| Computing Meeting, Wednesday, January 16, 2008 [message #5622] |
Mon, 14 January 2008 10:05  |
Silvia Masciocchi
Messages: 162
Registered: May 2006
|
first-grade participant |
From: *gsi.de
|
|
Meeting at 12:30, room 4.161
Minutes by Dariusz Miskowiec
Preliminary Agenda:
===================
1. New results from speed tests - Misha
2. Status of transfer of December data
3. Third-party-copy (Fabrizio Furano)
4. Data storage and management at GSI
Care of data stored under xrootd
? priviledges for individual users implemented?
? news about the fuse interface?
? when usable space in lustre?
5. lxb255 and other failures of the batch machines
6. Second GSI contact person for GRID matters
7. GRID analysis jobs (Ana, Kilian)
[Updated on: Tue, 15 January 2008 16:41] Report message to a moderator |
|
|
|
|
|
| minutes [message #5645 is a reply to message #5622] |
Wed, 16 January 2008 16:44  |
Dariusz Miskowiec
Messages: 70
Registered: August 2004
|
continuous participant |
From: *gsi.de
|
|
1. New results from speed tests (Misha)
hardware reading limit: 70 MB per disk
reading Root files (Pythia): 20-30 MB/s per job
reading from RAID: 64 MB/s aggregate
reading from tuned RAID: 183 MB/s aggregate
Peter M.: the tests are not yet conclusive.
Reading ESD files and other tests will follow.
2. Status of transfer of December data (Kilian): 50% of the December raw data were transfered to Karlsruhe (maximum speed 45 MB/s), the other part went to IN2P3. No copy to GSI was possible. Reason: AliEn and xrootd services have been separated. Xrootd redirector for the fileserver cluster runs now on grid2. The cluster is accessible with xrootd methods, but so far not via AliEn. Reason: probably central configuration. Kilian's call for help before Christmas remained without answer from CERN. Marian: Karlsruhe is not better than CERN, the data need to be at GSI because initially many iterations are needed.
3. New xroot scheme by Fabrizio Furano (Silvia): superredirector at CERN, talking to redirectors at centers, should make all xrootd systems look like one single xrootd. Kilian will provide dedicated hardware at GSI for the corresponding tests.
4. Data storage and management at GSI (Silvia): xrootd is not really suitable for storing user data because normal users have limited access to their files (e.g. cannot overwrite) and have no means to monitor the location and sanity of files e.g. in case of hardware failure. Fabrizio says xrootd is not thought for this. Peter M.: for disk data storage at GSI nfs is supposed to be replaced by lustre. Kilian will talk to Walter Schoen and see if a usable lustre system can be set up within a reasonable time. Kilian will report on this next week.
5. AOB
Silvia asks Kilian to send computing relevant messages (like informations about hardware problems) to alice-gsi@gsi.de .
CERN people need another contact person at GSI for the cases when Kilian is not available.
Kilian will contact Ana so they can meet and figure out the origin of Ana's problems with running AliAnalysisTask jobs on the farm.
Jacek will test running AliAnalysisTask jobs on the GSI proof cluster.
[Updated on: Thu, 17 January 2008 10:34] Report message to a moderator |
|
|
|
GSI Forum
Search
Help
Members
Login
Home
