180 likes | 292 Views
ATLAS Computing. Introduction + Overview. Norman McCubbin (RAL) Computing Co-ordinator. ATLAS Computing: Introduction + Overview. Organisation: people/task matrix Main milestones Summary of major activities “Projectisation” of Computing: schedule, PBS;
E N D
ATLAS Computing. Introduction + Overview Norman McCubbin (RAL) Computing Co-ordinator McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • Organisation: people/task matrix • Main milestones • Summary of major activities • “Projectisation” of Computing: • schedule, PBS; • effort; ‘holes’; software agreements; • hardware resources: • National Computing Board; World-Wide Computing; • MONARC; CERN Review • Conclusions McCubbin: DoE/NSF Review 16Nov00
Computing organization Comp. Oversight Board National Comp. Board Comp. Steering Group Physics Technical Group Event filter QC group simulation reconstruction database Arch. team simulation reconstruction database coordinator Detector system McCubbin: DoE/NSF Review 16Nov00
Offline Coordinator Reconstruction Simulation Database Chair N. McCubbin D. Rousseau K.Amako D.Malon/ RD Schaffer Inner Detector D. Barberis D. Rousseau F.Luehring S.Bentvelsen Liquid Argon J. Collot S.Rajagopalan M. Leltchouk S.Simion/R.Sobie Tile calorimeter A. Solodkov F. Merritt A. Solodkov T. LeCompte Muon G. Poulard J.F. Laporte A. Rimoldi S. Goldfarb T. Hansl-Kozanecki S. Tapprogge S. George LVL2 trigger H.P. Beck Trigger/DAQ Event Filter V. Vercesi F. Touchard ATLAS Detector/Task matrix Physics Co-ordinator: F.Gianotti Chief Architect: D.Quarrie McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • Some aspects of organisation still need to be developed: • software ‘support’: librarian, Quality Assurance, documentation; • World-Wide Computing: already have working-groups (WWC, Regional Centres, GRID, MONARC ). Need to develop some organisational structure to deliver Mock Data Challenges etc. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction and Overview • The major schedule items (with indicative dates) for the next 2-3 years are: • Develop framework (Athena) through series of releases, leading to first ‘fully-functional’ version by end 01. • Profit from TestBeam requirements, starting this year. • Major effort to validate Geant4 for ATLAS, with simulation workshop Dec. 2000 and Physics workshop Spring September 2001. • Have first version of reconstruction software ready (start 01) for Event Filter analysis for TDAQ TDR. • MDC0: 100k event ‘continuity test’ for end 01. • Chain (G4 simulation and ATLAS code) ready for MDC1 (“0.1% PB”) to start in 02. (Open issues: how to ‘trigger’; involvement of Regional Centres (RC)) • Computing TDR: deliver end 02. (Open issues: date) • MDC2: First half 03. “10% of 1PByte”. Must, inter alia, validate computing model. (Open issues: how to ‘trigger’; scale; collaboration with other experiments.) McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • Simulation: • major effort is focussed in Geant4 (G4) physics validation for ATLAS; • very detailed studies (G4/G3/Test-Beam) are now in progress: G4 looks promising but of course needs tuning/refining. ATLAS results presented at several conferences/work-shops. • Close and productive collaboration with G4 team (monthly meetings); • ATLAS G4-physics workshop Dec. 4+5; • major goal is G4 ‘good enough’ for Mock Data Challenge 0 (end 2001). G3 as ‘insurance’? McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • Reconstruction: • first target is to have C++ versions of reconstruction working via framework (Athena) for end of this year; muon code will be MuonBox (Fortran); • Lar code already available via Athena; MuonBox and xKalman++ have been interfaced ‘privately’ to Athena, some tidying still needed. • these versions of recon. codes will be used (on existing Geant3 data) for TDAQ TDR. (Maybe later versions, depending on TDAQ TDR date); McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • ‘Projectisation’ of Computing: • Scale, scope, duration, and cost of ATLAS computing project all argue for a greater formal ‘projectisation’ of computing than has been done in previous experiments (LEP, HERA, UA1,2 , etc.) • Indeed, my time as Computing Co-ordinator has strengthened my conviction that this is not only desirable but essential. • Details of our planning etc. so far can be seen at: http://cern.ch/Atlas/GROUPS/SOFTWARE/OO/planning • Considerable effort to harmonise with USA at PBS level to avoid duplication and confusion. (See presentations of Meinhard and Wenaus) • We have prepared a first version of a schedule, which includes the major milestones mentioned above, but also gives much much more detail. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • This schedule is by no means ‘final’ or complete. (e.g. ‘World-wide computing’ not yet entered in detail, and ‘dependencies’ need much more detail). But it does give significant detail through to end 2001, and is already proving valuable for Computing Steering Group (CSG): part of each (monthly) meeting is devoted to schedule ‘follow-up’. • A first set of LHCC and ATLAS Exec. Board (EB) milestones has been defined. This list will surely grow. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing; Introduction + Overview • PBS is used to estimate effort required to deliver the computing project. • It is clear that the entire project (including ‘core’ software and detector-specific software) requires approx. 100 ‘full-time-equivalent’ (fte) per year from now until…… (see next slide) • It is also clear that we have some urgent and obvious ‘vacancies’ right now. • Hope that ‘software agreements’ (see later) will help to fill these. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • ‘Core’ effort estimate (fte’s): • 2000 2001 2002 2003 2004 2005 • sub-T O T A L (1-8) 30.2 33.4 30.6 24.2 23.3 23.4 • ====================================== • T O T A L (1-12) 34.6 47.2 46.6 42.2 39.3 39.4 • (1-8) includes framework, dBase, core management,… • (1-12) adds in world-wide computing, GRID, MDC’s. • N.B. above is requirement, we don’t have it all! McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • It has also been recognised for some time that computing (hardware definitely, software probably) will require, and benefit from, Computing MoU. • This emphasises the point that CERN cannot provide the major share of these resources; they have to come from the other institutes. (CERN will provide some, of course..) • More on the hardware side in a moment; on the software side the main benefit (in my opinion) of an MoU is to commit institutes in the same way as they feel committed by detector MoU. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • Since the signing of MoU is some way off, and because the whole idea is new for computing, we have been developing the idea of “software agreements”, which could, eventually, form part of Computing MoU. • The policy paper on software agreements has been approved by the National Computing Board, and by the ATLAS CB in June 2000. • We have just launched, through the NCB, a list of tasks which could be covered by ‘core’ software agreements. • Idea of software agreement (swa) is to get formal institute commitment. Thus it helps ATLAS; it may also help ATLAS members in arguing for resources in their institute. • Hope is to get first swa (Control Framework with USA) signed by end of the year. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • On hardware resources, there is a lot of work going on, stimulated by the following (related) points: • as mentioned above, much of the money (and effort) for computing hardware and infrastructure will have to come from outside CERN; scale is such that Funding Agencies will have to be asked ‘officially’ for support; • a possible model for ‘World-Wide Computing’ was discussed already at the time of the ATLAS and CMS Computing TPs: major computing centres (‘Regional Centres’) in a few regions which act as computing ‘hubs’ for their region(s). Thus you do not ‘copy from CERN’ every time; required data is copied to your Regional Centre ‘once’, etc. etc. • This thinking has been elaborated in considerable detail in the MONARC project. ( Participation from all LHC experiments) • It is also now clear that estimates of resource requirement will be one of the most important outcomes of the CERN Review. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • H.Hoffmann launched idea of CERN Review of LHC Computing in second half of 1999. Announced at Marseille workshop in September 99. • Aims: • assess state of work and preparations in experiments and IT; • provide input for Computing MoUs, which should be put in place in 2001 (?); • recommend actions and in particular common actions between experiments and IT Division that will help to achieve desired goals within existing resources; • first draft of report December 2000 (?). • See H.Hoffmann talk yesterday(?). McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • Much of the subject matter discussed by CERN Review is also of prime concern to the ATLAS National Computing Board (NCB), Chairman: Alois Putzer (Heidelberg) • Stimulated in part by CERN Review and various GRID activities, NCB has now formed w/groups which, as a first step, will gather information (‘what’s going on’) on GRID, Regional Centres etc. • For several (related) reasons the whole topic of ‘world-wide computing’ has, rightly, moved up the agenda in the last few months. • In ATLAS we have used a plausible estimate of event size (1MB) and trigger rate (100Hz) for several years. (As have CMS) Following the very detailed Physics (TDR) and Trigger (TP) studies, we have much more refined and understood estimates…. But they are higher (‘270x2’ at 10^33). This is being addressed, vigorously. McCubbin: DoE/NSF Review 16Nov00
ATLAS Computing: Introduction + Overview • There has been a lot of progress since the DoE/NSF review in January: • framework/architecture • Reconstruction Code • Simulation • organisational infrastructure (planning etc.) • And the USA contribution has been very significant. • ATLAS Computing effort needs your continuing and ongoing support. McCubbin: DoE/NSF Review 16Nov00