created. dscli> showvolgrp v0 Name ITSO_p7_ETS_vg ID V0 Type SCSI Mask Vols 4C00 4C01 4C02 4C03 dscli> mkhostconnect -dev IBM.2107-75ZA571 -wwpn 10000090fa1f7a50 -hosttype pSeries -volgrp V0 ITSO_p7_ETS_0 CMUC00012I mkhostconnect: Host connection 0044 successfully created. dscli> mkhostconnect -dev IBM.2107-75ZA571 -wwpn 10000090fa1f7a51 -hosttype pSeries -volgrp V0 ITSO_p7_ETS_1 CMUC00012I mkhostconnect: Host connection 0045 successfully created. dscli> mkhostconnect -dev IBM.2107-75ZA571 -wwpn 10000090fa263476 -hosttype pSeries -volgrp V0 ITSO_p7_ETS_2 CMUC00012I mkhostconnect: Host connection 004C successfully created. dscli> mkhostconnect -dev IBM.2107-75ZA571 -wwpn 10000090fa263477 -hosttype pSeries -volgrp V0 ITSO_p7_ETS_3 CMUC00012I mkhostconnect: Host connection 004D successfully created. dscli> lshostconnect Name ID WWPN HostType Profile portgrp volgrpID ESSIOport =================================================================================================== ITSO_p7_ETS_0 0044 10000090FA1F7A50 pSeries IBM pSeries - AIX 0 V0 all ITSO_p7_ETS_1 0045 10000090FA1F7A51 pSeries IBM pSeries - AIX 0 V0 all ITSO_p7_ETS_2 004C 10000090FA263476 pSeries IBM pSeries - AIX 0 V0 all ITSO_p7_ETS_3 004D 10000090FA263477 pSeries IBM pSeries - AIX 0 V0 all dscli> lshostconnect -login WWNN WWPN ESSIOport LoginType Name ID =========================================================================== 20000120FA263476 10000090FA263476 I0000 SCSI ITSO_p7_ETS_2 004C 20000120FA1F7A50 10000090FA1F7A50 I0130 SCSI ITSO_p7_ETS_0 0044 20000120FA263477 10000090FA263477 I0200 SCSI ITSO_p7_ETS_3 004D 20000120FA1F7A51 10000090FA1F7A51 I0330 SCSI ITSO_p7_ETS_1 0045 For more information about DS8870 virtualization concepts and the LUN mapping process, see IBM System Storage DS8870 Architecture and Implementation, SG24-8085. Note: There is no special setting requirement for Easy Tier Server regarding storage area network (SAN) connectivity. You just need to ensure that clients and servers are properly zoned on the SAN. Chapter 4. Easy Tier Server implementation 57 4.1.2 Setting up an AIX host for Easy Tier Server client After the DS8870 storage system has been configured to be an Easy Tier Server coherency server, you need to install the appropriate drivers and configure the AIX host to be an Easy Tier Server coherency client. In summary, these are the steps that you need to follow for the client-side deployment: 1. 2. 3. 4. 5. 6. Install DS8870 multipathing software and its required device drivers. Install Easy Tier Server coherency client driver. Start Easy Tier Server coherency client driver. Format and manage the Power Systems direct-attached storage (DAS) flash devices. Configure the flash (SSD) devices or arrays as Easy Tier Server cache devices. Enable the read cache function on DS8870 provisioned hdisks. The details for all these steps follow. Installing multipathing software As indicated in Chapter 3.1.2, “Easy Tier Server coherency client requirements” on page 35, the host must have either AIX native Multipath Input/Output (MPIO), Subsystem Device Driver (SDD), or Subsystem Device Driver Path Control Module (SDDPCM) to support DS8870 LUNs. Therefore, you can have any of this supported multipathing software installed on AIX. Example 4-4 shows how to prepare SDDPCM and its host attachment driver for installation. Example 4-4 Installing SDDPCM and a host attachment for DS8000 # ls devices.fcp.disk.ibm.mpio.rte.tar devices.sddpcm.71.rte.tar devices.sddpcm.71.2.6.3.2.bff.tar # tar -xvf devices.fcp.disk.ibm.mpio.rte.tar x # tar -xvf devices.sddpcm.71.2.6.3.2.bff.tar x # tar -xvf devices.sddpcm.71.rte.tar x # inutoc . For more information about SDDPCM installation files and required host attachment drivers, see the SDDPCM user’s guide at this website: http://www-01.ibm.com/support/docview.wss?uid=ssg1S4000201 Figure 4-1 on page 59 illustrates the smit install_all command used for the actual installation and the System Management Interface Tool (SMIT) settings that you can use for the fileset installation. Instead of using the AIX SMIT, you can also use the installp command with the following syntax for the filesets installation: # installp -acgX -d <directory_path> <fileset_name> 58 IBM System Storage DS8000 Easy Tier Server # smit install_all Install and Update from ALL Available Software Type or select values in entry fields. Press Enter AFTER making all desired changes. * INPUT device / directory for software * SOFTWARE to install PREVIEW only? (install operation will NOT occur) COMMIT software updates? SAVE replaced files? AUTOMATICALLY install requisite software? EXTEND file systems if space needed? OVERWRITE same or newer versions? VERIFY install and check file sizes? DETAILED output? Process multiple volumes? ACCEPT new license agreements? PREVIEW new LICENSE agreements? [Entry Fields] . [ALL] no yes no yes yes no no yes yes yes no + + + + + + + + + + + + Figure 4-1 Fileset installation via AIX SMIT To ensure that the filesets were properly installed, the lslpp command can be used, as shown in Example 4-5. Example 4-5 Validating SDDPCM and host attachment installation # lslpp -l devices.sddpcm.71.rte Fileset Level State Description ---------------------------------------------------------------------------Path: /usr/lib/objrepos devices.sddpcm.71.rte 2.6.3.2 COMMITTED IBM SDD PCM for AIX V71 Path: /etc/objrepos devices.sddpcm.71.rte 2.6.3.2 COMMITTED IBM SDD PCM for AIX V71 # lslpp -l devices.fcp.disk.ibm.mpio.rte Fileset Level State Description ---------------------------------------------------------------------------Path: /usr/lib/objrepos devices.fcp.disk.ibm.mpio.rte 1.0.0.24 COMMITTED IBM MPIO FCP Disk Device Installing the Easy Tier Server coherency client driver Easy Tier Server coherency client driver is installed by the bos.etcacheDD.rte fileset. This installable file is provided in an optical media shipped with IBM System Storage DS8870. Otherwise, it can be downloaded from the IBM website: http://www.ibm.com/support Chapter 4. Easy Tier Server implementation 59 After obtaining the media, the fileset itself can be transferred to the AIX host. Alternatively, the ISO image file can be copied on the host and mounted by using the loopmount command, as shown in Example 4-6. Example 4-6 Mounting an ISO file on AIX for bos.etcacheDD.rte fileset installation # ls -l total 1688 -rw-r----- 1 root system 864256 Jun 3 12:51 ETCC-7.7.10.287.iso # mkdir /ETS # loopmount -i /tmp/ETS/ETCC-7.7.10.287.iso -o "-V cdrfs -o ro" -m /ETS # cd /ETS # ls bos.etcachedd The fileset can be installed using the # smit install_all command with the SMIT settings that are indicated in Figure 4-1 on page 59, or the installp command. Example 4-7 demonstrates how to check the fileset installation on AIX. Example 4-7 Checking Easy Tier Server coherency client driver installation # lslpp -l bos.etcacheDD.rte Fileset Level State Description ---------------------------------------------------------------------------Path: /usr/lib/objrepos bos.etcacheDD.rte 7.7.10.287 COMMITTED Easy Tier Server Path: /etc/objrepos bos.etcacheDD.rte 7.7.10.287 COMMITTED Easy Tier Server Starting Easy Tier Server coherency client driver The cfgmgr AIX command is used to configure and load all Easy Tier Server coherency client kernel extensions installed by the bos.etcacheDD.rte fileset. After running the cfgmgr command, the device etcdd is configured and the service etcdaemon is automatically started, as shown in Example 4-8. Example 4-8 Starting Easy Tier Server coherency client driver # cfgmgr # lsdev -Ct etcdd etcdd Available Easy Tier Cache Parent # lssrc -s etcdaemon Subsystem Group etcdaemon PID 1378020 Status active Note: The Easy Tier Server daemon is designed to be started or stopped by the Easy Tier Server client driver configuration and deconfiguration methods. Do not issue any command to manually start or stop the Easy Tier Server daemon. If the Easy Tier Server daemon is not in the expected state, contact IBM support and report the problem. 60 IBM System Storage DS8000 Easy Tier Server Configuring direct-attached storage flash (SSD) devices Easy Tier Server is now enabled and the next deployment step is to define the SSD cache devices. However, before doing so, you might want to examine the current one-to-one RAID 0 array formatting (pdisk to hdisk) that is recognized by default on the AIX host. The physical adapter on the Power System host that is connected to the DAS expansion enclosure determines the Redundant Array of Independent Disks (RAID) level supported by the SSD devices within the enclosure. Currently, most of the supported adapters allow RAID 0, 5, 6, and 10 for creating SSD arrays. Best Practice: In terms of client DAS cache devices for Easy Tier Server, any RAID configuration is supported. A general recommendation is to use RAID 5 for creating SSD arrays to be used as cache devices. Although, RAID 0 and RAID 10 would result in excellent performance as well, RAID 0 does not tolerate any disk failure, and using RAID 10 can be too costly, provided that 50% of the raw capacity of the SSD disks would be used purely for redundancy. Consider a scenario with several hdisks assigned as cache devices to the Easy Tier Server coherency client. If one drive fails within a RAID 0 array, the entire client cache directory (all cache devices) has to be destroyed and recreated to exclude the failed RAID 0 array. Because the client DAS cache is a repository filled only with data copies from DS8870 “hot” data and it is only supporting read operations from the application, there will be no impact to the system if the entire cache is lost, other than increased read response times, when the applications can no longer benefit from the read cache. However, you might want to avoid the Easy Tier Server caching directory being vulnerable to a single disk failure. Although the application data integrity is not impacted by a full cache directory loss, you cannot afford having your application running without the performance improvements inherent to Easy Tier Server, even for a while. Such situations can be prevented by formatting the SSD devices into RAID 5 arrays. Also, the fewer RAID arrays that you create, the more usable space there will be. For example, if eight 387 GB SSD devices are split into two 4-drives RAID 5 arrays, each one would provide roughly 1 TB of usable capacity. Thus, it would be around 2 TB total capacity with those two arrays. Instead, if all eight 387 GB-SSD devices are formatted into a single RAID 5 array, the total usable capacity of the array would be around 2.5 TB. Alternatively, although a two RAID 5 array DAS cache directory would tolerate the failure of two drives, one per array, a single RAID 5 array configuration tolerates just one drive failure. Therefore, it is a trade-off that you should consider when planning your cache devices. All the details for managing and formatting SSD devices into RAID arrays are described in Chapter 5.1.2, “Managing and configuring direct-attached storage” on page 77. Chapter 4. Easy Tier Server implementation 61 Defining cache devices At this point, we use the etcadmin command to create cache devices based on the newly created SSD arrays. The bos.etcacheDD.rte fileset implements the etcadmin command. This etcadmin command is the management tool for Easy Tier Server coherency client. The command is extensively used in the next implementation steps and it is also covered in details in 5.1.1, “Easy Tier Server coherency client management tasks” on page 70. Example 4-9 first shows a listing of the current SSD arrays’ hdisks reported by the lsdev command. Use the etcadmin -a list command to ensure that there is no cache device currently configured in the environment, as expected. In our scenario, there are two RAID 5 arrays: hdisk4 and hdisk5. The first cache device is then created by the etcadmin -a create command using /dev/hdisk4 as a parameter. Although hdisk4 was selected for creating the first cache device, we could have used any SSD hdisk available. Eventually, other SSD hdisks are added to the cache directory subsequently, by using the etcadmin -a add command, as we did for /dev/hdisk5. Note: When dealing with cache devices as parameters of the etcadmin command, the prefix /dev must always be used. Example 4-9 Creating a client caching directory and adding cache devices # lsdev -Cc disk | grep SSD hdisk4 Available 0N-00-00 SAS RAID 5 SSD Array hdisk5 Available 0N-00-00 SAS RAID 5 SSD Array # etcadmin -a list No cache device has been created # etcadmin -a create -d /dev/hdisk4 Cache device has been created with Flash Cache /dev/hdisk4 # etcadmin -a add -d /dev/hdisk5 Flash Cache device /dev/hdisk5 has been added to cache device # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | Enabling the read cache function for DS8870 hdisks Easy Tier Server coherency client cache devices are ready, as previously reported by the etcadmin -a list command. The next step is to assign DS8870 hdisks to be read-cache enabled for Easy Tier Server. Example 4-10 on page 63 shows a list of the DS8870 LUNs recognized by our particular host. The etcadmin -a start command is used to start the read cache function on the LUNs and the etcadmin -a query is used next to validate that the read cache function has been enabled for the selected devices. 62 IBM System Storage DS8000 Easy Tier Server Example 4-10 Enabling the Easy Tier Server read cache function on DS8870 hdisks # lsdev hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 -Cc disk | grep 2107 Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-00-02 IBM MPIO MPIO MPIO MPIO MPIO MPIO MPIO MPIO FC FC FC FC FC FC FC FC 2107 2107 2107 2107 2107 2107 2107 2107 # etcadmin -a start -D hdisk16 hdisk23 SAN device hdisk16 read_cache function SAN device hdisk17 read_cache function SAN device hdisk18 read_cache function SAN device hdisk19 read_cache function SAN device hdisk20 read_cache function SAN device hdisk21 read_cache function SAN device hdisk22 read_cache function SAN device hdisk23 read_cache function is is is is is is is is enabled enabled enabled enabled enabled enabled enabled enabled # etcadmin -a query Total read cache enabled SAN devices is 8 ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 From the moment when a device read cache function is started on DAS cache, the Easy Tier Server coherency client and server initiate a learning phase of the volumes’ data pattern for about 15 minutes. After that, read caching effectively starts. Refer to “Managing and enabling the read cache function on DS8870 hdisks” on page 72, and 6.1.2, “AIX operating system IOSTAT Tool” on page 87 for more details and examples about read cache function enablement and startup. Chapter 4. Easy Tier Server implementation 63 4.2 Uninstalling Easy Tier server coherency client driver If for any reason you no longer want or need the coherency client driver on a particular AIX host, you can uninstall the driver. Before uninstalling the Easy Tier Server coherency client driver, you must stop the read cache function for all DS8870 hdisks that have it enabled. For that purpose, use the etcadmin -a shutdown command as shown in Example 4-11. Example 4-11 Shutting down Easy Tier Server coherency client driver # etcadmin -a shutdown -s no Total read cache enabled SAN devices is 8 -------------------------------------------SAN device hdisk16 read_cache function is disabled SAN device hdisk17 read_cache function is disabled SAN device hdisk18 read_cache function is disabled SAN device hdisk19 read_cache function is disabled SAN device hdisk20 read_cache function is disabled SAN device hdisk21 read_cache function is disabled SAN device hdisk22 read_cache function is disabled SAN device hdisk23 read_cache function is disabled Removing etcdd device... etcdd device is deleted successfully The shutdown parameter from etcadmin stops all read-cache enabled DS8870 devices and removes the cache directory (and so its cache devices). In addition, it unconfigures and removes the etcdd device from the client. In Example 4-12, you can see that the etcdd device has been removed and the etcdaemon has been stopped. As a result, no cache device or read-cache enabled ones are displayed on etcadmin command outputs. Example 4-12 Validating Easy Tier Server coherency client status after etcadmin -a shutdown # lsdev -C | grep etc # lssrc -s etcdaemon Subsystem Group etcdaemon PID Status inoperative # etcadmin -a list Fail the call since etcdd device is not configured # etcadmin -a query No read_cache enabled SAN device has been found. When the client driver is not operational anymore, the smit remove AIX command can be used to have the bos.etcacheDD.rte fileset removed off the operating system. Figure 4-2 on page 65 displays the AIX SMIT menu that is opened for removing the fileset. 64 IBM System Storage DS8000 Easy Tier Server Remove Installed Software Type or select values in entry fields. Press Enter AFTER making all desired changes. * SOFTWARE name PREVIEW only? (remove operation will NOT occur) REMOVE dependent software? EXTEND file systems if space needed? DETAILED output? [Entry Fields] [bos.etcacheDD.rte] no no no no + + + + + Figure 4-2 SMIT menu for fileset removal After entering the fileset name and pressing the Enter key, AIX removes the fileset. If everything runs fine, an OK message is returned. Otherwise, if the removal status displayed is Failed, either scroll through the logs or check the smit.log file to identify the root cause of the failure. Note: Contact your AIX support representative for more assistance on handling fileset installation or removal failures. Upon fileset removal, the lslpp command will not find that fileset any longer and the etcadmin command is not recognized by the operating system either, as Example 4-13 demonstrates. Example 4-13 Validating bos.etcacheDD.rte fileset removal from the operating system # lslpp -l bos.etcacheDD.rte lslpp: Fileset bos.etcacheDD.rte not installed. # etcadmin ksh: etcadmin: not found 4.3 Upgrading Easy Tier Server coherency client driver An upgrade of the Easy Tier Server coherency client driver consists of uninstalling the current version and installing the new one. Therefore, before upgrading it, you must stop all read-cache enabled DS8870 hdisks and remove the client etcdd device. Despite that the current driver must be uninstalled, the Easy Tier Server configuration for cache devices and read-cache enabled DS8870 hdisks can be saved. You just need to issue the etcadmin -a shutdown -s yes|cacdev|rdcac command to save either the configuration of the cache devices, of the enabled read cache devices, or both. The saved client configuration is automatically restored after the new driver is installed, during the etcdd device configuration on cfgmgr. When the new package is installed, you can either reboot the system or issue the cfgmgr command. Chapter 4. Easy Tier Server implementation 65 In the next examples, we cover all the upgrade procedures, starting on checking the current configuration for both cache devices and read-cache enabled DS8870 hdisks, as shown in Example 4-14. Example 4-14 Querying the current Easy Tier Server coherency client configuration # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | # etcadmin -a query Total read cache enabled SAN devices is 8 ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 The etcadmin -a shutdown command takes care of stopping all read-cache enabled DS8870 hdisks and of removing the client etcdd devices. The -s yes flag was used to save the current configuration of both the cache devices and read-cache enabled DS8870 hdisks, as shown in Example 4-15. Example 4-15 Shutting down Easy Tier Server coherency client driver # etcadmin -a shutdown -s yes Total read cache enabled SAN devices is 8 -------------------------------------------SAN device hdisk16 read_cache function is disabled SAN device hdisk17 read_cache function is disabled SAN device hdisk18 read_cache function is disabled SAN device hdisk19 read_cache function is disabled SAN device hdisk20 read_cache function is disabled SAN device hdisk21 read_cache function is disabled SAN device hdisk22 read_cache function is disabled SAN device hdisk23 read_cache function is disabled Removing etcdd device... etcdd device is deleted successfully Because the Easy Tier Server coherency client has been shut down while saving configuration data, the uninstallation procedures described in 4.2, “Uninstalling Easy Tier server coherency client driver” on page 64 can be followed to remove the current version of the bos.etcacheDD.rte fileset from the operating system. 66 IBM System Storage DS8000 Easy Tier Server Next, with the new client driver available, refer to “Installing the Easy Tier Server coherency client driver” on page 59 for procedures for the driver installation. The cfgmgr AIX command loads the driver’s kernel extensions, configures the etcdd device, and automatically starts etcdaemon, as demonstrated in Example 4-16. If you properly used the etcadmin -a shutdown -s yes|cacdev|rdcac command to shut down the client driver while saving some configuration, as we did in Example 4-15 on page 66, cfgmgr will also restore the saved configuration. Attention: Do not issue the etcadmin -a cfgetcdd -r command if you are migrating or using the Easy Tier Server client package. It will not load the saved configuration when the driver was shut down. Example 4-16 Restoring client driver configuration after an upgrade # cfgmgr # lsdev -C | grep etc etcdd Available # lssrc -s etcdaemon Subsystem Group etcdaemon Easy Tier Cache Parent PID 2293776 Status active # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | # etcadmin -a query Total read cache enabled SAN devices is 19 ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 Chapter 4. Easy Tier Server implementation 67 68 IBM System Storage DS8000 Easy Tier Server 5 Chapter 5. Managing Easy Tier Server This chapter describes how to properly manage and maintain an Easy Tier Server environment after it is operational. © Copyright IBM Corp. 2013. All rights reserved. 69 5.1 Managing Easy Tier Server Easy Tier Server coherency client functions are managed by the etcadmin command, which is installed with the bos.etcacheDD.rte fileset. In the next sections of this chapter, we cover several management tasks enabled by the etcadmin command. 5.1.1 Easy Tier Server coherency client management tasks As a starting point, Figure 5-1 lists the etcadmin command usage and its actions and options. # etcadmin USAGE: etcadmin -a <action> [-d <devName>] -a create -d </dev/devName> -a destroy -a add -d </dev/devName> -a start -d <devName> -a stop -d <devName> -a start -D <devName1> <devName2> -a stop -D <devName1> <devName2> -a stopall -a query -a list -a iostat -d <devName> -a shutdown -s <yes|cacdev|rdcac|no> -a cfgetcdd -r <yes|cacdev|rdcac|no> Figure 5-1 etcadmin command usage Starting Easy Tier Server coherency client driver The Easy Tier Server coherency client driver can be started using three different methods: AIX cfgmgr command, AIX restart, or etcadmin -a cfgetcdd. In general, the cfgmgr command is used to initially start the client driver, just after bos.etcacheDD.rte installation, for instance. See Example 5-1. Example 5-1 AIX cfgmgr usage for starting Easy Tier Server coherency client driver # cfgmgr # lsdev -Ct etcdd etcdd Available Easy Tier Cache Parent # lssrc -s etcdaemon Subsystem Group etcdaemon PID 1836592 Status active # etcadmin -a list No cache device has been created # etcadmin -a query No read_cache enabled SAN device has been found. 70 IBM System Storage DS8000 Easy Tier Server An operating system restart operation works exactly like the cfgmgr command, in terms of starting the client driver, when the cfgmgr command runs upon AIX startup. Creating a cache directory and adding DAS cache devices The client cache directory with cache devices is a key component of the Easy Tier Server feature. With the etcadmin command, you can designate which direct-attached storage (DAS) solid-state drive (SSD) devices are used as cache devices. First, you might want to list the available SDD devices on the client, using the lsdev command. Then, if there is no cache device designated for this client yet, the etcadmin -a create command must be used in order to create the cache directory. Later, other SSD devices are added to the cache directory as cache devices by the etcadmin -a add command. Example 5-2 shows the usage of both the etcadmin -a create and etcadmin -a add commands to initially create a caching directory and then to add other cache devices. It also shows the etcadmin -a list command that is used to check the newly created cache devices for this Easy Tier Server coherency client. Note: When referring to DAS SSD cache devices on the etcadmin command, you must use the prefix /dev/ in the hdisk name parameter. By contrast, when using the etcadmin command to manage the DS8870 hdisks read cache function, you must use only the hdisk name, without the /dev/ prefix. Example 5-2 Creating or adding DAS cache devices # lsdev -Cc disk | grep SSD hdisk4 Available 0N-00-00 SAS RAID 5 SSD Array hdisk5 Available 0N-00-00 SAS RAID 5 SSD Array # etcadmin -a create -d /dev/hdisk4 Cache device has been created with Flash Cache /dev/hdisk4 # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | # etcadmin -a add -d /dev/hdisk5 Flash Cache device /dev/hdisk5 has been added to cache device # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | Tip: The etcadmin command will not allow you to designate a non-SSD device as a cache device. If you use a DS8870 hdisk in the etcadmin -a add or etcadmin -a create commands, the following error is displayed: hdiskX is not a valid Flash Cache device. Chapter 5. Managing Easy Tier Server 71 Managing and enabling the read cache function on DS8870 hdisks DS8870 hdisk read caching does not start automatically after installing the cache device. You still need to define which DS8870 hdisk is to benefit from the Easy Tier Server feature. Across a given set of DS8870 hdisks mapped to this host, you might want to just select a limited group of logical unit numbers (LUNs) to be read-cache enabled, or you can enable them all. In environments where a single host uses LUNs from different storage systems, ensure that you select only for read cache LUNs from a DS8870 with the Easy Tier Server feature enabled. An Easy Tier Server coherency client can be served by as many Easy Tier Server coherency servers as you would like. It is just a matter of enabling the Easy Tier Server feature on the DS8870 systems that map LUNs to the client. The opposite is partially true: a DS8870 can act as an Easy Tier Server coherency server for at most 16 clients. In Example 5-3, we list the DS8870 LUNs recognized by the host. Referring to this list, the etcadmin -a start command is used to start the read cache function for those hdisks. Note: An AIX host with the Easy Tier Server coherency client driver enabled is capable of using LUNs from different DS8870 storage systems, whether they have the Easy Tier Server feature enabled or not. If all DS8870 storage systems that map LUNs to this AIX host are Easy Tier Server-enabled, the DAS cache in the client is shared by all of the Easy Tier Server coherency servers whose LUNs have been read-cache enabled on the client. Example 5-3 Enabling the read cache function on DS8870 hdisks # lsdev hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 -Cc disk | grep 2107 Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-01-02 IBM Available 0k-00-02 IBM MPIO MPIO MPIO MPIO MPIO MPIO MPIO MPIO FC FC FC FC FC FC FC FC 2107 2107 2107 2107 2107 2107 2107 2107 # etcadmin -a start -D hdisk16 hdisk21 SAN device hdisk16 read_cache function SAN device hdisk17 read_cache function SAN device hdisk18 read_cache function SAN device hdisk19 read_cache function SAN device hdisk20 read_cache function SAN device hdisk21 read_cache function is is is is is is enabled enabled enabled enabled enabled enabled # etcadmin -a start -d hdisk22 SAN device hdisk22 read_cache function is enabled # etcadmin -a start -d hdisk23 SAN device hdisk23 read_cache function is enabled # etcadmin -a query Total read cache enabled SAN devices is 8 72 IBM System Storage DS8000 Easy Tier Server ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 With the etcadmin -a start command, you can specify either a contiguous range of hdisks, by using the -D flag, or single hdisks to have the read cache function started, by using -d. Note: It will take up to 15 minutes for the read cache function to start for a just-enabled DS8870 hdisk. In these first minutes, the Easy Tier Server coherency client and server will be exchanging advices and learning about the data pattern access on the LUNs. After the devices were read cache enabled by the etcadmin -a start command, there is no user intervention required during this first learning cycle. As soon as the read caching function is effectively started, an asterisk is displayed by the side of each hdisk on the output of the etcadmin -a query command. Disabling the read cache function on DS8870 hdisks At any time, the Easy Tier Server coherency client driver allows you to stop the read cache function for either a single DS8870 hdisk, a contiguous set of these hdisks, or for all of the currently read-cache enabled hdisks. Refer to Example 5-4 for illustration. Example 5-4 Disabling the read cache function on DS8870 hdisks # etcadmin -a stop -d hdisk16 SAN device hdisk16 read_cache function is disabled # etcadmin -a stop -D hdisk17 hdisk20 SAN device hdisk17 read_cache function SAN device hdisk18 read_cache function SAN device hdisk19 read_cache function SAN device hdisk20 read_cache function is is is is disabled disabled disabled disabled # etcadmin -a stopall Total read cache enabled SAN devices is 3 -------------------------------------------SAN device hdisk21 read_cache function is disabled SAN device hdisk22 read_cache function is disabled SAN device hdisk23 read_cache function is disabled # etcadmin -a query No read_cache enabled SAN device has been found. # etcadmin -a list Chapter 5. Managing Easy Tier Server 73 ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | Even if the read cache function is stopped for all read-cache enabled hdisks, the cache devices are expected to remain active. Shutting down the Easy Tier Server coherency client driver The etcadmin -a shutdown command allows you to stop both the read cache function for enabled DS8870 hdisks and the DAS cache devices. Using the -s flag and depending on the parameter specified allows you to either save the cache devices configuration, read-cache enabled devices configuration, both configurations, or none before a shutdown. The respective parameters with the -s flag are as follows: yes: Saves both the client cache devices and read-cache enabled DS8870 hdisk configurations before shutdown. cacdev: Saves only the client cache devices configuration before shutdown. rdcac: Saves only the client read cache enabled DS8870 hdisks configuration before shutdown. no: Does not save any configuration before shutdown. Example 5-5 demonstrates the etcadmin -a shutdown command with the -s flag yes parameter to save all Easy Tier Server coherency client configuration before the driver is shut down. In this example following the etcadmin -a shutdown command, the etcdd device is unconfigured from AIX and the etcdaemon is stopped. Example 5-5 Shutting down the Easy Tier Server coherency client driver while saving all configuration # etcadmin -a shutdown -s yes No rd_cache enabled SAN device found. Removing etcdd device... etcdd device is deleted successfully # lsdev -Ct etcdd # # lssrc -s etcdaemon Subsystem Group etcdaemon PID Status inoperative # etcadmin -a list Fail the call since etcdd device is not configured # etcadmin -a query No read_cache enabled SAN device has been found. Important: Upon a restart of the AIX operating system as an Easy Tier Server coherency client, there is no need to perform any additional operations on the client driver. It will automatically stop the read cache function for enabled devices and the cache devices before AIX shuts down, and upon AIX boot, the cfgmgr command will automatically recover the configuration. 74 IBM System Storage DS8000 Easy Tier Server Restoring a saved configuration and restarting the client driver If the client driver was shut down by etcadmin -a shutdown -s no, having no configuration saved, a cfgmgr command is the best way to restart the driver. Instead, if any configuration (yes, cacdev, or rdcac) was saved, the etcadmin -a cfgetcdd command must be used with the appropriate -r flag parameter. The -r flag parameter of the etcadmin -a cfgetcdd command matches the etcadmin -a shutdown -s ones: yes, cacdev, rdcac, or no. In Example 5-6, we restart the client driver restoring all the saved configuration. Example 5-6 Restoring Easy Tier Server coherency client saved configuration and restarting the driver # etcadmin -a cfgetcdd -r yes etcdd Configuring etcdd device... etcdd device is configured to Available state # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | Tip: If you saved all configuration using the etcadmin -a shutdown command, by using -s yes, you can also choose among recovering all configuration (yes), cached devices configuration (cacdev), read-cache enabled devices configuration (rdcac), or no configuration (no) on the etcadmin -a cfgetcdd command. In Example 5-5 on page 74, when the etcadmin -a shutdown command was run, there were no read-cache enabled devices as the output of the command displayed the informational message: No rd_cache enabled SAN device found. Therefore, although we use etcadmin -a cfgetcdd -r yes to restore all saved configuration, no read-cache enabled device is restored, as expected. Attention: The AIX cfgmgr command will not restore the saved configuration after the client driver is shut down by etcadmin -a shutdown -s yes|cacdev|rdcac. If you want the configuration to be restored, the etcadmin -a cfgetcdd command must be used. If cfgmgr is run after the etcadmin -a shutdown -s yes|cacdev|rdcac execution, it starts the client driver properly, but it deletes the saved data, as shown in Example 5-7. Example 5-7 Saving configuration data before driver shutdown and using cfgmgr inappropriately # etcadmin -a shutdown -s yes No rd_cache enabled SAN device found. Removing etcdd device... etcdd device is deleted successfully # cfgmgr # lsdev -Ct etcdd etcdd Available Easy Tier Cache Parent # etcadmin -a list Chapter 5. Managing Easy Tier Server 75 No cache device has been created There is an exception to this rule, in which cfgmgr will load the saved configuration. This exception applies for Easy Tier Server coherency client driver upgrades and migrations. Refer to Chapter 4.3, “Upgrading Easy Tier Server coherency client driver” on page 65 for more information. Destroying DAS cache devices If the goal is to reformat the existing cache devices, you can use the etcadmin -a destroy command to remove all cache devices at one time, as shown in Example 5-8. Example 5-8 Destroying DAS cache devices and querying etcadmin outputs # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | # etcadmin -a destroy Cache device has been destroyed DAS device hdisk4 cache device is set to non-Active DAS device hdisk5 cache device is set to non-Active # etcadmin -a list No cache device has been created # etcadmin -a query Total read cache enabled SAN devices is 8 ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 If read-cache enabled DS8870 hdisks are defined, they remain defined, waiting for new cache devices to be created. 76 IBM System Storage DS8000 Easy Tier Server 5.1.2 Managing and configuring direct-attached storage As stated in “Configuring direct-attached storage flash (SSD) devices” on page 61, Redundant Array of Independent Disks 5 (RAID 5) is the recommended configuration for Easy Tier Server coherency client cache devices. When a supported expansion enclosure is attached to the host, each SSD disk is formatted as a single RAID 0 array hdisk. For best practices, you might want to destroy the current RAID 0 SSD arrays’ hdisks and create new RAID 5 ones. Refer to Example 5-9. First, we list the current SSD hdisks (arrays) available on AIX, using the lsdev command. Next, the sissasraidmgr AIX command is used for listing the current array formatting and destroying these arrays. Attention: Before reformatting SSD arrays that are already defined and currently in use as cache devices, shut down the client driver or destroy the cache directory using the etcadmin command. Example 5-9 Using the sissasraidmgr AIX command to list and delete RAID arrays configuration # lsdev hdisk4 hdisk5 hdisk6 hdisk7 hdisk8 hdisk9 hdisk10 hdisk11 -Cc disk | grep SSD Available 0M-00-00 SAS Available 0M-00-00 SAS Available 0M-00-00 SAS Available 0M-00-00 SAS Available 0M-00-00 SAS Available 0M-00-00 SAS Available 0M-00-00 SAS Available 0M-00-00 SAS RAID RAID RAID RAID RAID RAID RAID RAID 0 0 0 0 0 0 0 0 SSD SSD SSD SSD SSD SSD SSD SSD Array Array Array Array Array Array Array Array # sissasraidmgr -L -j1 -l sissas7 -----------------------------------------------------------------------Name Resource State Description Size -----------------------------------------------------------------------sissas7 FEFFFFFF Primary PCIe2 3.1GB Cache RAID SAS Enclosure 6Gb x8 sissas3 FEFFFFFF HA Linked Remote adapter SN 002B9004 hdisk4 pdisk0 FC0000FF 000000FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk5 pdisk2 FC0100FF 000002FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk6 pdisk3 FC0200FF 000003FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk7 pdisk4 FC0300FF 000004FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk8 pdisk5 FC0400FF 000005FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk9 pdisk6 FC0500FF 000006FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB Chapter 5. Managing Easy Tier Server 77 hdisk10 pdisk7 FC0600FF 000007FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB hdisk11 pdisk1 FC0700FF 000001FF Optimal Active RAID 0 Array SSD Array Member 387.9GB 387.9GB # sissasraidmgr -D -l sissas7 -d hdisk4 hdisk4 deleted pdisk0 Defined # sissasraidmgr -D -l sissas7 -d hdisk5 hdisk5 deleted pdisk2 Defined # sissasraidmgr -D -l sissas7 -d hdisk6 hdisk6 deleted pdisk3 Defined # sissasraidmgr -D -l sissas7 -d hdisk7 hdisk7 deleted pdisk4 Defined # sissasraidmgr -D -l sissas7 -d hdisk8 hdisk8 deleted pdisk5 Defined # sissasraidmgr -D -l sissas7 -d hdisk9 hdisk9 deleted pdisk6 Defined # sissasraidmgr -D -l sissas7 -d hdisk10 hdisk10 deleted pdisk7 Defined # sissasraidmgr -D -l sissas7 -d hdisk11 hdisk11 deleted pdisk1 Defined # sissasraidmgr -L -j1 -l sissas7 -----------------------------------------------------------------------Name Resource State Description Size -----------------------------------------------------------------------sissas7 FEFFFFFF Primary PCIe2 3.1GB Cache RAID SAS Enclosure 6Gb x8 sissas3 FEFFFFFF HA Linked Remote adapter SN 002B9004 pdisk0 pdisk4 pdisk7 pdisk2 pdisk3 pdisk6 pdisk1 pdisk5 78 000000FF 000004FF 000007FF 000002FF 000003FF 000006FF 000001FF 000005FF Active Active Active Active Active Active Active Active IBM System Storage DS8000 Easy Tier Server SSD SSD SSD SSD SSD SSD SSD SSD Array Array Array Array Array Array Array Array Candidate Candidate Candidate Candidate Candidate Candidate Candidate Candidate 387.9GB 387.9GB 387.9GB 387.9GB 387.9GB 387.9GB 387.9GB 387.9GB After the sissasraidmgr command has destroyed the previously defined RAID arrays, each SSD device is represented only as a pdisk, with no associated hdisk. The pdisks are then SSD Array Candidates and they are ready to be assigned to new RAID arrays. The sissasraidmgr command is used with the appropriate flags to create new arrays. In Example 5-10, we create two RAID 5 arrays, each with four SSD devices. Example 5-10 Using the sissasraidmgr AIX command to create new SSD arrays # sissasraidmgr -C -r 5 -s 256 -z 'pdisk0 pdisk1 pdisk2 pdisk3' . # sissasraidmgr -C -r 5 -s 256 -z 'pdisk4 pdisk5 pdisk6 pdisk7' . # sissasraidmgr -L -j1 -l sissas7 -----------------------------------------------------------------------Name Resource State Description Size -----------------------------------------------------------------------sissas7 FEFFFFFF Primary PCIe2 3.1GB Cache RAID SAS Enclosure 6Gb x8 sissas3 FEFFFFFF HA Linked Remote adapter SN 002B9004 hdisk4 pdisk0 pdisk2 pdisk3 pdisk1 FC0000FF 000000FF 000002FF 000003FF 000001FF Rebuilding Active Active Active Active RAID 5 Array SSD Array Member SSD Array Member SSD Array Member SSD Array Member 1163GB Create 25% 387.9GB 387.9GB 387.9GB 387.9GB hdisk5 pdisk4 pdisk7 pdisk6 pdisk5 FC0100FF 000004FF 000007FF 000006FF 000005FF Rebuilding Active Active Active Active RAID 5 Array SSD Array Member SSD Array Member SSD Array Member SSD Array Member 1163GB Create 25% 387.9GB 387.9GB 387.9GB 387.9GB # lsdev -Cc disk | grep SSD hdisk4 Available 0N-00-00 SAS RAID 5 SSD Array hdisk5 Available 0N-00-00 SAS RAID 5 SSD Array Just after launching the command, the RAID array status is Rebuilding. It takes a few minutes for the array creation to complete, depending on the number of drives contained by the arrays and the drive sizes. When the rebuild operation completes, the array status changes to Optimal, as shown in Example 5-11. Example 5-11 Using the sissasraidmgr AIX command to list RAID arrays configuration and their status # sissasraidmgr -L -j1 -l sissas7 -----------------------------------------------------------------------Name Resource State Description Size -----------------------------------------------------------------------sissas7 FEFFFFFF Primary PCIe2 3.1GB Cache RAID SAS Enclosure 6Gb x8 sissas3 FEFFFFFF HA Linked Remote adapter SN 002B9004 hdisk4 pdisk0 pdisk2 pdisk3 FC0000FF 000000FF 000002FF 000003FF Optimal Active Active Active RAID 5 Array SSD Array Member SSD Array Member SSD Array Member 1163GB 387.9GB 387.9GB 387.9GB Chapter 5. Managing Easy Tier Server 79 pdisk1 000001FF Active SSD Array Member 387.9GB hdisk5 pdisk4 pdisk7 pdisk6 pdisk5 FC0100FF 000004FF 000007FF 000006FF 000005FF Optimal Active Active Active Active RAID 5 Array SSD Array Member SSD Array Member SSD Array Member SSD Array Member 1163GB 387.9GB 387.9GB 387.9GB 387.9GB At this point, the SSD RAID arrays, represented by hdisks, are available for use as Easy Tier Server coherency client cache devices. In Example 5-12, the newly created arrays are displayed by the lsdev command and then, the etcadmin command is used to create the cache devices. Example 5-12 Creating SSD DAS cache directory and enabling read cache function for DS8870 hdisks # lsdev -Cc disk | grep SSD hdisk4 Available 0N-00-00 SAS RAID 5 SSD Array hdisk5 Available 0N-00-00 SAS RAID 5 SSD Array # etcadmin -a create -d /dev/hdisk4 Cache device has been created with Flash Cache /dev/hdisk4 # etcadmin -a add -d /dev/hdisk5 Flash Cache device /dev/hdisk5 has been added to cache device # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 1163869028352 | /dev/hdisk5 | 1163869028352 | # etcadmin -a query No read_cache enabled SAN device has been found. # etcadmin -a start -D hdisk16 hdisk23 SAN device hdisk16 read_cache function SAN device hdisk17 read_cache function SAN device hdisk18 read_cache function SAN device hdisk19 read_cache function SAN device hdisk20 read_cache function SAN device hdisk21 read_cache function SAN device hdisk22 read_cache function SAN device hdisk23 read_cache function is is is is is is is is enabled enabled enabled enabled enabled enabled enabled enabled # etcadmin -a query Total read cache enabled SAN devices is 8 ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== *hdisk16 *hdisk17 80 IBM System Storage DS8000 Easy Tier Server *hdisk18 *hdisk19 *hdisk20 *hdisk21 *hdisk22 *hdisk23 For more information about the sissasraidmgr command, see the IBM AIX 7.1 Information Center: http://pic.dhe.ibm.com/infocenter/aix/v7r1/index.jsp Chapter 5. Managing Easy Tier Server 81 82 IBM System Storage DS8000 Easy Tier Server 6 Chapter 6. Easy Tier Server monitoring This chapter covers the Easy Tier Server monitoring in terms of overall workload and distribution of read I/Os, either sent to direct-attached storage (DAS) solid-state drive (SSD) cache devices or the DS8870. We discuss the following tools that can be used for monitoring Easy Tier Server operations: Storage Tier Advisor Tool (STAT) AIX iostat command Easy Tier Server coherency client etcadmin command © Copyright IBM Corp. 2013. All rights reserved. 83 6.1 Monitoring Easy Tier Server In general, the performance improvements generated by Easy Tier Server operation are mostly observed at the application layer, while monitoring the response time from the application perspective. In addition, the DS8870 Storage System, AIX operating system, and Easy Tier Server etcadmin command offer specific tools for monitoring DAS cache hits, overall input/output operations per second (IOPS) performance, and general workload characteristics. The results shown by these monitoring tools along with the application’s performance metrics give a comprehensive understanding of Easy Tier Server operation and the performance improvements that this feature can use. 6.1.1 DS8870 Storage Tier Advisor Tool The Storage Tier Advisor Tool (STAT) is a Windows application that can be used to analyze the characteristics of the workload running on DS8870 storage systems. It provides capacity planning information associated with the current or future use of the Easy Tier. The STAT processes data that is collected by Easy Tier and Easy Tier Server monitors. Monitoring statistics are gathered and analyzed at least every 24 hours. The results of this data are integrated in a summary report data that can be downloaded from the DS8870 storage system for reporting purposes using the STAT. Note: STAT version 7.7.10.xxx enhancements enable users to monitor Easy Tier Server coherency clients’ IOPS statistics, regarding overall read IOPS and DAS cache hit ratio. STAT can be downloaded from the following File Transfer Protocol (FTP) server address: ftp://ftp.software.ibm.com/storage/ds8000/updates/DS8K_Customer_Download_Files/Sto rage_Tier_Advisor_Tool/ Refer to IBM System Storage DS8000 Easy Tier, REDP-4667 for more information about the Storage Tier Advisor Tool concepts and usage. DS8870 Easy Tier summary report The collected Easy Tier data can be downloaded at any time from the DS8870, by using the offloadfile command. The offloadfile command creates two output files, one file for each DS8870 internal server. These output files consist of two compressed files, which are known as heat data files, and are processed by the STAT tools to generate a graphical report that can be viewed in a web browser. You can also use the data storage graphical user interface (DS GUI) for exporting DS8870 Easy Tier Summary Report by clicking Home System Status Right-Click the Storage Image Storage Image Export Easy Tier Summary Report, as illustrated by Figure 6-1 on page 85. Decompress the heat data files generated by DS8870 and use them as input files for STAT.exe, using the following syntax: >STAT.exe inputfile1 inputfile2 The index.html file in the STAT installation directory is updated with new heat data. 84 IBM System Storage DS8000 Easy Tier Server Figure 6-1 Exporting Easy Tier Summary Report via DS GUI Easy Tier Server overview in STAT The Storage Tier Advisor Tool (STAT) System Summary Report provides an overview of Easy Tier monitored DS8870 extent pools. It also shows all Easy Tier Server coherency clients managed by this server, as highlighted in Figure 6-2. In this view, STAT displays the average read IOPS and DAS Cache Hit Ratio for a given client in the last 24-hours data collection time frame. It also shows the Configured DAS Cache Size for each client. Important: The output of the STAT is based on data collected by the Easy Tier monitoring function and a migration plan is only created after a data collection period of at least 24 hours. Easy Tier Server information about STAT is not immediately reclaimed by STAT data collection. It will be available in the next report cycle after Easy Tier Server deployment on the Easy Tier Server coherency clients. Figure 6-2 STAT System Summary view Chapter 6. Easy Tier Server monitoring 85 Detailed information for each Easy Tier Server coherency client is displayed by clicking the client’s correspondent worldwide port name (WWPN), in the HOST ID column. The view represented by Figure 6-3 shows a comprehensive graph containing the overall read IOPS for one client over the time. Note: The IOPS count brought by STAT for Easy Tier Server coherency clients is the average of overall read IOPS for all DS8870 LUNs in a given client, considering both DAS read IOPS and DS8870 read IOPS. Cache Hit Ratio is the percentage of the read IOPS responded by DAS cache. Write I/Os are not taken into account in these summaries for Easy Tier Server. Figure 6-3 STAT detailed view for a given Easy Tier Server coherency client Still, STAT breaks down the IOPS and Cached Hit Ratio statistics in the LUN level. Hence, it allows you to see the individual results for all DS8870 LUNs mapped to this particular client, as shown in Figure 6-4. To get the detailed breakdown per-LUN statistics, expand the Existing Attached Volume Status option. Figure 6-4 STAT breakdown per-LUN statistics for a given Easy Tier Server coherency client Storage Tier Advisor Tool provides several other monitoring views for Easy Tier internal data placement functions. IBM System Storage DS8000 Easy Tier, REDP-4667 can be consulted for more information. 86 IBM System Storage DS8000 Easy Tier Server 6.1.2 AIX operating system IOSTAT Tool The iostat AIX command is used to monitor the system I/O devices’ statistics from the operating system perspective. Besides monitoring the DS8870 hdisks, iostat can also provide I/O statistics for the SSD arrays’ hdisks that are designated as cache devices. By monitoring iostat outputs, we can see a breakdown, per hdisk (whether it is a DS8870 LUN or an SSD array) on reads and writes IOPS. In a measurement interval, iostat displays hdisks’ utilization, amount of IOPS, throughput, response time, and more relevant information. Using iostat flags -D and -l shows a long listed extended utilization report, continuously updating the performance information at one-second intervals. Use the commands as follows: # iostat -Dl 1 Example 6-1 illustrates the monitoring scenario with iostat. In this scenario, there are eight DS8870 LUNs, represented by hdisk16 to hdisk23. All of them are read-cache enabled by Easy Tier Server coherency client cache device, which is a single 8-SSD device RAID 5 array, represented by hdisk4. Example 6-1 Listing the DAS cache devices and read cache enabled DS8870 hdisks # lsdev hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 hdisk23 -Cc disk | grep 2107 Available 0E-00-02 IBM Available 0E-00-02 IBM Available 0E-00-02 IBM Available 0E-00-02 IBM Available 0E-00-02 IBM Available 0E-00-02 IBM Available 0E-00-02 IBM Available 0E-00-02 IBM MPIO MPIO MPIO MPIO MPIO MPIO MPIO MPIO FC FC FC FC FC FC FC FC 2107 2107 2107 2107 2107 2107 2107 2107 # lsdev -Cc disk | grep SSD hdisk4 Available 0N-00-00 SAS RAID 0 SSD Array # etcadmin -a list ------------------------------------------------------------------------------Device Name | Caching Capacity (bytes) | ------------------------------------------------------------------------------/dev/hdisk4 | 2715694399488 | # etcadmin -a query Total read cache enabled SAN devices is 8 ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== hdisk16 hdisk17 hdisk18 hdisk19 hdisk20 hdisk21 hdisk22 Chapter 6. Easy Tier Server monitoring 87 hdisk23 After starting a random online transaction processing (OLTP)-like workload, we collected initial iostat output, as shown in Figure 6-5. This first iostat output was captured during the initial 15-minute interval, immediately after enabling the read cache function on the DS8870 hdisks. At this time, the Easy Tier Server coherency client and server are learning about the data patterns and exchanging advices to effectively start the caching mechanisms. Therefore, we can see that all read I/Os were still being handled by the DS8870 hdisks, while there was no I/O at all to hdisk4, which is the cache device for this client. Disks: xfers read write -------------- -------------------------------- ------------------------------------ -----------------------------------%tm bps tps bread bwrtn rps avg min max time fail wps avg min max time fail act serv serv serv outs serv serv serv outs hdisk18 96.0 24.8M 898.6 20.5M 4.3M 722.1 1.3 0.1 518.9 0 0 176.5 0.3 0.2 0.5 0 0 hdisk16 93.4 21.7M 923.6 17.7M 3.9M 723.4 1.4 0.1 515.7 0 0 200.2 0.3 0.2 0.5 0 0 hdisk21 96.5 28.8M 1425.3 23.5M 5.3M 1104.2 1.3 0.1 486.3 0 0 321.1 0.3 0.2 5.2 0 0 hdisk17 96.2 21.8M 1203.7 17.5M 4.3M 909.3 1.3 0.1 507.4 0 0 294.4 0.2 0.2 8.3 0 0 hdisk22 93.0 30.7M 1069.1 25.8M 5.0M 879.9 1.2 0.1 507.5 0 0 189.2 0.3 0.2 0.5 0 0 hdisk23 100.0 28.3M 1136.2 23.4M 5.0M 906.6 1.2 0.1 504.5 0 0 229.6 0.3 0.2 0.5 0 0 hdisk19 100.0 20.9M 909.0 17.0M 3.9M 706.2 1.5 0.1 517.5 0 0 202.8 0.3 0.2 33.5 0 0 hdisk20 97.8 26.2M 1073.1 21.4M 4.8M 842.1 1.4 0.1 149.5 0 0 231.0 0.3 0.2 0.5 0 0 hdisk4 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 0.0 0.0 0.0 0.0 0 0 Figure 6-5 Iostat output collected when the read cache function on DS8870 hdisks had not effectively started yet After the initial 15-minute learning time, we can see with the etcadmin -a query command output that the read cache function has been effectively started on the read-cache enabled devices. Now, these devices have asterisks by their sides, as displayed in Example 6-2. Example 6-2 Listing read-cache enabled DS8870 hdisks when read_cache is started # etcadmin -a query Total read cache enabled SAN devices is 8 ------------------------------------------'*': indicates SAN device read_cache has started ' ': indicates SAN device read_cache has not started '~': indicates failing to get SAN device read_cache status Device Name =========== *hdisk16 *hdisk17 *hdisk18 *hdisk19 *hdisk20 *hdisk21 *hdisk22 *hdisk23 88 IBM System Storage DS8000 Easy Tier Server Another iostat output was collected and you can see the amount of read IOPS being served by the cache device to the application in Figure 6-6. As most of the read I/Os requested by the application were hitting the DAS cache, cache device read throughput and IOPS contrast greatly with these same statistics for the read-cache enabled DS8870 hdisks. For instance, while the read IOPS on the DS8870 hdisks were in the 200 range, the read IOPS for the cache device were above 5000. As expected, we can see the same behavior for the overall throughput. The average DS8870 hdisks throughput was around 60 Mbps, while the cache device’s was higher than 500 Mbps. Disks: xfers read write -------------- -------------------------------- ------------------------------------ -----------------------------------%tm bps tps bread bwrtn rps avg min max time fail wps avg min max time fail act serv serv serv outs serv serv serv outs hdisk18 74.3 39.6M 292.3 34.6M 5.1M 137.0 8.6 0.1 320.1 0 0 155.3 0.3 0.2 0.6 0 0 hdisk16 94.7 57.1M 295.6 53.9M 3.3M 196.4 9.9 0.1 237.7 0 0 99.2 0.3 0.2 0.6 0 0 hdisk21 83.9 87.2M 486.6 82.5M 4.6M 345.2 3.8 0.1 101.6 0 0 141.4 0.3 0.2 4.8 0 0 hdisk17 97.5 60.8M 312.9 58.8M 2.1M 249.3 9.2 0.1 570.1 0 0 63.6 0.3 0.2 0.6 0 0 hdisk22 46.7 53.1M 385.9 45.1M 8.0M 142.1 3.6 0.1 52.8 0 0 243.8 0.3 0.2 0.6 0 0 hdisk23 100.0 52.0M 365.8 45.1M 6.9M 156.1 4.2 0.1 72.7 0 0 209.7 0.3 0.2 0.6 0 0 hdisk19 100.0 55.4M 297.3 52.1M 3.3M 197.0 9.4 0.1 369.7 0 0 100.3 0.3 0.2 0.6 0 0 hdisk20 60.2 54.4M 380.0 48.0M 6.4M 184.3 4.0 0.1 100.7 0 0 195.7 0.3 0.2 0.6 0 0 hdisk4 98.5 547.4M 6735.4 168.6M 378.8M 5290.5 0.6 0.0 42.8 0 0 1444.9 1.9 1.3 8.3 0 0 Figure 6-6 Iostat output displaying the I/O behavior in a Easy Tier Server coherency client These results demonstrate that Easy Tier Server cache devices are effectively serving most of the application’s read requests, at the same time, they are fetching data from DS8870 to populate the cache. Although the Easy Tier Server coherency client cache device is a read-only cache for the application, you might notice some amount of write IOPS. These write IOPS reflect the caching population logic of Easy Tier Server, in which data is read from DS8870 and written to the cache devices. Attention: With Easy Tier Server enabled, the read service times of DS8870 hdisks do not represent the real response time observed at the application layer. This is because most of the read requests made for read-cache enabled DS8870 hdisks are directly handled by the cache devices, whose average service time is in the microseconds range. The high number of read transactions on the read-cache enabled DS8870 hdisks are for populating the DAS cache. Upon population, there will be read I/Os to DS8870 LUNs and write I/Os to the cache device. For more information about each iostat statistic, consult IBM AIX 7.1 Information Center at the following site: http://pic.dhe.ibm.com/infocenter/aix/v7r1/index.jsp 6.1.3 Monitoring with etcadmin The Easy Tier Server coherency client management tool etcadmin can be used to monitor read-cache enabled DS8870 hdisks, using the etcadmin -a iostat command. It displays I/O statistics, classifying read and write I/Os to DS8870, and read I/Os to the DAS cache device. Chapter 6. Easy Tier Server monitoring 89 The output shown in Example 6-3 was collected during the Easy Tier Server initial 15-minute learning time. Thus, although the device is read-cache enabled, the read cache function has not started yet. Example 6-3 Etcadmin iostat output collected before the read cache function started # etcadmin -a iostat -d hdisk16 ETS Device I/O Statistics -- hdisk16 --------------------------------------------------Read Count: 449020 Write Count: 85921 Read Hit Count: 0 Partial Read Hit Count: 0 Read Bytes Xfer: 14713487360 Write Bytes Xfer: 2815459328 Read Hit Bytes Xfer: 0 Partial Read Hit Bytes Xfer: 0 Promote Read Count: 0 Promote Read Bytes Xfer: 0 Example 6-4 shows the etcadmin -a iostat command for DS8870 hdisk16 when the read cache function has already been started. You can use the numbers of this output to get individualized per-LUN statistics like Read/Write Ratio, and DAS Cache Read Hit Percentage (Read Count/Read Hit Count + Partial Read Hit Count), for example. Example 6-4 Etcadmin iostat output from a given DS8870 hdisk with the read cache function started # etcadmin -a iostat -d hdisk16 ETS Device I/O Statistics -- hdisk16 --------------------------------------------------Read Count: 3385448 Write Count: 914249 Read Hit Count: 2186632 Partial Read Hit Count: 5934 Read Bytes Xfer: 83262103552 Write Bytes Xfer: 18077851648 Read Hit Bytes Xfer: 47124586496 Partial Read Hit Bytes Xfer: 194445312 Promote Read Count: 110750 Promote Read Bytes Xfer: 116129792000 In this output, the Read Hit Count is the total number of read operations that were issued to the driver that were full DAS cache read hits. The Partial Read Hit Count is the total number of read operations that were issued to the driver that were partial DAS cache read hits. The Partial Read Hit is a number of instances in which a read request had part, but not all, of the data requested in cache. The remainder of the data that is not available in cache must be acquired from DS8870. For more information about etcadmin -a iostat output, consult the Easy Tier Server User’s Guide, available in IBM System Storage DS8000 Information Center, at this site: http://publib.boulder.ibm.com/infocenter/dsichelp/ds8000ic/index.jsp 90 IBM System Storage DS8000 Easy Tier Server Related publications The publications listed in this section are considered particularly suitable for a more detailed discussion of the topics covered in this paper. IBM Redbooks The following IBM Redbooks publications provide additional information about the topic in this document. Note that some publications referenced in this list might be available in softcopy only: IBM System Storage DS8870: Architecture and Implementation, SG24-8085 IBM System Storage DS8000 Host Attachment and Interoperability, SG24-8887 IBM System Storage DS88870 Product Guide IBM System Storage DS8000 Easy Tier, REDP-4667 IBM System Storage DS8000 Easy Tier Heat Map Transfer, REDP-5015 IBM System Storage DS8000: Easy Tier Application, REDP-5014 You can search for, view, download, or order these documents and other Redbooks, Redpapers, Web Docs, drafts, and additional materials, at the following website: ibm.com/redbooks Other publications These publications are also relevant as further information sources: IBM System Storage DS8700 and DS8800 Introduction and Planning Guide, GC27-2297-07 IBM System Storage DS8700 Performance with Easy Tier, WP101675 IBM System Storage DS8700 and DS8800 Performance with Easy Tier 2nd Generation, WP101961 IBM System Storage DS8800 and DS8700 Performance with Easy Tier 3rd Generation, WP102024 Online resources These websites and URLs are also relevant as further information sources: IBM data storage feature activation (DSFA): http://www.ibm.com/storage/dsfa Documentation for the DS8000 system: http://www.ibm.com/systems/storage/disk/ds8000/index.html © Copyright IBM Corp. 2013. All rights reserved. 91 IBM System Storage Interoperation Center (SSIC): http://www.ibm.com/systems/support/storage/config/ssic/index.jsp IBM announcement letters (search for R6.1): http://www.ibm.com/common/ssi/index.wss IBM Techdocs Library - The IBM Technical Sales Library: http://www.ibm.com/support/techdocs/atsmastr.nsf/Web/Techdocs IBM System Storage DS8000 Information Center: http://publib.boulder.ibm.com/infocenter/dsichelp/ds8000ic Help from IBM IBM Support and downloads ibm.com/support IBM Global Services ibm.com/services 92 IBM System Storage DS8000 Easy Tier Server Back cover ® IBM System Storage DS8000 Easy Tier Server Redpaper Unified storage caching and tiering solution Leverage AIX direct-attached flash devices Cache management and workload monitoring IBM Easy Tier Server is one of several Easy Tier enhancements introduced with the IBM DS8000 Licensed Machine Code 7.7.10.xx.xx. Easy Tier Server is a unified storage caching and tiering solution across IBM AIX servers and supported direct-attached storage (DAS) flash drives. Easy Tier Server manages placing a copy of the “hottest” volume extents on flash drives attached to an AIX server. Data can be read directly from flash drives local to the application host rather than from cache or disk drives in the DS8870, while maintaining other advanced feature functions. ™ INTERNATIONAL TECHNICAL SUPPORT ORGANIZATION This IBM Redpaper publication explains the Easy Tier Server concept and explores key aspects of its architecture, design, and implementation. BUILDING TECHNICAL INFORMATION BASED ON PRACTICAL EXPERIENCE From a more practical standpoint, this publication also contains numerous illustrations and examples that help you set up, manage, and monitor Easy Tier Server. IBM Redbooks are developed by the IBM International Technical Support Organization. Experts from IBM, Customers and Partners from around the world create timely technical information based on realistic scenarios. Specific recommendations are provided to help you implement IT solutions more effectively in your environment. For more information: ibm.com/redbooks REDP-5013-00