CERN/COCO/2003-17
  COCO/181
  28 October 2003
 
 
 

D R A F T



Computer Time Allocation Group

Minutes of the hundred eighty first meeting held on 28 October 2003

Closed session: 28 October 2003, 10:00 h

Present:  L. Betev, H.F. Hoffmann (Chairman), A. Martin, A. Masoni, A. Morsch (Secretary), S. O'Neale, G. Poulard, W. von  Rüden, J. Salicio, T. Smith, T. Wildish
Invited:  B. Panzer-Steindel, J. Shiers
Apologies:  J. Closier



1. Minutes of the 180th meeting held on 6 June 2003: Approval of minutes, action list, matters arising

Cocotime took note that the minutes ( url ) are approved.

Actions:

A replacement for M. Stavrianakou has been found thanks to the help of J. Salicio. The chairman welcomed Gilbert Poulard as a new experiment representative.

The web form for the 2004 requests has been prepared by Tim Smith. All experiments have sent their requests.

2003 Acquisition Status

(See slides presented by Tim Smith as (pdf))

Interactive services

As agreed by Cocotime 70x1GHz PCs with 1 GB memory running RH7 are in production.

The small Solaris cluster SUNDEV has been upgraded to 10x1GHz Sun210s.

Batch

No major upgrades since June 2003. Lxbatch is running with ~700k Si2k.

EIDE Disk servers

100% of the servers for analysis and Solaris replacement and 70% of the required production capacity have been installed.

AFS

Installed capacity in accordance with request, but on different hardware due to problems with Solaris servers.

Network

One item (structured cabling LHCb Bldg. 156) pending due to other priorities of network group.

W. von  Rüden asked how many people use Sundev. There are 45-50 users per day, 41 are from CMS.

Computer Centre Status Report

(See slides presented by Tim Smith as (pdf) )

AFS

Number of AFS users is plateauing at around 14000. After some plateauing in the first half of 2003 due to a aggressive deletion policy user space rises again with the usual slope. The ratio between assigned and used quota stays constant at a quite high value of about 1.8 as remarked by W. von  Rüden.

CASTOR

Approximately 11 M files have been stored now in the CASTOR system corresponding to 1.7 PB of data. The decrease in the total data volume with respect to June 2003 is due to file deletion after the ALICE data challenge.

Tapes

Compass has the highest number of tape mounts followed by LHCb, NA48, Aleph and Delphi; LHCb due to their data challenge whereas NA48 and Delphi simply lack disk space as has been remarked by L. Betev.

LXBATCH

Whereas there was still some free capacity during summer Lxbatch is now running at 100% of its capacity.

Case studies comparing the pattern of CPU and clock time for different experiments have been shown. Particularly Compass uses only a relatively small fraction of the allocated time. W. von  Rüden remarked that the LHC experiments perform systematically better than the other. S. O'Neale expressed his opinion that the main problem is waiting for tape mounts. T. Smith explained that several jobs are put on the same CPU to reduce idle time. However, more appropriate would be a classification of batch job stages for correct monitoring. W. von  Rüden suggested to develop a batch template to be distributed to the users to make this monitoring possible. T. Wildish cautioned that one has to consider the efficiency for all jobs of a production run to determine the total efficiency (waiting for the last job finishing sometimes limits the overall efficiency).

2004 Allocations and Referee Reports

B. Panzer-Steindel showed an overview of the 2004 requests endorsed by the referees. He pointed out that the preliminary sharing of resources is 1.2 MCHF for LHC and 0.4 MCHF for non-LHC experiments. W. von  Rüden explained that this year IT will have to pay 150 kCHF for tapes assuming that the total costs will be shared between IT and EP divisions.

W. von  Rüden observed that the LEP request is on the same level as last year. The main requests from the non-LHC experiments come from Compass, NA45 and NA48. L. Betev suggested not to put Compass and the other non-LHC experiments under the same hat in order to avoid that all have to suffer reduction. W. von  Rüden recalled that there is a MoU which states that Compass will pay a significant fraction of its needed computing capacity. He asked which fraction of the capacities are taken over by institutes outside CERN. A. Martin answered that about two third is done outside CERN on small farms.

H.F. Hoffmann proposed a meeting between the Compass Spokesperson, the Director for Fixed Target Programmes and him to clarify the issue of the Compass MoU (action item).

Action list

Responsible Minuted Action
H.F. Hoffmann 2003/10/18 Organize meeting between Compass and CERN Management to clarify issue of Compass contribution to computing resources as agreed in the MoU.
H.F. Hoffmann 2003/06/06 Send Memorandum to Compass stating approval of increase of 2003 request.

Andreas Morsch