The Gridbus Toolkit for Service-Oriented [Cluster/Grid] Computing “Economic Paradigm for Clusters and Grids-based Utility Computing” Rajkumar Buyya Grid Computing and Distributed Systems (GRIDS) Lab. The University of Melbourne Melbourne, Australia www.gridbus.org Collaboration: Nimrod-G Project WW Grid Collaborations Nimrod-G Resource Broker 2 Virtual Lab Collaborations Yuko Mizuno-Matsumoto3 Susumu Date 3 Acknowledgements: Co-authors Nimrod-G David Abramson, Monash Jon Giddy, DSTC Slavisa Garic (nirmod 1.1 update) Manzur Murshed, Monash (GridSim core 1.0) Kim Branson, WEHI for Structural Biology Susumu Date, Osaka University Wolfgang Gentzsch, Sun Microsystems GridSim Virtual Lab NeuroGrid Industry Collaborator Gridbus Project Members Srikumar Venugopal (Data Grid Scheduler) Jia Yu (Grid Market Directory) Martin Placek (Web-based Monitoring and Control Station) Shoaib Burq (Visual Tool for Parameter Sweep Application Creation) Alex Burmouta (Grid Bank) Anthony Sulistio (GridSim 2.0 Visual modeller) Chee Shin Yeo (GridSim 2.0 Vistual modeller) Steve Melnikoff (HEP data catalogue and apps parameterization) Lyle Winton, School of Physics (HEP data catalogue and BASF application help) Libra 1.0: CPM/Jxta: 4 Jahanzeb Sherwani, Nosheen Ali, Nausheen Lotia, Zahra Hayat, Lahore University of Management and Science Fazilah Haron, Chan Huah Yong, Liew Sun, U. Sains Malaysia Agenda Grid Challenges and The Project Vision Grid Architecture for Computational Economy Gridbus and Nimrod-G Technologies Scheduling Experiments on the World Wide Grid testbed New Gridbus Tools Conclusion Grid Grid Economy 5 Scheduling Economics A Sample Grid Computing Environment: Resource Sharing & Aggregation Grid Information Service Grid Resource Broker R2 2 R5 R3 Application database R4 RN Grid Resource Broker R6 Grid Information Service 6 R1 Resource Broker Why Grids ? Large Scale Exploration needs them—Killer Applications. Solving grand challenge applications using computer modeling, simulation and analysis Aerospace Internet & Ecommerce Life Sciences CAD/CAM 7 Digital Biology Military Applications Grid Characteristics and Challenges Entities/Issues Users, Resources, Owners Grid Information Service Geographically Grid Resource Broker Distributed resources, users, apps R2 2 R3 Heterogeneous Application database R4 Resource Varies with time Availability/Capability R5 R N Policies and strategies Grid Resource Broker R6 QoS requirements Cost / Price Grid Information Service challenges 8 Characteristics Heterogeneous & decentralised R1 Heterogeneous Resource Broker Varies: different resources, users, time * Resource Management * Application Construction Grid Challenges and Technologies Security Computational Economy Uniform Access Resource Discovery 9 System Management Data locality Resource Allocation & Scheduling Application Construction Network Management Some Grid Projects & Technologies Australia Nimrod-G Gridbus GridSim Virtual Lab DISCWorld GrangeNet ..new coming up Europe UNICORE Cactus UK eScience EU Data Grid EuroGrid MetaMPI XtremeWeb and many more. India Japan 10 Korea... I-Grid Ninf DataFarm N*Grid USA Cycle Stealing & .com Initiatives Globus Legion OGSA Sun Grid Engine AppLeS NASA IPG Condor-G Jxta NetSolve AccessGrid and many more... Distributed.net SETI@Home, …. Entropia, UD, Parabon,…. Public Forums Global Grid Forum Australian Grid Forum IEEE TFCC CCGrid conference P2P conference http://www.gridcomputing.com mix-and-match Object-oriented Internet/partial-P2P Network enabled Solvers Economic-based Utility Computing 11 Nimrod-G The Gridbus Vision: To Enable Service Oriented Grid Computing & Bus iness! WW Grid Gridbus World Wide Grid! + marketplace for Service-Oriented Computing 12 Gridbus Vision Scales: Breaks Administrative Barriers P E R F O R M A N C E 2100 2100 2100 2100 2100 2100 2100 2100 2100 Administrative Barriers •Individual •Group •Department •Campus •State •National •Globe •Inter Planet •Universe + Q o S 13 Personal Device SMPs or SuperComputers Local Cluster Enterprise Cluster/Grid Global Grid Inter Planet Grid GRIDS Lab @ the U. of Melbourne, The Gridbus Project: www.gridbus.org R& D in Distributed Computational Economy and end-to-end infrastructure for Service-Oriented/Utility Computing: : Visual Tools for Creation of Distributed Applications Grid Economy & Scheduling (via Nimrod-G Broker) GridSim Toolkit: Grid Modeling and Simulation (Java based): Libra: Economic Cluster Scheduler Grid Bank: Accounting and Transaction Management Grid Market Directory and Web Services World Wide Grid (WWG) – A Global Grid Testbed Jxta-based P2P Compute Power Market (CPM) Application Enabler Projects: 14 Architecture, Specification, and “Open Source” Reference Implementation in Collaboration with interested Global Peers. Platform: Cluster, Grid, P2P for diverse applications. Virtual Laboratory Toolset for Drug Design High-Energy Physics and the Grid Network (HEPGrid) Brain Activity Analysis on the Grid (NeuroGrid) Future: GridEmail for Spam Management —read only if sender pays! Gridbus: Towards Building a Grid Economy for Sustained Resource Sharing 15 Offers Incentive and Enable the Creation and Promotion of: Grid Marketplace (competitive) ASP Service Oriented Computing ... And let users focus on their own work (science, engineering, or commerce)! Grid Economy: Incentive, Resource Allocation and Regulation 16 New challenges of Computational Economy Resource Owners Resource Consumers 17 How do I decide prices ? (economic models?) How do I specify them ? How do I enforce them ? How do I advertise & attract consumers ? How do I do accounting and handle payments? ….. How do I decide expenses ? How do I express QoS requirements ? How I trade between timeframe & cost ? …. Grids need to support the automation of the process. GRACE: A Reference Grid Economy Services Architecture GRid Architecture for Computational Economy (GRACE) GRACE: A Reference Service-Oriented Grid Architecture for Computational Economies Data Catalogue Grid Bank Programming Environments Applications Sign-on Info ? Grid Explorer Job Control Agent Grid Market Services Health Monitor Grid Node N Secure Schedule Advisor QoS Grid Node1 Pricing Algorithms Trade Server Trade Manager Trading … Deployment Agent JobExec Grid Resource Broker Misc. services Resource Allocation Storage Grid Middleware Services Accounting Resource Reservation R1 Grid Consumer 19 Information Service R2 … Rm Grid Service Providers Realizing the GRACE Vision: Development Methodology Used Applications Science Engineering Commerce … Portals High-level Services and Tools MPI-G CC++ … Parameter Specification Nimrod-G Higher Level Resource Aggregators ActiveSheet Visual Paramteric Modeller Gridbus Scheduler Grid Apps. Grid Tools Core Services MDS GRAM GASS GARA Jxta GMD GBank GTS Middleware PKI-based Grid Security Interface (GSI) 20 JVM TCP UDP Linux Irix Solaris Local Services Libra Condor LSF Grid PBS SGE Grid Fabric Using Gridbus + Nimrod-G: Brokering and On Demand Assembly of Grids Application Code Explore data 1 Data Source Visual Parametric Tool Data 10 2 (Instruments/dis tributed sources) Data Catalogue Data Replicator (GDMP) 5 6 (Data) Grid Resource Broker 3 Grid Info Service 4 ASP Catalogue 9 12 Grid Market Directory 7 8 Grid Service (GS) (Globus/CPM) Grid Service Bill GS Cluster Scheduler PE 21 Grid Service Provider (GSP) (e.g., CERN) CPU or PE GSP (e.g., IBM) Cluster Scheduler PE GSP (e.g., UofM) PE GTS GSP (e.g., VPAC) 11 GridBank GSP (Accounting Service) Gridbus Tools: Enabling Utility Computing for Distributed Data Intensive Applications Visual Environment for Parameter Application Creation Grid Market Directory (Portal for GSPs + Webservices for GRBs) Gridbus Data Grid Scheduler (built on Nimrod-Farming Engine) Gridbus Monitoring and Control Station (Web-based) GridBank: Grid Accounting Services Architecture GridSim Toolkit 2.0 Libra: Cluster Economy Scheduler Sample P-Sweep/Task/Data Farming Grid Applications Bioinformatics: Drug Design / Brain Activity Analysis Ecological Modelling: Combinatorial Control Strategies Optimization: for Cattle Tick Meta-heuristic Data Mining parameter estimation Neuroscience Computer Graphics: Ray Tracing High Energy Physics: Searching for Rare Events Electronic CAD: Field Programmable Gate Arrays VLSI Design: Finance: SPICE Simulations Investment Risk Analysis Civil Engineering: Building Design Automobile: Crash Simulation 23 Network Simulation Aerospace: Wing Design astrophysics Drug Design: Data Intensive Computing on Grid Molecules Protein 24 Chemical Databases (legacy, in .MOL2 format) It involves screening millions of chemical compounds (molecules) in the Chemical DataBase (CDB) to identify those having potential to serve as drug candidates. [Collaboration with WEHI for Medical Science, Melbourne] Gridbus Visual Tool for Parametric Application Creation (e.g., Docking) 25 NeuroGrid: QoS Requirements driven MEG Data (Brain Activity) Analysis on the Grid Analysis All pairs (64x64) of MEG data by shifting the temporal region of MEG data over time: 0 to 29750: 64x64x29750 jobs 64 sensors MEG 2 Data Generation 3 1 5 Data Analysis Grid Resource Broker (Nimrod-G+Gridbus) Results 4 Life-electronics laboratory, AIST [deadline, budget, optimization preference] World-Wide Grid •Provision of MEG (MagnetohyEncephaloGraphy) analysis 26 [Collaboration with Osaka University, Japan] The Grid Market Directory for Service Publication and Discovery Grid Market Directory (GMD) Web Server (Tomcat) GMD Portal Manager GMD Query Webservice Grid Service Info (RDBMS) Browse Publish/Manage Query(SOAP+XML) Job submission Grid Node Grid Node Provider (Web Client) Consumer (Grid Resource Broker) Grid Node 27 Consumer (Web Client) Nimrod-G : A Grid Resource Broker A resource broker for managing, steering, and executing task farming (parameter sweep/SPMD model) applications on the Grid based on deadline and computational economy. Based on users’ QoS requirements, our Broker dynamically leases services at runtime depending on their quality, cost, and availability. Key Features 28 A single window to manage & control experiment Persistent and Programmable Task Farming Engine Resource Discovery Resource Trading Scheduling & Predications Generic Dispatcher & Grid Agents Transportation of data & results Steering & data management Accounting Nimrod-G and Gridbus Scheduler Nimrod/G Client Nimrod/G Client Gridbus Client (Default Scheduler or Gridbus Scheduler) Nimrod/G Engine Schedule Advisor Trading Manager Grid Store Grid Dispatcher Grid Explorer Grid Middleware TM Globus, Legion, Condor, etc. TS GE GIS Grid Information Server(s) RM & TS RM & TS G RM & TS C L G Globus enabled node. 29 L Legion enabled node. RM: Local Resource Manager, TS: Trade Server G C L Condor enabled node. Deadline A Nimrod-G CostMonitor 66 Arlington Alexandria Legion hosts She na nd o a h Rive r 64 64 81 Ra p p a ha nno c k Po to m a c Rive r Rive r Ja m e s Rive r Roanoke Richmond Ap p o m a to x Rive r Hampton Norfolk Virginia Beach Portsmouth Chesapeake Newport News 77 VIRGINIA 85 Globus Hosts Bezek is in both Globus and Legion Domains 30 G-monitor – Web-based Station 31 Deadline and Budget Constrained Scheduling Algorithms 32 Algorithm/ Strategy Execution Time Execution Cost (Deadline, D) (Budget, B) Cost Opt Limited by D Minimize Cost-Time Opt Minimize when possible Minimize Time Opt Minimize Limited by B Conservative-Time Opt Minimize Limited by B, but all unprocessed jobs have guaranteed minimum budget Scheduling Experiments on the World Wide Grid The World Wide Grid Sites WW Grid Cardiff/UK Portsmoth/UK Manchester, UK Cambridge, UK TI-Tech/Tokyo ETL/Tsukuba AIST/Tsukuba ANL/Chicago USC-ISC/LA UTK/Tennessee UVa/Virginia Dartmouth/NH BU/Boston UCSD/San Diego Santiago/Chile 34 EUROPE: ZIB/Germany PC2/Germany AEI/Germany Lecce/Italy CNR/Italy Calabria/Italy Pozman/Poland Lund/Sweden CERN/Swiss CUNI/Czech R. Vrije: Netherlands Kasetsart/Bangkok Singapore U. Of Melbourne Monash U. VPAC/Melbourne U. Of Sydney Experiment Setup Workload: Deadline: 2 hrs. and budget: 396000 G$ Strategies: 1. Minimise cost 2. Minimise time Execution: 35 165 jobs, each need 5 minute of CPU time Optimise Cost: 115200 (G$) (finished in 2hrs.) Optimise Time: 237000 (G$) (finished in 1.25 hr.) In this experiment: Time-optimised scheduling run costs double that of Cost-optimised. Users can now trade-off between Time Vs. Cost. Resources Selected & Price/CPU-sec. 36 Resource & Location Grid services & Fabric Cost/CPU sec.or unit No. of Jobs Executed Linux Cluster-Monash, Melbourne, Australia Globus, GTS, Condor 2 64 153 Linux-Prosecco-CNR, Pisa, Italy Globus, GTS, Fork 3 7 1 Linux-Barbera-CNR, Pisa, Italy Globus, GTS, Fork 4 6 1 Solaris/Ultas2 TITech, Tokyo, Japan Globus, GTS, Fork 3 9 1 SGI-ISI, LA, US Globus, GTS, Fork 8 37 5 Sun-ANL, Chicago,US Globus, GTS, Fork 7 42 4 Time_Opt Cost_Opt. Total Experiment Cost (G$) 237000 115200 Time to Complete Exp. (Min.) 70 119 G-monitor Usage: HPC Challenge Demo @ SC 2002 Australia North America GMonitor MEG Visualisation Melbourne+Monash U: ANL: SGI/Sun/SP2 NCSA: Cluster Wisc: PC/cluster NRC, Canada Many others Gridbus+ Nimrod-G VPAC, Physics Solaris WS @ SC 2002/Baltimore Europe Grid Market Directory Asia AIST, Japan: Solaris Cluster Osaka University: Cluster Doshia: Linux cluster Korea: Linux cluster 37 WW Grid Internet ZIB: T3E/Onyx AEI: Onyx CNR: Cluster CUNI/CZ: Onyx Pozman: SGI/SP2 Vrije U: Cluster Cardiff: Sun E6500 Portsmouth: Linux PC Manchester: O3K Cambridge: SGI Many others Resource Scheduling for DBC Time Optimization Condor-Monash Linux-Prosecco-CNR Linux-Barbera-CNR Solaris/Ultas2-TITech SGI-ISI Sun-ANL 12 No. of Tasks in Execution 10 8 6 4 2 38 Time (in Minute) 68 72 60 64 52 56 44 48 36 40 28 32 20 24 16 8 12 4 0 0 Resource Scheduling for DBC Cost Optimization Condor-Monash Linux-Prosecco-CNR Linux-Barbera-CNR Solaris/Ultas2-TITech SGI-ISI Sun-ANL 14 No. of Tasks in Execution 12 10 8 6 4 2 Time (in Minute) 39 10 2 10 8 11 4 96 90 84 78 72 66 60 54 48 42 36 30 24 18 6 12 0 0 The GridSim Toolkit A Java based tool for Grid Scheduling Simulations Application, User, Grid Scenario’s Input and Results Application Configuration Resource Configuration User Requirements Grid Scenario ... Output Grid Resource Brokers or Schedulers GridSim Toolkit Application Modeling Resource Entities Information Services Job Management Resource Allocation Statistics Resource Modeling and Simulation (with Time and Space shared schedulers) Single CPU SMPs Clusters Load Pattern Network Reservation Basic Discrete Event Simulation Infrastructure SimJava Distributed SimJava Virtual Machine (Java, cJVM, RMI) PCs 40 Workstations SMPs Clusters Distributed Resources Selected GridSim Users! 41 Libra: Economic-based cluster scheduler User Application +deadline, budget Cluster Management System (e.g., SGE/PBS) Job Job Input Dispatch Control Control (node 1) Application User Budget Check Control Best Node Evaluator ... (node 2) Deadline Control Node Querying Module (node N) Scheduler (Libra) Server: Master Node 42 Cluster Worker Nodes with node monitor (e.g.,pbs_mom) CPM: P2P Market-Maker Technology (Jxta based Implementation) Market Server - Discovery - Membership User (Consumer) Market Repository CPM Agent Trader Bill Job Management Accounting 43 Resources (Provider) Summary and Conclusion 44 Grid Computing is emerging as a next generation computing platform for solving large scale problems. Resource management is a complex undertaking as systems need to be adaptive, scalable, competitive,…, and driven by QoS. Scheduling experiments on the World Wide Grid demonstrate our brokers ability to dynamically lease services at runtime based on their quality, cost, and availability and consumers QoS requirements. Gridbus tools allow Creation/Composition of Grid Applications with minimal or no effort. [Grids fail if we ask biologists to create applications using Globus/MPI—assembly language of grid computing). Economic-based resource management and scheduling is essential for pushing Grids into mainstream computing and creating the World-Wide Grid Marketplace. The End, Thank You! ? For further info: Visit our home in cyberspace @: 45 http://www.gridbus.org Visit our Research Exhibit @ SC 2002 for a live demo. Visit our home on planet earth @ Melbourne, Australia! W are happy to collaborate with anybody in the universe. PS: When we deploy our Economic-based GridEmail application, our Inbox accepts only paid or bartered emails and the rest will be sent to junk box automatically. This helps in handling Nigerian emails!