USCMS

1
US-CMS User Facilities Vivian ODell FNAL
Software and Computing PMG May 11, 2001
2
Production Status in US

• US-CMS is responsible for reconstruction of 2M
  JPG PRS group events.
• Weekly coordination meetings between T1 (FNAL)
  and T2 (CalTech/UCSD).
• Many problems
• At FNAL we have had many hardware problems
• T2 have personnel issues
• At FNAL we have daily production meetings
• Keeps team focused
• Good collaboration between CMS/ISD/OSS/ODS

3
Tier 1 Production Status
4
CMS Cluster
WORKERS
SERVERS
GALLO,WONDER,VELVEETA, CMSUN1
Popcrn01 - popcrn40
5
Production Status in US

• Fermilab is hosting the JPG User Federation
• Being used by people at FNAL/Wisconsin for HLT
  studies
• Keep documentation up to date with federation
  contents
• Required AMS-Enstore interface for automatic
  migration of files
• See
• http//computing.fnal.gov/cms/Monitor/cms_producti
  on.html
• Production scripts written by FNAL being used by
  all of CMS (including CERN!)
• These scripts will allow us to make a distributed
  system.

6
FNAL/UF in International CMS

• It is essential we have good integration in
  international CMS
• Natalia Ratnikova (UF Code librarian) is also
  spending part of her time as CAS engineer
  working on SCRAM the CMS code distribution tool
• Greg Graham/Hans Wenzel spend part of their time
  on UF issues
• Pal Hidas (PPD-Guest Scientist working on HLT
  studies for JPG) is sitting with us.
• Mixing roles according to expertise really helps
  people stay integrated with international CMS

7
Hardware Spending Proposal

• We have put together a hardware spending proposal
  for T1 consistent with user and RD needs
• This was presented to the ASCB in April and
  general principles agreed to
• Some concern was raised that FNAL buys expensive
  stuff
• We have a meeting of CMS computing people this
  afternoon to hash out the details and start
  writing requisitions

8
Summary of Current UF tasks

• Digitization of JPG fall production with Tier 2
  sites
• New MC (spring) production with Tier 2 sites
• Hosting JPG User Federation at FNAL
• For fall production, this implies 4 TB storage
• Implies 1TB on disk, 3 TB in tape storage
• Hosting MPG User Federation at FNAL?
• For fall production, this implies 4 TB storage
• Implies 1TB on disk, 3 TB in tape storage
• Also hosting User Federation from spring
  production, AOD or even NTUPLE for users
• Objectivity testing/RD in data hosting

9
Hardware Equipment Needs

• Efficient use of CPU at Tier 2 sites so we
  dont need additional CPU for production
• Fast, efficient, transparent storage for hosting
  user federation
• Mixture of disk/tape
• RD for RAID/disk/OBJY efficient matching
• This will also serve as input to RC simulation
• Build RD systems for analysis clusters

10
Funding Request for 2001-2003
(This is the total with FNAL added overhead)
11
Funding Request for 2001
12
Funding Reality for 2001
13
Proposed Spending FY01

• 1.6.1 Test Systems 48,500
• Sun Netra-T1 3,500.00 (?)
• 1 TB RAID (ZYZZX) 45,000.00

14
Proposed Spending FY01

• 1.6.2.2.1 Prototype User Analysis System (RD)
  53,000
• Dell 6400 Server 16,000.00
• Powervault Storage Array
  7,000.00
• Eight worker nodes 30,000.00

15
Proposed Spending FY01

• 1.7.2 Disk support for User Federation 45,000
• 1 TB ZYZZX 45,000.00
• 1.6.2.2.3 Storage RD Systems 188,000
• 2 Dell 6400 Server 32,000.00
• 1 TB RAIDZONE 25,000.00
• 1 TB 3WARE 13,000.00
• 2 TB WINCHESTER 65,000.00
• Farm Test System 53,000.00
• Dell 6400 Server 16,000.00
• Powervault Storage Array 7,000.00
• Eight worker nodes 30,000.00

16
Proposed Spending FY01

• Networking 52,000
• WS-C6509-2500AC Catalyst 6509 Chassis w/2500W AC
  Power 9,300.00
• WS-CAC-2500W/2 Catalyst 6000 Second 2500W AC
  PS 2,600.00
• WS-X6K-S2-PFC2 Catalyst 6500 Supervisor
• Engine-2,2GE,plus PFC-2 16,000.00
• WS-X6516-GBIC Catalyst 6500 16-port GigE
  ModFabric
• Enabled 16,500.00
• WS-X6548-RJ-21 Catalyst 6500 48-Port 10/100
  RJ21Fabric
• Enabled 13,300.00
• --------------------------------------------------
  --------------------------------------------------
  ----------
• TOTAL will be 52,000.00
• (if we trade in a couple our old switches)

17
Funding Proposal for 2001
Some costs may be overestimated (but) also
we may need to augment our farm CPU
18
Selection of 2nd T2 prototype

• Method
• Level 1 SC PM sends a letter to the US-CMS
  collaboration requesting institutes interested in
  becoming a Tier 2 prototype site to send a letter
  of interest to the Level 2 US-CMS User Facility
  Manager.
• Lothar did this March 28, 2001
• (Matthias did this May 25, 2000)
• The letter of interest should address the
  selection criteria (see below)
• The User Facility Manager will review all of the
  proposals, request more information if necessary
  and write a report detailing the results and
  recommending a selection.
• I did this April 22nd
• This report then goes to the ASCB and the L1
  project manager for US-CMS Software and
  Computing.
• Sent draft to Lothar Irwin on April 23
• The L1 Project Manager in consideration of the UF
  report and with input from the ASCB then makes
  the final decision on the prototype center
  location.

19
Tier 2 Selection

• Selection Criteria (as detailed in letter from
  Lothar to US-CMS)
• Activity in GRID research
• Amount of personnel and degree of involvement in
  RD projects (GriPhyN, PPDG, Globus, testbeds)
• Existing resources that can be leveraged, i.e.
  local and wide area network connectivitiy,
  availability of support staff, and existing
  hardware such as mass storage systems,
  processors, disk storage

20
Tier 2 Selection

• Results and Responses
• 6 universities responded in May, 2000
• 2 of the original 6 responded in April, 2001
• I considered all 6 and used updates where I had
  them. I also requested information from a 7th
  (MIT) and clarifications from most of the
  original 6
• I then rated the responses according to the
  selection criteria
• University of Florida
• Iowa State/University of Iowa
• University of Maryland
• Massachusettes Institute of Technology
• University of Minnesota
• Northeastern University
• University of Wisconsin
• Recommended University of Florida for next pT2
  center

21
Tier 2 Hardware Status
From CalTech 20 dual 800MHz PIIIs, 0.5 GB
RAM Dual 1GHz CPU DS, 2GB RAM 2 X 0.5 TB fast
(Winchester) RAID 70 MB/sec sequential disk
access Total cost 140k at each site. Installed
CMS software, in process of commissioning. Tested
ooHits. UCSD has similar system in similar
status Plans to buy another 20 duals this year
at each site, 65k each(?) This makes a total
spending of 410k out of projected costs of 430k
22
Florida Cluster

• A CMS Computing Cluster is being assembled at the
  University of Florida by Paul Averys Group.
• 72 computational nodes
• Dual 1GHz PIII
• Based on Intel STL2 Motherboards
• Built in Ethernet, SCSI, and VGA
• Serverworks ServerSet III LE Chipset
• 512MB PC133 MHz SDRAM
• 76GB IBM IDE Disks
• Purchased from Ace Computers of Chicago
• (http//www.acecomputers.com)
• Per system cost of 1950
• Sun Dual Fiber Channel RAID Array 660GB (raw)
• Connected to Sun Data Server
• Not yet delivered. Performance numbers to follow.

23
CMS Milestones (as pertinent to UF)

• CMS is starting from a working OO system and
  slowly building in complexity. Our goals for the
  next few years are to fully OO the software chain
  and implement grid tools.
• 6/01/2000 Roughly 1 M events produced for Higher
  Level Trigger Studies (0.1 MDC, though not
  full complexity). Full OO reconstruction
  pileup.
• 12/01/2000 5M events produced for HLT Studies,
  representing 0.5 MDC. CERN, INFN, FNAL,
  Caltech and Moscow are participating in
  deploying grid tools to share simulated data.
  
• 12/01/2000 Trigger TDR.
• 01/30/2001 create AOD at FNAL for JPG group. For
  this we have to run over 2M final jets
  events. Support re- reconstruction. Needs
  5-6k Si95s for this.

24
CMS Milestones (as pertinent to UF)

• 06/01/2001 1 data challenge. This means 10M
  events, or gt2M generated in US. 2M at FNAL, 2M
  at T2? Continue with grid implementation.
  Non-networked based data movement?
• 12/01/2001 1-2 data challenge. Fully support
  prototpye Tier 2s. Implement any
  reconfiguration or design from previous data
  challenge.
• 11/01/2002 DAQ TDR
• 12/01/2003 Software and Computing TDR
• 03/01/2003 Tier 1 Regional Center 5 prototype.
• 06/01/2003 5 Data Challenge. Goals not yet
  defined.

25
CMS Milestones (as pertinent to UF)

• Longer Term Milestones
• 10/01/2003 Begin Tier 1 implementation phase
  (FY2004)
• 12/01/2004 Physics TDR
• 12/01/2004 20 Data Challenge
• 04/01/2006 Pilot LHC run (2-4 weeks at 1032)
• 08/08/2006 LHC running (1033)
• -02/27/07

26
Near Term Plans

• Define hardware strategy for FY2001. DONE
• We have the project plan as the basis.
• Refining using input from
• US-CMS physics coordinator and groups
• US-CMS users
• ASCB
• CMS SCB
• Production Organization DONE
• Scripts developed at Fermilab have been adopted
  as the official CMS distributed production
  scripts
• Working with T2 on production and prototype grid
  systems
• Strategy for dealing with strong authentication
• In progress
• Organize another CMS software tutorial early
  summer,
• coinciding with kerberizing CMS machines

27
Summary

• Lots of progress in UF
• US-CMS computing beginning to coalesce
• Good interaction with international CMS
• Production status
• US CMS taking a lead role here.
• Would like stronger connections with physics
  groups
• By next PMG
• Hardware reqs out and equipment coming in
• RD projects underway
• Continue documentation and collaborative efforts
• Production at 1st T2 site ramped up
• Florida prototype ready to begin production
• Critical items
• Fix milestones to new schedule and update WBS
• Signed SoWs for T2 sites