The UK National Grid Service and EGEE
Total Page:16
File Type:pdf, Size:1020Kb
Enabling Grids for E-sciencE The UK National Grid Service and EGEE David Meredith and Andrew Richards 10 May 2007 Manchester www. eu-egee.org EGEE-II INFSO-RI-031688 EGEE and gLite are registered trademarks Mission of the UK NGS Enabling Grids for E-sciencE The Mission of the National Grid Service • To provide coherent electronic access for UK researchers to compp,utational and data resources, and to (scientific) facilities required to carry out their research. • Do this Independently of resource or user location. EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 2 Who is the NGS? Enabling Grids for E-sciencE • The National Grid Service, funded by JISC, EPSRC and CCLRC ((),now STFC), was created in October 2003 and the service entered full production in September 2004 • The NGS is led and coordinated by the STFC in collaboration with the University of Manchester, the University of Oxford, the University of Edinburgh and the White Rose Grid at the University of Leeds EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 3 History of the UK NGS Enabling Grids for E-sciencE • 2001 - e-Science Grid Program started in UK – Inc. GridPP and others • 2003 - Initial grid service ITT – 4 independent clusters to investigate provision of a grid service • Apr il 2004 - NGS pre-prodiduction servi ce – EGEE, GridPP-2 • August 2004 – GOSC proposed – Coordinating NGS and providing central services • September 2004 - NGS production service / GOSC • April 2006 – NGS/GOSC phase 1 review • May 2006 - NGS phase-2 approved – More integrated programme – EGEE-2 started in April • Oct ob er 2006 – NGS p hase-2 EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 4 NGS Key Themes Enabling Grids for E-sciencE • Provide the users what they need – aim to do this with as low an ‘access barrier’ as possible • “Content” – Services user driven in demand (data + computation services at present, but also bring online additional services such as visualisation). • NGS will only grow if both the content + community grows – Not the job of 1 person or organisation – Job of everyone – Being part of EGEE helps increase community awareness and overall benefits. EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 5 NGS and EGEE Relationship Enabling Grids for E-sciencE • NGS roadmap includes interoperability with EGEE (and others e.g. Teragrid) • NGS/EGEE people are the same – NGS staff provide matching effort for EGEE related work – collaborating and sharing experience across projects where possible. – UKIROC and N GS Support Centre and mostly the same people • Current NGS middleware is not ‘EGEE’ but a subset – minimal set of common middleware to enable NGS grid to grow • Interoperate fully now in areas such as GGUS • Utilise tools such as GOCDB • Adopt parts of ‘EGEE’ for NGS use – e.g. Resource Brokers (working with EGEE on parallel job submission) – VOMS (using GridPP / LCG knowledge at Manchester) – Accounting – Monitoring – SFT Tests. EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 6 NGS Management Structure Enabling Grids for E-sciencE GridPP/EGEE NGS Board Directors JISC e-Infrastructrure Members Advisory Board Review Group NGS Technical Board Operations Board Development Board User Forum Work Package Work Package Work Package Engineering Task Force External Body Reports to NGS Work Package EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 7 Enabling Grids for E-sciencE Using the NGS EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 8 Resources available to users Enabling Grids for E-sciencE • NGS core nodes (()Linux IA32 clusters) • data nodes at RAL and Manchester – each 15 TB, 40 processors • 128p compu te nodes at O xford and Leeds • PHASE 2 NOW BEING INSTALLED! National HPC services • HPCx • free at point of use • HeCToR (expected 2007) – Users apply through NGS web site Must apply separately to – ~500 registered users Research Councils – accept conditions of use – light-weight peer review 1-2 weeks • access is through digital X.509 certificates – from UK e-Science CA – or recognized peer EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 9 NGS Compatibility Enabling Grids for E-sciencE In addition to core sites, resource providers can also join the NGS by : Two levels of membership: 1. Affiliates – run compatible NGS minimum software stack (3 levels of conformance) – verify with compliance tests – adopt NGS security policies – Providinggpp support and securit y contacts – Providing site-specific information to users, through, but not limited to, the NGS web site 2. PtPartners also – make “significant resources” available to NGS users – enforce NGS acceptable use policies – provide accounting information – define commitments through Service Level Descriptions – Gain influence + NGS direction through representation on NGS Technical Board EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 10 NGS-2 Core Hardware Enabling Grids for E-sciencE Compute Nodes – 48 nodes with Dual Socket Dual Core AMD Opteron 280 processors, 8GB memory 2x80GB disks, Myrinet 2000 – 8 nodes with Quad Socket Dual Core AMD Opteron 280 processors, 32GB memory 2x80GB disks, Myrinet 2000 Storage – 8 storage nodes Dual Socket Dual Core AMD Opteron 280 processors, 8GB memory 2x80GB disks, Myrinet 2000, Fibre HBA – 5 x 12 TB Infotrend Storage Arrays – Qlogic 5200 SANbox Initial BenchMarks – Initi al testi ng usi ng HPL (t op 500 gi gafl ops t est er) on 240 Cores gives 934Gflops (81% peak). The full system should have 256 Cores available but some nodes have hardware problems EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 11 NGS and Partner sites 2007 Enabling Grids for E-sciencE EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 12 Acceptable Use Policies Enabling Grids for E-sciencE • NGS currently has its Terms and Conditions of Use that all users have to accept – http://www.ngs.ac.uk/NGS-tacu.html • We are looking towards adopting the VO Policies of EGEE/LCG/GridPP/… (something common) for VO’s that we host and support on our VOMS infrastructure, to be consistent. • For support of external VO’s, if we accept the VO, then the VO ‘Manager’ has to accept our TACU , but not each individual member. EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 13 Access Mechanisms Enabling Grids for E-sciencE • Direct GSISSH login to site head nodes and UI • Globus tools (e.g. globusrun) • Portals ( NGS Application Repository, P-Grade Portal ) • Resource Brokers / UI machines • AHE (App lica tion Hos ting Env ironmen t) • GridSAM (from OMII-UK, hosted b y Belfast) . • ….your own custom application … EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 14 GSISSHTerm Enabling Grids for E-sciencE • http://www.ngs.ac.uk (Grid Utilities) EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 15 Resource Brokers Enabling Grids for E-sciencE • EGEE gLite RB (+ UI) – The NGS Resource Bro ker an d User In ter face is base d on the g Lite 3. 0 WMS-LB and UI nodes from the EGEE Project. The UI is accessible to all members of the NGS by GSISSH on port 2222 at the following node: ngsui01.ngs.rl.ac.uk • Gridway –http://asds.dacya.ucm.es/GridWay/ –MtMetasc hdliheduling tec hno log ies for the grid EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 16 NGS Applications Repository Enabling Grids for E-sciencE https://portal.ngs.ac.uk EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 17 NGS App Repository: https://portal.ngs.ac.uk Enabling Grids for E-sciencE 1. A portal to browse applications described in JSDL (personal and shared). • JSDL can be browsed for, selected and loaded in order to run applications on the Grid (loaded either ‘out-of-the-box’ or, more usually loaded and modified/tweaked as required). • JSDL can be searched for by category of interest in the portal (e. g bioinformatics, chemistry, tutorials/examples). • JSDL documents can be pre-configured and published by domain experts and made available to all other users (users benefit from expertise captured in JSDL). 2. A JSDL GU I ed it oor for co n str uctin g, v ali datin g, sh arin g, upl oadin g jobs described in JSDL. 3. A Grid job submission and monitoring application (currently, only Globus but more Grid middleware providers are being added, e. g. GridSam/WSRF). 4. Portal is designed to be generic and not tied to any particular set of Grid technologies. Extend portal to support more middleware (SAGA / GridSAM, and Grid + Web protocols for staging). EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 18 NGS App Repository: https://portal.ngs.ac.uk Enabling Grids for E-sciencE Active Job Detail Input fields are already filled out for pre-configured applications. Changes to the parameters in the GUI will update the generated JSDL automatically. Input fields are taken from the JSDL and JSDL- POSIX extension schemas. EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 19 NGS App Repository: https://portal.ngs.ac.uk Enabling Grids for E-sciencE Environment Variables / Arguments Add required env vars, e.g. ‘NGSMODULES’ – used to configure application environment Paste and parse command line arguments (space and/or line separated values) EGEE-II INFSO-RI-031688 UK National Grid Service and EGEE: OGF20 20 NGS App Repository: https://portal.ngs.ac.uk Enabling Grids for E-sciencE Named file systems used Named File Systems to declare mount points that are required on the consuming system.