Introduction to Grid Computing with High Performance Computing Mike Griffiths White Rose Grid e-Science Centre of Excellence

Introduction High Performance Grid Computing e-Science The Evolving Grid The Local Compute Node Iceberg Registration Outline

Objectives What is grid computing? How the grid assists with problem solving lifecycle Identify and Explain Buzzwords Remove Hype

Problem solving lifecycle Problem definition and requirements capture Model development –Languages (FORTRAN, C, C++, Java etc.) –Model Building SDK’s –Matlab and clones –Packages (ANSYS, FLUENT, CFX)

Problem solving lifecycle Problem solving environment –specialized software for solving one class of problems –Application user interface, portal Model testing –Validation, verification Results production –Scheduling tasks over the grid Analysis and Visualisation

Grid Technologies

Simulation of large complex systems Large scale multi site data mining, distributed data sets Shared virtual reality Interactive collaboration Real-time access to remote resources. Grid Technologies

What Is Grid Computing Virtualisation of resource Increase processing power Secure and flexible collaboration The Grid Problem

Electric Power Generation Analogy Information Generators Information Distributed Over the Grid Customer Access to Information Grid

Pcwebopedia.com A form of networking. Unlike conventional networks that focus on communication among devices, grid computing harnesses unused processing cycles of all computers in a network for solving problems too intensive for any stand-alone machine.networkingprocessing

IBM Definition Grid computing enables the virtualization of distributed computing and data resources such as processing, network bandwidth and storage capacity to create a single system image, granting users and applications seamless access to vast IT capabilities. Just as an Internet user views a unified instance of content via the Web, a grid user essentially sees a single, large virtual computer.

Sun Microsystems Grid Computing is a computing infrastructure that provides dependable, consistent, pervasive and inexpensive access to computational capabilities.

“The Grid Problem” “Grid problem,” flexible, secure, coordinated resource sharing among dynamic collections of individuals, institutions, and resources—what we refer to as virtual organizations. –From “The Anatomy of the Grid” by Foster, Kesselman and Tuecke.

Virtual Organisations

Grid Characteristics Networks – High Bandwidth Computing - Tflops Data storage Peta byte The Grid

Cluster Grid Beowulf clusters Enterprise Grid, Campus Grid, Intra-Grid Departmental clusters, servers and PC network Utility Grid Access resources over internet on demand Global Grid, Inter-grid White Rose Grid, National Grid Service, Particle physics data grid Types of Grids

Three Uses of Grid Computing Compute grids Data grids Collaborative grids

Distributed Supercomputing Compute Clusters –Schedulers sun grid engine, pbs Grid aggregates computational resources to compute large complex problems Fast networks enabling true parallel computation and shared memory processing Select compute resources according to Time and Financial constraints

Architectures for High Performance Computing Supercluster –e.g. Blue Gene (65536 dual processors in 64 cabinets) Clusters –e.g. iceberg –Parallel applications using MPI Symmetric multiprocessors –e.g. 4 processor shared memory V40 node on iceberg –Shared memory programming Open MP Vector Processor –E.g Amdhal VP at MCC (80’s and 90’s)

High Throughput Applications Problems divided into many tasks –Grid schedules tasks –The mother projects –Spin off for companies such as Entropia and United Devices projects Condor –Cycle scavenging from spare PC’s

Statistics for SETI at Home (13/09/2004) TotalLast 24 Hours Users Results received Total CPU time years years Floating Point Operations e e+19 ( TeraFLOPs/sec) Average CPU time per work unit 11 hr 41 min 24.2 sec6 hr 46 min 10.6 sec

Most Promising Candidates

Grid Types Data Grid Computing Network stores large volume of data across network Heterogeneous data sources Engine flight data Airline Maintenance Centre European data center London Airport New York Airport American data center Grid Diagnostics centre

Grid Types - Collaborative Internet videoconferencing Collaborative Visualisation

e-Science More science relies on computational experiments More large, geographically disparate, collaborative projects More need to share/lease resources –Compute power, datasets, instruments, visualization

e-Science Centres Centres of Excellence Regional Centres

e-Science Organisations National e-Science Centre –To stimulate and sustain the development of e-Science in the UK, to contribute significantly to its international development and to ensure that its techniques are rapidly propagated to commerce and industry. Open Middleware Infrastructure Institute –Repository for UK Grid Middleware

e-Science Requirements Simple and secure access to remote resources across administrative domains Minimally disruptive to local administration policies and users Large set of resources used by a single computation Adapt to non-static configuration of resources

The Evolving Grid

Comprising of two data clusters and two compute clusters. Offer a significant resource for the UK e-Science community. Clusters are located at – Manchester (data cluster), – Oxford (compute cluster), – CCLRC (data cluster) and – White Rose Grid (compute cluster). More sites –Lancaster –Wesc –Bristol

EGEE The EGEE project brings together experts from over 27 countriesexperts – Build on recent advances in Grid technology. –Developing a service Grid infrastructure in Europe, available to scientists 24 hours-a-day.

Available Grid Services Access Grid White Rose Grid –Grid research –HPC Service National Grid Service –Compute Grid –Data Grid (SRB) National HPC Services –HPCx and CSAR (part of NGS) Portal Services

Sheffield Grid Node: Hardware AMD based supplied by Sun Microsystems Processors: 320 Performance: 300GFLOPs Main Memory: 800GB Filestore: 9TB Temporary disk space: 10TB Physical size: 8 racks Power usage: 50KW

Sheffield Grid Node: Hardware,part Processors Grid pp community 160 Processors General Use –20 x V40 each with 4x64 bit AMD Opteron (2.4GHz) and 16GB shared main memory. –40 x V20 each with 2x64 bit AMD Opteron (2.4 GHz) and 4GB shared main memory Comparing L2 Cash –AMD Opteron 1MB –Ultrac sparc III Cu (Titania) 8MB

Sheffield Grid Node: Hardware, part 3 Inside a V20 unit.

Sheffield Grid Node: Hardware 4 Two main Interconnect types gigabit (commodity), Myrinet (more specialist) –Gigabit – Supported as standard good for job farms, and small to mid size systems –Myrinet – High End solution for large parallel applications has become defacto standard for clusters (4Gb/s)

Sheffield Grid Node: Hardware 64bit v 32 bit –Mainly useful for programs requiring large memory – available on bigmem nodes –Greater Floating Point accuracy –Future-proof: 32-bit systems are becoming obselete in HPC

Sheffield Grid Node: Software 1 Opteron Redhat 64bit Scientific Linux Portland, GNU DDT MPICH Sun Grid Engine v6 Ganglia

Sheffield Grid Node: Software 2 Maths and Statistical –Matlab7.0, scilab 3.1 –R Engineering and Finite Element –Fluent , and als gambit, fidap and tgrid –Ansys v90 –Abaqus –CFX –DYNA 91a Visualisation –IDL 6.1 –OpenDX

Sheffield Grid Node: Software 3 Development –MPI, MPICH-gm –OpenMP –Nag, 20 –ACML Grid –Globus (via gpt 3.0) –SRB s-client tools to follow

Registration Local User Account Obtain an e-Science Certificate Register with the White Rose Grid Apply for NGS Resource Go to the link

Why obtain an e-Science Certificate Enables secure single sign on to the White Rose Grid Use portals e.g. the WRG Application portal Access WRG, NGS, egee

For More Information The White Rose Grid – The National e-Science Centre – The Globus Project™ – Global Grid Forum –

Grid Computing References The Grid: Computing Without Bounds –Ian Foster, Scientific American, April “The Anatomy of the Grid” – Grid Services – “The Physiology of the Grid” – 22.pdf Research Agenda for the Semantic Grid –