25th october 2006tim adye1 ral tier a tim adye rutherford appleton laboratory babar uk physics...
TRANSCRIPT
25th October 2006 Tim Adye 1
RAL Tier ARAL Tier ATim Adye
Rutherford Appleton Laboratory
BaBar UK Physics MeetingQueen Mary, University of London
25th October 2006
25th October 2006 Tim Adye 2
Outline
• CPU Usage
• CPU Allocations
• Disk Status
• The bleak future
• Summary
25th October 2006 Tim Adye 3
BaBar Batch CPU Use at RAL
0
50,000
100,000
150,000
200,000
250,000
300,000
350,000
400,000
Week Beginning
BaB
ar C
PU
Ho
urs
per
Wee
k(N
orm
alis
ed to
P45
0)
GridClassic MC ProductionUK UsersNon-UK Users
25th October 2006 Tim Adye 4
BaBar Batch Users at RAL(running at least one non-trivial job each week)
0
5
10
15
20
25
30
35
40
45
50
Week Beginning
BaB
ar U
sers
per
Wee
k
Grid UsersUK UsersNon-UK Users
A total of 237 new BaBar users registered since December 2001
25th October 2006 Tim Adye 5
CPU Allocations
CPU Allocation(MAUI fairshare target)
CPU Usage(MAUI fairshare usage)
BaBar CPU Allocation and Usage
Farm Capacity
25th October 2006 Tim Adye 7
Requests and Allocations
BaBar Jan06Request (MoU)
BaBar Mar06 Request(after Tau/QED -> SLAC)
GridPPAllocation
End of Disk(TB)
CPU(kSI2k)
Disk(TB)
CPU(kSI2k)
Tape
(TB)
Tape bandwidth(MB/s)
Disk(TB)
CPU(kSI2k)
05 Q4 100 435 100 435
06 Q1 140 495 95 435 180 14 95 435
06 Q2 155 555 120 500 200 16 95 435
06 Q3 170 625 120 550 220 17 95 435
06 Q4 190 660 135 600 250 19 95 200
25th October 2006 Tim Adye 8
Data and Storage
• Keeping up-to-date with new production• Uses disk space freed up by:-
• Tau/QED skims removed in February• Converted R18b pointer skims to R18c deep-copy• Removed AllEvents in August
• All old files still accessible from tape• Except old SP5/SP6 generics, now deleted from tape
• Currently problems with user data disk• /stage/babar-user1 offline since 13 Oct• Recovering the data going slowly – hope to be done by the
end of the week• Three 1.9 TB AWG disks
• /stage/babar-awg1/Quasi2body (was TauQED)• /stage/babar-awg2/Quasi2body• /stage/babar-awg3/ThreeBody
25th October 2006 Tim Adye 9
Three bullets we try to dodge
1. BaBar disk, tape, and CPU requirements increase with luminosity.No change in GridPP allocations Jan06 to Dec08.
2. PPGP cuts removed BaBar/RAL support staff• 1.5 FTE -> 0.25 FTE in April 2007• No effort to import data, releases, help users, etc
• identified 21 BaBar-specific tasks needed to keep Tier A running
3. GridPP proposal to remove non-Grid access by September 2007• No RAL front-ends, no NFS access to user/AWG disks• Continued BaBar user analysis probably impossible
• SP and/or skimming might still be possible via the Grid
25th October 2006 Tim Adye 10
Summary
• We are making good use of the resources we have• Apart from an (understandable) lull over the
summer
• The service works well most of the time• Current disk problem is severe, but rare
• We are fighting hard for1. the resources we need2. the staff we need3. the non-Grid access we need