Fusion-io directCache Non-Virtualized Server Caching Solution NetApp Confidential – Limited Use 1 Why Deploy Flash Technology? Need low latency (<1ms) – Lower application run time – Provide better user experience Need more IOPS – Offload IOPS from spinning disk to flash – Increased workload on same infrastructure Green computing – Reduce spindle/rack requirements – Reduce power requirements 2 Server Flash as Cache Flash emerging in server as IOPS tier • Significant performance benefits, but relatively costly However, flash alone is just fast DAS (Direct Attached Storage) • No sharing across servers • No high availability, data protection, disaster recovery • Management complexity Ideal: Server flash as a cache for network storage • Flash to improve performance, network storage to protect and manage data • Optimal $/GB and $/IOPS Target use cases • Application-specific IOPS acceleration, OLTP, virtualized OLTP, VM density 3 Virtual Storage Tiering Server Flash Virtual Storage Tier Data-driven Real-time Self-managing Array Flash Intelligently place “hot” data on the highest performing media in real time Efficient use of flash Simple to install Self-managing Nondisruptive operations Caching vs. data migration Minimal HDD I/Os Highly granular HDD Storage Real-time responsiveness 4 Server Side Flash as Cache Utilizing flash without compromising enterprise array capabilities – Application integrated Snapshot®/SnapRestore® technologies SnapDrive® for Windows®/SnapManager® for SQL Server ® – Disaster recovery capabilities SnapMirror®/MetroCluster ™ – High availability CFO/MetroCluster – Shared storage Virtualization and host clustering 5 Server Cache Partner Alliance – Software Data Coherency Implications with Caching What is data coherency? Consistency between data that is being presented in cache and data that actually resides on disk Why is data coherency Important? Storage Array Physical Server Application Cache 0100100100101 0101010101101 0101010010100 1000100101001 Data Coherency 0100100100101 0101010101101 0101010010100 1000100101001 Caches that don’t recognize changes to the primary data copy (e.g. via SnapRestore®) will cause data corruption issues Data coherency issues can vary in severity based on the cache implementation. Key criteria include: • When does the cache invalidate data? • Is the cache a write-through or a write-back cache? • Does the cache retain data across a reboot, etc? 6 Server Cache Partner Alliance Creating Customer Value • Provide industry's broadest coverage for use cases, operating systems and hypervisors - today • Provide partners with integration points, to expose NetApp value at the server Enabling “Best of Breed” solutions • Provides customers with best of breed solutions • Provide qualification criteria (SW & HW) to ensure interoperability, performance, and data management Simplifying Deployment • “Best Integrated” into solutions such as FlexPod® for rapid customer deployment 7 Server Cache Partner Alliance Program for joint testing/validation of server side caching solutions – Ensures existing data coherency issues are known and understood – Collaboration between NetApp® and partner to design resolution for data coherency issues 8 Fusion-io Bare Metal Cache Solution 9 Fusion-io Solution Components ioMemory Virtual Storage Layer (VSL) – Block interface for ioMemory (driver) ioMemory device – ioDrive2 – Hardware PCIe device providing flash capacity directCache software – Software that transforms ioMemory into a block based cache device ioSphere – GUI for ioMemory management, alerting, and performance monitoring 10 Fusion-io ioMemory VSL & ioDrive2 Enterprise reliability from MLC flash Simplified data path from CPU to flash = lower latency and increased data integrity Lower CPU and RAM utilization Flash endurance-extending capabilities 11 Fusion-io directCache Integrated with ioMemory VSL Caches frequently read data closer to the application Caches block storage (not NAS) Simple software configuration Maintain existing SAN configuration Physical Server Application OS Disk Controller VSL NetApp® Storage ioMemory 12 Fusion-io directCache directCache operates as a transparent filter driver in the host Primary storage device identity is maintained: no changes in existing storage identity – Support for NetApp® SnapDrive® for Windows® and SnapManager® for SQL Server® 13 Fusion-io directCache Application NetApp® Storage For bare-metal Windows and Linux Reads are cached to ioMemory Writes are passed through to primary storage and then cached for later reads All following reads of same data are from ioMemory Maintains existing SAN configuration 14 Multi-Volume Caching Support One ioMemory device caches multiple LUNs 64 backing stores per ioMemory device Independent settings for each volume ioMemory Cached data blocks 15 SnapDrive for Windows Data Coherency Maintaining data coherency during SDW SnapRestore® operation requires the NetApp® SnapRestore Support feature of directCache be installed When a SDW SnapRestore operation is detected, cache is automatically invalidated for the affected device Cache must be manually invalidated if a SnapRestore is performed manually at the array 16 directCache Product Features Product Type Add-on software Cache type Fully associative cache Cache mode Write-through Minimum cacheable block size 512 bytes Max cache instances per host 64 Additional CPU&RAM overhead No noticeable increase over the ioMemory VSL driver Product Architecture Independent Linux® kernel module Windows® volume/disk filter driver Management CLI and ioSphere-based GUI 17 Fusion-io ioSphere (for Bare Metal) Fusion-io offers two different ioSphere products: – ioSphere – for bare metal environments using ioMemory and directCache software – ioSphere VM Edition – management system for ioTurbine software in virtual environments – These are different products Customers with valid Fusion-io support contracts are entitled to download ioSphere 18 Fusion-io ioSphere (for Bare Metal) GUI management and monitoring software for ioMemory devices Real-time visibility into all deployed ioMemory Historical performance monitoring Warranty forecasting Intuitive Web user interface ioSphere - VM or Physical Host ioDrive2 ioDrive2 ioDrive2 19 Fusion-io ioSphere (for Bare Metal) ioSphere capabilities Interface Web GUI for remote access Authentication Local users and/or LDAP accounts Monitoring Multihost ioMemory device monitoring, auto discovery Alerting SMS/Email, customizable and searchable history ioMemory Performance metrics Live display and historical reports Configuration Configure ioMemory devices Usage warranty forecasting 20 Supported Systems and Requirements 21 NetApp ioTurbine Validation NetApp® validation includes verification of certain product combinations and features – Data coherency for snapshot restore events – SAN interoperability – see note in IMT Policies & Guidelines http://support.netapp.com/matrix/ Other operating systems, product combinations, and protocols are fully supported by Fusion-io but have not been validated by NetApp 22 NetApp directCache Validation NetApp directCache data coherency validation SnapDrive® for Windows® SnapDrive® for UNIX® SnapProtect® directCache 1.1.6 Validated – Requires directCache NetApp® SnapRestore® Support feature Not supported* To be validated Current as of August, 2012 * SDU is inoperable when directCache caching is enabled 23 NetApp directCache Validation Configuration used for NetApp validation directCache 1.1.6 with directCache NetApp® SnapRestore® Support feature installed ioMemory VSL 3.1.5 SnapDrive® for Windows® 6.4.1 SnapManager® for SQL® 5.2 Server Operating System Windows Server® 2008 R2 SP1 Storage Connectivity FCoE 24 Fusion-io directCache System Support Other operating systems and features supported by Fusion-io for directCache Fusion-io directCache 1.1.6 Compatible cache devices ioDrive1 and ioDrive2 family devices ioDrive, ioDrive Duo, and ioDrive Octal products, and mezzanine form factors Operating Systems - Windows® 64-bit Microsoft® Server 2008/2008 R2 Operating Systems - Linux® RHEL 5.x, RHEL 6.x, SLES 11.x, Debian Squeeze, Cent OS 6.2 Host clustering MSCS, VCS 5.x and above on RHEL, Linux native cluster Supported primary storage connectivity FC, iSCSI (does not cache NAS) 25 ioDrive2 Requirements Fusion-io ioDrive2 device requirements Form Factor Low-profile PCI Express Connectivity PCIe 2.0 x4 electrical x8 physical 26 Fusion-io NetApp Resell NetApp Confidential – Limited Use 27 Fusion-io ioMemory – ioDrive2 ioMemory devices available from NetApp – ioDrive2 365GB MLC – ioDrive2 785GB MLC ioDrive2 Capacity 365GB MLC 785GB MLC Read Bandwidth (1 MB) Write Bandwidth (1 MB) 910 MB/s 590 MB/s 1.5 GB/s 1.1 GB/s Random Read IOPS (512B) 137,000 141,000 Random Write IOPS (512B) 535,000 535,000 Device raw throughput numbers NetApp Confidential – Limited Use 28 Hardware & Software Resell NetApp resell of Fusion-io hardware and software – Hardware: ioDrive2 365GB & 785GB MLC – Software: ioTurbine, directCache & ioSphere Note: No SKU for ioSphere needed. Hardware support contract provides ioSphere access SKU Description SW-FS0106 directCache Standard (per host) SW-FS0120 ioTurbine (per host) X-F00-001-365G-CS-0001-R6 ioDrive2 365GB MLC X-F00-001-785G-CS-0001-R6 ioDrive2 785GB MLC NetApp Confidential – Limited Use 29 Product Support Hardware and software support provided by Fusion-io directly to customer: – NetApp® support staff trained to help triage – NGS has Fusion-io hardware/software – Joint customer issues managed through TSANet License and account registration performed by customer/Fusion-io All software provided by Fusion-io from fusionio.com – No Fusion-io software hosted on NetApp site NetApp Confidential – Limited Use 30 Product Support SKUs Fusion-io support contracts resell by NetApp Software Support SW-SSPVN-FS0110 FusionIO SSP, directCache Gold 1Yr SW-SSPVN-FS0111 FusionIO SSP, directCache Gold 3Yr SW-SSPVN-FS0124 FusionIO SSP, ioTurbine Gold 1Yr SW-SSPVN-FS0125 FusionIO SSP, ioTurbine Gold 3Yr Hardware Support SVCVN-FS1-GLD-320-HE-1YR HW Svc, FusionIO ioDrive1 320GB SLC Gold 1Yr SVCVN-FS1-GLD-320-HE-3YR HW Svc, FusionIO ioDrive1 320GB SLC Gold 3Yr SVCVN-F00-GLD-365G-CS-1YR HW Svc, FusionIO ioDrive2 365GB MLC Gold 1Yr SVCVN-F00-GLD-365G-CS-3YR HW Svc, FusionIO ioDrive2 365GB MLC Gold 3Yr SVCVN-F00-GLD-785G-CS-1YR HW Svc, FusionIO ioDrive2 785GB MLC Gold 1Yr SVCVN-F00-GLD-785G-CS-3YR HW Svc, FusionIO ioDrive2 785GB MLC Gold 3Yr NetApp Confidential – Limited Use 31 directCache Administration and Management 32 directCache Terms and Definitions Important directCache terms and definitions Cache device The ioMemory ioDrive2 device used for flash memory capacity ioMemory control device The device name for the ioMemory ioDrive2 device. Example: \\.\fct5 Backing store The primary storage device to be cached. A device name, drive letter, or mount point. Example: PhysicalDrive3, F:\, or C:\MSSQL\tempdb Bind To configure a backing store to be cached Unbind To remove a backing store from a cache device and drop cache contents 33 directCache Installation Overview Installation high-level steps 1. 2. 3. 4. Install ioDrive2 card in host Install ioMemory VSL Install directCache software If using SDW install the directCache NetApp ® SnapRestore® Support feature 5. Install ioSphere software (optional) on different windows host 6. Configure storage for caching 34 Configuring directCache 1. Identify control device name for ioMemory C:\Users\Administrator> fio-status Found 1 ioMemory device in this system Fusion-io driver version: 3.1.3 build 27 Adapter: Single Controller Adapter Fusion-io ioDrive2 785GB, Product Number:F00-001-785G... Connected ioMemory modules: fct5: Product Number:F00-001-785G-CS-0001, SN:1210D0132 fct5 Attached as 'fct5' (block device) ioDrive2 Adapter Controller, Product Number:F00-001-785G... PCI:05:00.0 Firmware v6.0.0, rev 107004 Public 785.00 GBytes block device size Device name \\.\fct5 35 Configuring directCache 2. Nondisruptively bind cache device to backing store C:\Users\Administrator>dc-bind C:\MSSQL\tpce_tempdb \\.\fct5 About to create binding using backing store 'C:\MSSQL\tpce_tempdb’ caching device '\\.\fct5’ write-through mode WARNING: Binding will destroy any existing data on the caching device \\.\fct5! Do you wish to continue [y/n]? y Binding: [====================] (100%) / Successfully bound backing store 'C:\MSSQL\tpce_tempdb’ with cache device '\\.\fct5’. Warning occurs only for first backing store to bind to this device Note: binding a system/boot drive is not supported by directCache. 36 Removing Cache Configuration Nondisruptively unbind cache device from backing store C:\Users\Administrator>dc-unbind C:\MSSQL\tpce_tempdb About to tear down binding: backing store 'C:\MSSQL\tpce_tempdb’ cached by '/dev/fct5’ cache UUID: 52c1edca-0914-40f7-9275-b23aded2a2e8 Do you wish to continue [y/n]? y Unbinding: [====================] (100%) \ Successfully unbound: backing store 'C:\MSSQL\tpce_tempdb’ cached by '/dev/fct5' Note: Unbinding the backing store invalidates cache contents for that backing store. 37 Viewing Cache Status & Statistics Cache status and statistics can be displayed with dc-status command C:\Users\Administrator>dc-status –s fct5: Fusion-io ioDrive2 785GB, SN:1210D0132 Device fct5 cache bindings: Cache device: Backing store: Cache instance: VSU 8: Status: enabled, Mode: Write-through C:\MSSQL\tpce_tempdb dab3fedd-7338-4ea0-b181-9b1b352fb182 dda8d294-7d71-49bf-9b55-1099ed3aa59a In-use: 13.98GBytes Write stats: Hits: requests 0%, sector 0%, Misses: requests 100% Latency: hits 0.0s, misses 2.37247ms, partials 0.0s, total 2.37247ms Read stats: Hits: requests 72%, sectors 93%, Misses: requests 27% Latency: hits 327.666us, misses 50.3810ms, total 14.1122ms 38 Cache Statistics in Windows Perfmon Standard performance counters in Windows® Perfmon—reads/sec, sec/read—will not show correct stats – Cache hits bypass the part of the Windows stack where statistics are gathered – Standard Perfmon counters show stats only for IOPS that were writes or read cache misses directCache provides WMI Perfmon counters for monitoring various cache statistics 39 Cache Statistics in Windows Perfmon directCache provides Perfmon counters for monitoring various cache statistics 40 directCache Cache Invalidation Data in directCache cache is invalidated in the following cases: – System reboot – Binding/unbinding the backing store (dc-bind/dc-unbind) – Disabling/enabling caching (dc-disable/dc-enable) – During SDW SnapRestore® operation (only when directCache data coherency add-on is installed) 41 Collecting Data For Support/Troubleshooting fio-bugreport generates a detailed report about the state of the device and directCache – fio-bugreport runs several ioMemory VSL and directCache commands which must be located in your PATH environment variable for commands to complete successfully C:\Users\Administrator> fio-bugreport Report output: fio-bugreport-20120723_153906.cab Generating bug report. Please wait, this may take a while... --------------------------------------------------Gathering all Windows Event Logs...DONE Gathering Fusion-io Windows Event Logs...DONE Gathering System Information... 42 Recommendations & Best Practices 43 Recommendations for Server Caching If using SnapDrive® for Windows®, you must install the directCache NetApp® SnapRestore® Support feature Manual revert of storage at the array requires manual cache invalidation When performing manual storage reverts: 1. 2. Stop applications using storage Invalidate cache data (dc-unbind or dc-disable) 3. Disconnect storage devices from host or shut down host 4. Perform SnapRestore® 5. Reconnect storage devices to host or boot host 6. Re-enable cache 44 Recommendations for Server Caching Do not cache LUNs that are heavily written to – Example: Database logs, Snapinfo LUNs Do not use cache reservation feature – Let cache software manage cache utilization When using multiple cache devices, spread busiest backing stores across cache devices 45 Resources NetApp Confidential - Internal Use Only 46 References Server Caching Partner Alliance binder on field portal Overview of Server Caching Program Server Caching Partner Alliance FAQ Server Caching Program Customer Presentation NetApp Confidential - Internal Use Only 47 © 2012 NetApp, Inc. All rights reserved. No portions of this document may be reproduced without prior written consent of NetApp, Inc. Specifications are subject to change without notice. NetApp, the NetApp logo, Go further, faster, MetroCluster, SnapDrive, SnapInfo, SnapManager, SnapMirror, SnapProtect, SnapRestore, and Snapshot are trademarks or registered trademarks of NetApp, Inc. in the United States and/or other countries. Microsoft and Windows are registered trademarks of Microsoft Corporation. Linux is a registered trademark of Linus Torvalds. All other brands or products are trademarks or registered trademarks of their respective holders and should be treated as such. 48