Project Status Report : SAMGrid

1
Project Status Report SAMGrid

• SAMGrid Management, Status, Operations Merritt
• SAMGrid Development I. Veseli
• SAMGrid Development II. Kennedy
• SAMGrid Future Plans St. Denis

2
SAMGrid Project Description

• Purpose Provide data handling services to Run
  II experiments and other interested experiments
  with similar problems. These services should
  scale in performance and convenience for
  cataloging and delivery of Petabyte-sized
  datasets, and should evolve to availability in
  relevant Grid environments.
• Current stakeholders CDF, DØ, MINOS, CD
• Duration Development effort is expected to
  extend through 07 as the components move to
  become Grid services. High-level maintenance (
  I.e., effort that includes capability to respond
  to feature requests) is expected to continue
  through at least the data collection lifetime of
  the stakeholders.

3
The SAM-Grid Team

• Revised Management Plan went into effect Dec 03
• Project Co-Leaders Wyatt Merritt CD/DØCA
• Rick St. Denis CDF/ U Glasgow
• Project Technical Co-Leaders
• Rob Kennedy CD/CDFSinisa Veseli CD/DØCA
• CCF Andrew Baranovski, Gabriele Garzoglio, Igor
  Terekhov
• CEPA Carmenita Moore, Steve White (0.5 FTE)
• CDF Randy Herber, Art Kreymer, Stefan Stonjek
  (GS)
• DØCA Lauri Loebel Carpenter, Robert Illingworth,
  Adam Lyon

4
The SAM-Grid Team - Extended

• Database support (CSS-DSG)Diana Bonham, Anil
  Kumar
• Associated internal projectsRUNJOB (with
  CMS)Authorization Project (with CMS, still being
  defined)
• Associated external projectsPPDG Sankalp
  Jain, Aditya NishandarGridPP Morag
  Burgon-Lyon, Valeria Bartsch, Iain Bertram,
  Dave Evans, Peter Love SBIR II Matt
  Vranicar, Jeremy Simmons, Josh Gramlich,
  Ngan MacDonald, John Grace
•  

5
SAMGrid Project Management Organization

• Project co-leaders
• Represent largest stakeholders requirements
  priorities
• Run weekly design meetings
• Project technical leaders
• Run weekly operations meeting
• Conduct subproject assessments
• Active Subprojects C API, DBServer, JIM, H
  Stream Reco for CDF, Caching, ChainsLinks, CDF
  DFC, Test Harness, Linux deploy of DBServers,
  Config Man
• Planned Subprojects Request system, Autodest,
  Further monitoring (MIS)
• Related Subprojects d0tools, SBIR II, Condor
  mods, workflow packages for CDF D0,
  Authorization Accounting
• Recently completed Subprojects Python API, V5.1
  Schema Design, Batch Adapter, D0 Online dcache
  TDP, 1st Gen Monitoring Tools, Data Dimensions
  Grammar

6
SAMGrid Components

• Event/File Catalog for metadata (contents
  processing) and locations
• Dbservers for accessing catalog
• Station servers for file delivery to projects
• Optimizer
• File storage server
• Interface to station cache and MSS (samcp)
• JIM components for Grid job submission
  monitoring
• User API
• C client API

7
Status and deployments of SAMGrid For DØ

• Operational _at_ FNAL online, reco farm, d0mino,
  cab, new cab, clued0
• Operational _at_ Monte Carlo production sites
• Operational _at_ remote analysis sites 20 active,
  40 deployed
• Operational 11/03 2/04 for remote
  reconstruction IN2P3, UKGrid (Manchester/ICL/RAL)
  , WestGRID, GridKA, NIKHEF -- 97M events
  reprocessed remotely
• Stats 78K proj FNAL, gt14K proj remote (since
  1/1/03) 60 billion evts, 3 PB, 8 M files
  consumed (all D0 stations)

8
D0
9
D0 Files
4000-8000 Files/Day
10
D0 Files Per Month By Year
1999
2000
2001
2002
2003
100,000 files
Run II Start
11
D0 Total Files
2.5Million Files Served
12
D0 Total Data Moved
700TB moved
13

Status and deployments of SAMGrid For CDF

• Operational 24/7 to store online metadata
• Operational at remote stations 15 active, 30
  deployed Large recent increase Fla.
  Wkshp!
• In testing for Monte Carlo production
• File delivery tests up to 20 TB on testcaf
• Statistics 3000 proj total (since 1/1/03)
• Note CDF usage pattern is different from DØ CDF
  moves more GB (but not more events) because it
  does not use small summary format like DØ
  thumbnail.

14
CDF Florida DH Workshop
Now 20!

• 11 installations in about 2 hours. Integrated
  with dCAF in 2 cases in 2 days.
• 3 in Asia, 4 in Europe
• 6 sites committed to summer 2004 usage of their
  facilities for all of CDF (mostly MC)
• Sam installation now initsam cdf ltstationnamegt
• Follow-up on April 1.
• Each site has a local user support person to
  reduce load on core development team.
• Generally Security ate 80 of the effort!

15
(No Transcript)
16
2TB/Day Karlsruhe
17
CDF Dcache on CAF
ALL CDF on CAF reads 20TB/Day
18
(No Transcript)
19
(No Transcript)
20
Status and deployments of JIM

• Job broker, execution and submission site
  software, job monitor, client software for grid
  job submission
• Deployment plan for DØ Monte Carlo
• Test at 3 sites (Manchester, CCIN2P3, Wisconsin)
  with basic functionality and measure efficiency
  of job completion
• Verify use by experimenter for job submission
  (this week)
• Add merging
• Move to production at these 3 sites (DØ
  milestone Mar 1)
• Add remainder of DØ MC sites (Lancaster, SAR,
  NIKHEF, Prague)
• Improve brokering algorithm

Manchester CCIN2P3 Wisconsin
JIM eff gt99 gt99 gt99
Site eff x Code eff 85 60 60
21
JIM Issues

• Site operational requirements (e.g. clock synch,
  disk node reliability, OS issues)
• Experiment operational requirements (e.g. code
  footprint may exceed site capability and is
  variable w/ release)
• File transfer capabilities policies cf. mtg
  this week w/ GridKA rep
• Allocation of services to head node vs worker
  nodes
• Sandboxing mechanisms (last week design mtg)
• Merging mechanism, brokering (this week design
  mtg)

22
Operational Model

• Experiments provide shifters for 1st line problem
  fielding and solving
• Project provides on-call list from developers
• At DØ, on average 60 80 of problems are
  answered by shifters
• Classes of problems
• Routine jobs like adding info to database
• Less routine cleanup after failed stores
• Answering user questions regarding usage
• Updating documentation
• Investigating user reports of problems, and
  problems visible in project monitoring tools
• Providing solutions for problems

23
Operations Outlook

• Improve documentation with aim of improving
  shifter user ability to diagnose/solve problems
• Expect doubling of central station capacity at DØ
• Expect transition to more SAM usage at CDF
• Expect Grid operations in production for
  simulation, first at DØ then at CDF