gridchina lau

Category: Education

Presentation Description

No description available.


By: saura (128 month(s) ago)

looking for a ppt of grid computing

Presentation Transcript

Grid Infrastructure and Key Technologies (+ Grid Computing Research in Hong Kong): 

Grid Infrastructure and Key Technologies (+ Grid Computing Research in Hong Kong) Francis C.M. Lau 刘智满 (& Cho-li Wang) Systems Research Group (SRG) Department of Computer Science The University of Hong Kong 6 April 2005 - Beijing


Acknowledgement HKU Systems Research Group (SRG)


Outline Hong Kong Grid Status Report Hong Kong Grid Initiatives HKU CC, HKBU, HKU CS clusters China National Grid Project Asia Pacific Grid Project Grid Research Projects in HKU CS SLIM and InstantGrid G-JavaMPI and G-PASS JESSICA2 Summary and The Future

Hong Kong Grid 

Hong Kong Grid HKGrid Initiative Launched in Cluster2003 (Dec. 2003) CPU power, Memory, Network, Storage, Data.. Services.. Resource providers End users

HKGrid - Constituents : 

HKGrid - Constituents A 4 Tflop/s theoretical maximum computing power Links to China National Grid (CNGrid) and Asia-Pacific Grid (ApGrid) via CERNET and APAN respectively Internet2 connection to the Abilene backbone at Chicago, USA

Grid Research Projects in Hong Kong: 

Grid Research Projects in Hong Kong HKUST: Resource allocation and scheduling, topology optimization HKBU: Knowledge grid, autonomous grid service composition CityU: Agent-based wireless grid computing PolyU: Peer-to-peer grid, meta-scheduling, fault tolerance HKU CC: Scientific applications running across the ApGrid CC: Biosupport project with HKU-Pasteur Research Centre CS : Robust Speech Recognition (Dr. Q. Huo) CS : Simulation for the DNA Shuffling Experiment (Dr. T.W. Lam) CS : Approximate String Matching on DNA Sequences (L.L. Cheng) CS : Whole Genome Alignment (Dr. T.W. Lam) CS : China National Grid (CNGrid) project - HKU Grid Point ME: Parallel Simulation of Turbulent Flow Model (Dr. C.H. Liu, Dept. of Mechanical Engineering) ETI: Modeling of Air Quality in Hong Kong (with the Environmental Protection Department, HKSAR) ETI: RFID Grid (Prof. C.J. Tan)

HKU Computer Centre : 

HKU Computer Centre External Networking HARNET-Internet : 240Mbps (26 September 2004). The 1Gbps connection between HARNET and local Internet via HKIX (Hong Kong Internet Exchange) 300Mbps connection between the HARNET, connected to any of the eight tertiary institutions HARNET-Internet2 : Fast Ethernet October 20, 2004 : Inaugural Ceremony of HPC Cluster on Windows Platform hpcpower: 128 nodes (IBM x335) of dual Xeon 2.8GHz CPUs GigaEth connection (CISCO 4506)

HKU-Pasteur Research Centre Biosupport Project : 

HKU-Pasteur Research Centre Biosupport Project Collaboration between HKUCC, HKU-Pasteur Research Centre and Centre de Ressources INFOBIOGEN (France). Bioinformatics Tools:   The sequence analysis packages installed include EMBOSS, NCBI tools, FASTA, STADEN, PHYLIP, READSEQ, ClustalW/ClustalX, DIALIGN2 and the PHRAP/PHRED/CONSED package. Some tools installed also have on-line web interface, such as JEMBOSS, EMBOSS-GUI, NCBI-BLAST, FASTA and GenoList

Hong Kong Baptist University High Performance Cluster Computing Centre : 

Hong Kong Baptist University High Performance Cluster Computing Centre Message Passing Interface MPICH, LAM/MPI Mathematical: fftw (fast fourier transform) pblas (parallel basic linear algebra software) atlas (a collections of mathematical library) sprng (scalable parallel random number generator) MPITB -- MPI toolbox for MATLAB Quantum Chemistry software gaussian, qchem Molecular Dynamic solver NAMD, gromacs, gamess Weather modelling: MM5 64 nodes (Dual Intel Xeon 2.8GHz with 2GB RAM), Network: 65-port Extreme BlackDiamond 6816 Gigabit Ethernet switch Source: Quantum Chemistry

HKU Computer Science “Self-Made” Gideon 300 Linux cluster: 

300 Pentium 4 PCs @355 Gflops; Ranked #175 in TOP500 (11/2002) HKU Computer Science “Self-Made” Gideon 300 Linux cluster (Built in Oct. 2002)

HKU CS Grid Point: Grid and Cluster Software: 

HKU CS Grid Point: Grid and Cluster Software Gatekeeper Remote job submission Gideon Ostrich Srgdell Real - Globus Toolkit (GT) 2.0, 2.4, 3.0.1 Local Job Scheduler IPC / Network communication OpenPBS 2.3.16 Maui 3.2.5 HPF, Fortran 90 C, C++, Java with MPI JESSICA2 (HKU) WireGL, MatMPI - MPICH-G2 Grid middleware Job scheduling Programming Communication Lib

Performance Monitoring with Ganglia: 

Performance Monitoring with Ganglia URL:

China National Grid : HKU Grid Point: 

China National Grid : HKU Grid Point Supporting software: VEGA (织女星) GOS: dynamic service deployment, single-sign-on, data replication, and performance monitoring. Developed by Institute of Computing Technology, Chinese Academy of Sciences 中科院计算所开发的网格系统软件将与香港大学网格节点连接在一起First test run : Dec. 26, 2004… Dawning4000A (2560 Opteron proc, now 17th in TOP500) DeepComp 6800 (1024 I2 proc, now 38th in TOP500)


(2004. Nov. 28) : HKU supports G-JavaMPI, JESSICA2, WireGL, MatMPI


160 80 32 32 40 16 Asia Pacific Grid (APGrid) 10 countries 21 organizations 22 clusters 853 CPUs


Weather Forecast Demonstration on HKU Open Day – (Oct 2003)

Grid Research at HKU SRG: 

Grid Research at HKU SRG Selected Projects SLIM + InstantGrid { G-JavaMPI + G-PASS } JESSICA2 To setup To run To optimize

Our Goal: 

Our Goal Utility computing: to aggregate and make use of distributed computing resources transparently Traditional means: to utilize the dedicated HPC facilities distributed across institutions Performance and reliability are key Pervasive means: any user can be resource provider (e.g., idle PCs, etc.) or consumer, or both Convenience and security are key To construct an advanced grid computing platform to accommodate utility-like computing via traditional and “pervasive” means

An Advanced Grid Computing Platform: 

An Advanced Grid Computing Platform G-JavaMPI JESSICA InstantGrid SLIM Load balancing AGP On-demand Grid point construction (ODGPC) Research Issues Single- system image Performance and Reliability Objectives User’s convenience Grid point construction Convenient system administration (Programming Environment) (Execution Environment)


SLIM Single Linux Image Management (LinuxPilot 2004/04) On-demand construction of customized execution environments URL:


SLIM Utility computing decouples computing platforms (resources) and computing logic (applications) I.e., a single platform can run completely different applications Problem: different applications demand different execution environments (OS, shared libraries, supporting apps, etc.) SLIM is a network service for managing and constructing EE’s, and disseminating them to remote computing platforms

SLIM – System design: 

SLIM – System design How it works? A node sends a EE specification across the network to find the Boot server Boot server delivers the requested Linux kernel Image server constructs an EE by collecting shared libraries, user data, etc. Linux kernel boots, and contacts the Image Server to “mount” the EE via a file synchronization protocol such as NFS Aggressive caching techniques are deployed to optimize performance EE specification Linux kernel

InstantGrid on SLIM: 

InstantGrid on SLIM “pervasive utility computing” InstantGrid/SLIM On-Demand Grid Point Construction (Linux+ GT+ PBS+ Ganglia +MPI)

InstantGrid Performance: 

InstantGrid Performance To construct a 256-node grid point from scratch (PXE enabled) through Fast Ethernet in three (copy-if-needed) to five (full-copy to hard disk) minutes using one SLIM server To construct 256 standalone grid points takes longer time. The overhead mainly comes from the generation of host certificates.

SLIM – Ongoing and future work: 

SLIM – Ongoing and future work SLIM has been managing: the HKU CS grid point (350 nodes) for various grid research projects an addition 300+ lab machines for teaching purpose (different courses have different requirements) Future work To overcome the challenges in deploying SLIM over broadband links for realizing the “pervasive utility computing”

SLIM/InstantGrid – Key references: 

SLIM/InstantGrid – Key references R.S.C. Ho, C.M. Lee, D.H.F. Hung, C.L. Wang, and F.C.M. Lau, “Managing Execution Environments for Utility Computing,” Network Research Workshop, APAN 2004, July, 2004 R.S.C. Ho, K.K. Yin, C.L. Wang, and F.C.M. Lau, “InstantGrid: A Framework for Automatic Grid Point Construction,” The International Workshop on Grid and Cooperative Computing (GCC 2004), pp. 301-308, Oct 21-24, 2004, Wuhan, China. Download:


G-JavaMPI A grid-enabled Java-MPI system with dynamic load-balancing via process migration


G-JavaMPI A grid middleware that supports portable messaging-passing programming for achieving dynamic load-balancing and non-stop parallel computing in grid. Special feature: Transparent Java process migration State capturing and restoration through JVM Debugger Interface (JVMDI). No modification of JVM Facilitates more flexible task scheduling and more effective resource sharing. Avoids running hotspots. G-PASS: security enhancement for G-JavaMPI Performs identity mapping and access control while Java processes move across multiple grid points that are under different control policies. Avoids chain-delegation. Migration policies : Grid point CPU and network workload Application’s communication pattern Scheduled down time Data location

Preliminary Results at HKGrid: 

Preliminary Results at HKGrid Parallel BLAST program implemented by G-JavaMPI Three universities sharing CPU cycles and local bio-databases Executing 3 Blastp programs concurrently, total 18 processes Original no. of nodes: 5; 2 nodes join then 2 nodes quit The size of the migrated execution context is about 2.1 Kbytes. Total execution time : 566~911 seconds under different scheduling policies. PolyU HKU2 HKU1 CityU Single process migration is less than 0.5% of the total execution time under different CPU load. New node Migration Overhead Analysis

G-JavaMPI – Key references: 

G-JavaMPI – Key references Lin Chen, Tianchi Ma, Cho-Li Wang, Francis C.M. Lau, and Shanping Li, ``G-JavaMPI: A Grid Middleware for Transparent MPI Task Migration,'' to appear in Engineering the Grid: Status and Perspective, Nova Science Publisher. Tianchi Ma, Lin Chen, Cho-Li Wang, and Francis C.M. Lau, ``G-PASS: Security Infrastructure for Grid Travelers, The International Workshop on Grid and Cooperative Computing (GCC 2004), pp. 301-308, Oct 21-24, 2004, Wuhan, China. L. Chen, C.L. Wang, and F.C.M. Lau, “A Grid Middleware for Distributed Java Computing with MPI Binding and Process Migration Supports,” Journal of Computer Science and Technology (China), Vol. 18, No. 4, July 2003, pp. 505-514. Ricky K. K. Ma, Cho-Li Wang, and Francis C. M. Lau, ``M-JavaMPI : A Java-MPI Binding with Process Migration Support,'' The Second IEEE/ACM International Symposium on Cluster Computing and the Grid (CCGrid 2002), Berlin, Germany.


JESSICA2 “Java-Enabled Single-System-Image Computing Architecture”, project started in 1996. First version (JESSICA1) in 1999


JESSICA2 JESSICA2 is a distributed Java Virtual Machine (DJVM) which consists of a group of extended JVMs running on a distributed environment to support true parallel execution of a multithreaded Java application. Java threads can freely move across node boundaries and execute in parallel to achieve more scalable high-performance computing using clusters The JESSICA2 DJVM provides standard JVM services, that are compliant with the Java language specification, as if running on a single machine – Single System Image (SSI).

JESSICA2 Architecture : 

JESSICA2 Architecture Thread Migration Global Object Space JESSICA2 JVM A Multithreaded Java Program JESSICA2 JVM JESSICA2 JVM JESSICA2 JVM JESSICA2 JVM JESSICA2 JVM Master Worker Worker Worker Worker Worker JIT Compiler Mode Portable Java Frame

JESSICA2 Main Features: 

JESSICA2 Main Features Transparent Java thread migration Runtime capturing and restoring of thread execution context. No source code modification; no bytecode instrumentation (preprocessing); no new API introduced Enable dynamic load balancing on clusters Full Speed Computation JITEE: cluster-aware bytecode execution engine Operated in Just-In-Time (JIT) compilation mode Zero cost if no migration Transparent Remote Object Access Global Object Space : A shared global heap spanning all cluster nodes Adaptive migrating home protocol for memory consistency + various optimizing schemes. I/O redirection

Ray Tracing on JESSICA2 (64 PCs): 

Ray Tracing on JESSICA2 (64 PCs) Linux 2.4.18-3 kernel (Redhat 7.3) 64 nodes: 108 seconds 1 node: 4402 seconds ( 1.2 hour) Speedup = 4402/108=40.75

JESSICA – Key references: 

JESSICA – Key references Wenzhang Zhu, Weijian Fang, Cho-Li Wang, and Francis C.M. Lau, ``High Performance Computing on Clusters : The Distributed JVM Approach,''  to appear in High Performance Computing: Paradigm and Infrastructure, John Wiley & Sons, Inc. 2004. W.Z. Zhu , C.L. Wang, and F.C.M. Lau “A Lightweight Solution for Transparent Java Thread Migration in Just-in-Time Compilers,” The 2003 International Conference on Parallel Processing (ICPP-2003), pp. 465-472, Taiwan, Oct. 6-10, 2003 W.Z. Zhu, C.L. Wang and F.C.M. Lau, “JESSICA2: A Distributed Java Virtual Machine with Transparent Thread Migration Support,” IEEE Fourth International Conference on Cluster Computing (CLUSTER 2002), Chicago, USA, September 23-26, 2002, pp. 381-388.  M.J.M. Ma, C.L. Wang, F.C.M. Lau. “JESSICA: Java-Enabled Single-System-Image Computing Architecture,” Journal of Parallel and Distributed Computing, Vol. 60, No. 10, October 2000, pp. 1194-1222.


Summary Performance SLIM and InstantGrid: for high-speed construction of Grid computing environment, establish extensible grid platforms G-JavaMPI and JESSICA : Process/thread migration enables performance optimization and load balancing Reliability Java checkpointing (G-JavaMPI and JESSICA) SLIM helps construct platforms for failover Convenience G-JavaMPI and JESSICA enable users to utilize HPC facilities distributed across institutions via traditional means (e.g., message passing, Java) SLIM and InstantGrid fulfill on-demand Grid point construction, and simplify Grid point management.

What Next?: 

What Next? Grid computing today is like a “pot luck” supper Everyone brings and contributes a dish And … surprise! There really is “no free lunch” Everyone shares some of the costs Is it worth it?


To minimize the “surprises” (quality of service) Let the pros - the chefs - do it You sit back, relax, and enjoy, and pay for and only for what you consume Grid now is a private club Eventually it should be like … Starbucks Ubiquitous Invisible (the machinery behind) It’s my “cup of coffee”

The “pervasive grid” – everyone’s club: 

The “pervasive grid” – everyone’s club The grid (invisible computing) Thin clients “To use a computer is fun, but not to manage it”


Thanks! For more information: The HKU Systems Research Group Hong Kong Grid Grid Computing Research Portal

authorStream Live Help