Grid Computing 7700Fall 2005Lecture 1: About Grid ComputingGabrielle [email protected]://www.cct.lsu.edu/~gallen/TeachingNumerical Relativity Black holes, neutron stars, supernovae, gravitational waves Governed by Einsteins Equations: very complex, need tosolve numerically 10 coupled mixed elliptic-hyperbolic PDEs, thousands ofterms High fidelity solutions need more research innumerics/physics … but also larger computers, betterinfrastructure Physics currently limited by information technology!Numerical Relativity Good motivating example for Grid computing:– Large varied distributed collaborations need towork together– Need lots of cycles, storage (currently usingteraflops, terabytes)– Need to share results, codes, parameter files, …– Need advanced visualization, steeringwww.Googlism.com for grid computing grid computing is the next big thing? grid computing is one of the hottest areas in computing today grid computing is a flaky concept grid computing is the next evolutionary wave after the internet grid computing is an approach to distributed computing that spans locations as well as organizations grid computing is computing as a utility grid computing is still a few years off grid computing is like throwing darts at a dragonfly grid computing is no longer just a buzzword grid computing is still an emerging academic research field grid computing is to deliver computing and storage resources like utility companies deliver power grid computing is a way to create a virtual supercomputer by connecting large numbers of pcs indifferent locations over a shared network grid computing is going to enable things we can't even imagine grid computing is based on the concept of coordinated shared use of computers grid computing is applying the resources of many computers in a network to a single problem at thesame time grid computing is an emerging technology that allows seamless sharing of geographically distributedcomputational resources and networked instrumentation grid computing is an advanced approach to distributed computingGrid Computing Has also been called: metacomputing, scalablecomputing, global computing, internetcomputing What is it?– Distributed computing?– High performance computing?– Next generation internet?– Collaborative tools?– Open set of standards?Grid Computing“coordinated resourcesharing and problemsolving in dynamic,multi-institutional,virtual organizations”Foster, Kesselman, Tuecke, “The Anatomyof the Grid”, 2000.Grid Applications Grid Computing was originally about extendingscientific parallel computing on single machines todistributed systems In the end success depends on its application:– Complex problem solving– Large diverse collaborations– Gigantic data sets– Cybersecurity and cyberinfrastructure– Ultra fast networks– EconomicsAnalogies: DistributedInfrastructures Road, rail, telephones, power, banking, water, … All started locally, then regionally, thennationally, finally internationally Provide core infrastructure but cater todifferent end user devices Each provide reliable low cost access to astandardized service Available to the masses! Have hierachies and caching– Powerstation power lines to 110V home sockets– Interstates to dirt lanesFor Example: Electric Power Grid Single entity providing power to billions devices Relatively efficient, low cost, reliable US Grid links 10K generators Complex physical connections and trading mechanisms Components heterogeneous and operated/owned by differentcompanies Consumers differ in amount of power they use, the QoS theyrequire, and the price they will pay Economics important: grid driven by economic factors. Reservecapacities, trading power. Politics important: success depended on regulatory, political andinstitutional developments as much as technical innovation Control important: infrastructure for monitoring, management andcontrolGrid ChecklistA Grid … Coordinates resources that are not subject tocentralized control Uses standard, open, general purpose protocolsand interfaces Delivers non-trivial qualities of serviceIan Foster, “What is the Grid? A Three Point Checklist”, 2002Quality of Service Data management Resource discovery and information Authentication and authorization Accounting and tracking Job management Response time, security, throughput, …CCT GridGumbo Grid~10 mixed linux boxes for developmentCCT Condor FlockWorkstations aggregatedfor cycle harvestingSupermike1024 ProcessorIA32 Linux ClusterNemeaux64 ProcessorG5 MacOSX ClusterHelix256 ProcessorIA32 Linux ClusterMixture of machines, joinedby campus network … have incommon CCT administration.LONILouisiana Optical Network InitiativeTeraGridOpen Science Grid The Open Science Grid is a US grid computinginfrastructure that supports scientific computingvia an open collaboration of science researchers,software developers and computing, storage andnetwork providers. http://www.opensciencegrid.orgBasic Grid ModelGrid ApplicationsApplication Grid MiddlewareCommon InfrastructureGlobal ResourcesNew devicesSensorsWirelessCommonpoliciesGridEconomyGlobalnetworksCommon Infrastructure Most common core Grid infrastructure deployedtoday is called “Globus” Many higher level services are being researchedand built using Globus www.globus.org Originally from ANL/ISIThe Global Grid Forum Standards and best practices Promoting Grid technologies and applications Modeled around bodies such as IETF (internet engineeringtask force) Working groups and research groups in many different areas Meet 3 times a year http://www.gridforum.orgSome Application Areas Life sciences– Computational biology,bioinformatics, genomics– Access, collecting and miningdata, imaging Engineering– Aircraft design, modeling andmonitoring Data– High energy physics,astronomy Physical sciences– Numerical relativity, materialscience, geoscience Collaborations– Sharing, real timeinteractivity, visualization,communication Commercial– Gaming, idle workstations,climate predication, disaster,cyber security, portals Education and distancelearningSome Application Types Minimal communication (embarrassingly parallel) Staged/linked/workflow Access to Resources Fast throughput Large scale Adaptive