Grid Component Model and Platform An Overview Workshop

Grid Component Model and Platform: An Overview Workshop on Evolutions of GRIDs Towards SOKUs OGF 20, Manchester, 8 May 2007 Vladimir Getov Core. GRID STE Institute Leader University of Westminster, London, U. K. http: //www. coregrid. net V. S. Getov@westminster. ac. uk

Grid Research Projects under FP 6 supporting the Grid community EU Funding: 130 M€ trust, security Grid. Trust Challengers Grid. Coord Grid@Asia Bridge Degree EC-Gin Know. Arc Gredia Qos. Cos. Grid g-Eclipse Grid. Comp Bein. Grid business SIMDAT industrial simulations Grid. Econ A-Ware Next. GRID Akogrimo BREIN service architecture mobile services agents & semantics Core. GRID Xtreem. OS Linux based Grid operating system HPC 4 U Specific support action Edutain@ Grid Sorma business experiments virtual laboratories Wave 2 – start 2006 Wave 1 – start 2004 Grid services, business models Argu. Grid platforms, user environments Uni. Grids Echo. Grid Assess. Grid Provenance Grid 4 all Nessi-Grid international cooperation Integrated project Network of excellence data, knowledge, semantics Chemomen tum K-WF Grid Inteli. Grid Datamining Onto. Grid Specific targeted research project European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 2

Six highly-focused Research Institutes across 41 Research Labs Ø Preventing research fragmentation Ø Developing world-class scientific and technological excellence Ø Achieving sustainable integration Ø Contributing to the realisation of the European Research Area for Grid Research System architecture Programming models Knowledge and data management Resource Management & Scheduling Grid Information, Resource and Workflow Grid Systems, Monitoring Tools and Environments Gathering 145 researchers & 169 Ph. D students into a single EU Laboratory on Grid Technologies European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 3

Core. GRID Definition of Future Grids A fully distributed, dynamically reconfigurable, scalable and autonomous infrastructure to provide location independent, pervasive, reliable, secure and efficient access to a coordinated set of services encapsulating and virtualizing resources (computing power, storage, instruments, data, etc. ) in order to generate knowledge. European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 4

From Grids to SOKU Evolution of HPCN … Knowledge Technologies Current Grids Next Generation Grids Evolution of the Web Software Technologies … SOA Methodologies Service. Oriented Knowledge Utility Autonomic Computing European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 5

Background: How to build Grid Middleware • • • Proprietary middleware (Globus 1. 0, Legion, Unicore, …) – Resources exposed through an API – Non interoperable ! Object-based middleware – Resources exposed through distributed objects (Java, CORBA, etc. ) – Some interoperability issues with the communication protocols (CORBA IIOP) – Not anymore at the top of the hype ! Service-based middleware – Resources exposed through services – Strong support from the Industry – At the top of the hype ! – Need some extensions (stateful Web services) European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 6

One of the Main Research Challenges for Future Grids To develop the software design and development methodology of a generic component-based Grid platform for both applications and tools/systems/PSEs to have a single, seamless, “invisible” Grid software services infrastructure. Possible Solution: Grid Component Model (GCM): • • Proposal for a Grid Component Model - DPM 02 Basic Features of the Grid Component Model (assessed) DPM 04 Gore. GRID Institute on Programming Models European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 7

GCM: Main Features – – – Ø Ø Component hierarchy Extensibility of the model Support for adaptivity Language neutrality Interoperability Reflexivity Lightweight portable and compact implementations Well-defined semantics (allow future formalization) European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 8

GCM Technical Structure Component Specification as an XML schema Run-Time API defined in several languages C, Java, etc. Packaging described as an XML schema Information for Deployment (Virtual Nodes, … Variables, File Transfer, …) European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 9

Collective Interfaces Simplify the design and configuration of component systems Expose the collective nature of interfaces – Multicast, Gathercast, gather-multicast The framework handles collective behaviour at the level of the interface Based on Fractal API : – Dedicated controller – Interface typing Verifications European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 10

Content and Main Activities: GCM Reference Implementation 1 - Primitive Component Programming 2 - Legacy Code Wrapping, Interoperability 3 - Composition and Composites, Deployment 4 – Autonomic features 5 – IDE for GCM (Composition GUI, etc. ) European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 11

Research Example: Componentising an Application for the Grid INRIA and University of Westminster Pro. Active library –Java distributed object middleware for parallel and concurrent programming –Main features: Active objects, Asynchronous method invocation, Group communications, Descriptor-based deployment Jem 3 D – numerical solver for the 3 D Maxwell’s equations modelling the time domain propagation of electromagnetic waves – follows typical “geometric decomposition” parallelisation – distributed object application using Pro. Active European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 12

Research Example: Componentisation Process General, architecture-based process Object-based system Component-based system Early GCM using Pro. Active – extends Fractal with: • distributed components • multicast interfaces • configurable deployment on the Grid European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 13

Research Example: Performance Evaluation Comparison: Object-based vs. Component-based – experiments on Grid’ 5000 using up to 308 processors, allocated on up to 3 clusters – 7 experiments using different problem size and number of processors – execution times of two versions are similar European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 14

Domain-Specific Metadata for Model Validation and Performance Optimisation – Legacy Applications GENIE is an interactive, legacy code for Earth system modelling. Our hypothesis is that componentising the application and using domain-specific metadata will help transforming it into a scalable yet efficient Grid system. European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 15

Domain-Specific Metadata for Model Validation and Performance Optimisation Motivation: Enable legacy applications to evolve as a part of the scalable problem solving environments within modern Grid systems. Framework: Componentising existing applications along with domainspecific metadata so that issues arising thereof can be addressed using this metadata. Result: Used GENIE (Earth Simulation System) as a motivating example. Derived different domain- and component-specific metadata and optimisation strategies. Further Work: The principles we outlined are application-specific. A generic, but domain-restricted approach is required and potential performance benefits need to be demonstrated. Partners: - University of Westminster (UK) - Imperial College - London (UK) - Ongoing work: Core. GRID TR-0068 and a chapter in a Core. GRID Springer volume European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 16

Some Future Research Goals • Adoption of GCM for Grid applications development • Generic, lightweight component-based Grid platform design methodology • Use of GCM for Grid system software design • Integration of application and system components into a single adaptable Grid platform • Interoperability between peer-to-peer and clientserver paradigms in hybrid Grid systems European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 17

New Spin-off Project - Grid. COMP - FP 6 Call 5 Grid. COMP nvolves 6/12 Core. GRID partners: INRIA, ERCIM, UNIPI, UOW, ISTI/CNR, UCHILE Main goal: develop a GCM prototype platform Strong industrial involvement: IBM, Atos Origin, Grid Systems Worldwide partners: Tsinghua University (China), Melbourne University (Australia), and University of Chile (Chile) European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 18

Conclusions • Research work ongoing in both Core. GRID and Grid. COMP projects • Some other Core. GRID Institutes adopting GCM • More research results to be expected soon • Strong interest from the US community – joint focused yearly workshop • High industrial interest via the Grid. COMP project • ETSI interest to initiate further work on the GCM specification • An OGF activity - ? European Research Network on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies 19