CC-J Monthly Report

1
CC-J Monthly Report

• Shinya Sawada (KEK)
• for
• CC-J Working Group
• http//ccjsun.riken.go.jp/ccj/

2
Contents

• VRDC event generation at CC-J
• Hardware upgrade plan in near future
• Very preliminary plan towards CC-J operation

3
VRDC event generation _at_ CC-J

• http//ccjsun.riken.go.jp/hayashi/vrdc.html
• Au-Au
• Central arm, central bias 40k events
• Central arm, minimum bias 80k events
• Hijing-PISA99-PHOOLresp
• PISA files and PRDFs are being sent to HPSS_at_RCF
  via WAN. 100 files for central arm central
  bias already exist at RCF.
• proton-proton
• (Pythia-)PISA99-PHOOLresp
• Central arm (open heavy flavor) 40k events
• Muon arm (DY, min. bias) 318k events
• Trigger study
• Central arm (J/y f) 114k events

4
Throughput (CPUT)

• Numbers are reduced to 450MHz pentium CPU
• Hijing 30min./600ev (central bias)
• PISA99
• Au-Au central arm, minimum bias 200sec/event
• Au-Au central arm, central bias 500sec/event
• Proton-proton and trigger runs vary case by case
• Response chain
• Au-Au central arm, minimum bias 15sec/event
• Au-Au central arm, central bias 40sec/event
• Proton-proton and trigger runs vary case by case

5
Throughput (Data transfer)

• Data amount for Au-Au
• PISAout 300MB/200ev (min. bias) 800MB/200ev
  (cent. bias)
• PRDF 1GB/200ev
• rcp from CC-J to BNL (spin.riken.bnl.gov)
• 350kB/sec/1-rcp
• 700kB/sec/2-rcp
• 950kB/sec/3-rcp
• total throughput would be being saturated around
  here mainly due to the network bandwidth around
  RIKEN. APAN has much more band width.

6
Problems

• Successful event ratio
• PISA99
• BIMPCT Reported at PHENIX computing meeting on
  Janurary 2000 by Y.Watanabe. C.Maguire commented
  that it was due to FLUKA and his simulation group
  was going to patch this problem by skipping bad
  events.
• ZEBRA error A part of them could be reproduced
  again gt software problem? Another part of them
  could not be reproduced gt hardware (memory etc.)
  problem?
• PHOOL response chain
• segmentation violation Need more study.

7
Notes

• File name convention
• Tarun, Indrani, Tim and others have agreed with
  the rule, it might not be a eternal version.
  Runs at CC-J have 1E8 offsets.
• Run database
• Even now, many runs for various kinds.
• Simulation work will continue for the order of
  decade.
• Should be included in the to-do list.
• Script
• Stable and easy-to-understand job scripts (or
  their templates) should be written, especially
  for various people to use the farm.
• script builder in the to-do list.

8
Notes continued

• Read/write via NFS
• At RCF, Charlies team seems to read/write all
  the files via NFS, while we rcp (or pftp) files
  to/from the local disk before/after the
  simulation. CC-J experience shows NFS-write
  would be a bottle neck for the simulation with
  high I/O rates. (see http//ccjsun.riken.go.jp/cc
  j/doc/NFSbench99Apr.html)
• mu-camera for the muon arms
• might have to be developed soon.
• Memory usage for the response chain
• was about 300MB in some cases, while we have
  256MB real memory per CPU.

9
Hardware upgrade plan / CC-J schedule

• More 32 CPUs (700MHz?) will be purchased by the
  end of March.
• Total number of CPUs will be 96.
• How about the status of RH6.1 tests?
• More 1.6TB RAID disk will be purchased by the end
  of March.
• Total work disk will be 3.2TB.
• Acceptance tests of HPSS and other components
  will be in February and March.
• Test operation for PHENIX from around mid-March?
• Routine operation from May?

10
Very preliminary idea on CC-J operation

• See the Proposal for the PHENIX CC-J.
• http//ccjsun.riken.go.jp/ccj/doc/plan/
• Discussions are going among related people.
• CC-J Planing and Coordination Office will be
  established.
• Simulation should be proposed through PWGs?
• Each PWG should define responsible persons for
  the simulation?
• A simulation coordinator at the PHENIX side??
• A detailed announcement will be at the next core
  week.

11
To do

• Tests for replication of Objectivity/DB
• script builder?
• At least good templates are necessary. If one
  has a good template, he/she could easily develop
  a GUI/web based script builder.
• Run number database??
• So many simulation, so many runs of simulated
  events, various conditions for runs
• Documentation
• Explain CC-J specific configuration, rules and so
  on, published on the web and/or with pdf.