Analysis%20farm:%20past%20experience

1
Analysis farmpast experience

• Concezio Bozzi
• Padova, Oct 16th 2003

2
Farm usage

• Could have been better in terms of CPU cycles
• Low usage led to the decision to devote (part of)
  the farm for SP(45) starting in May2002
• SP5 was actually deployed outside SLAC at the
  INFN TierB and Caltech at the end of 2002
• This did not prevent bursts of CPU-intensive
  toy-MC studies (CPT violation, F.Martinez et al),
  and of course the discovery of new particles (A.
  Palano)

3
Main limitations

• Disk space
• We had only Run1 data (release 8) and SP3 MC
  until spring 2002 (Kanga production at SLAC was
  also a showstopper)
• Situation was better in summer 2002, when mostly
  all Run12 data (release 10) and half of SP4 MC
  were available
• Batch CPUs
• 13 biprocessors originally, then 29
• Never got all available 58 CPUs running at the
  same time
• No luck in HW procurement (lenghty, many
  failures, etc)
• Not enough manpower for hardware and software
  upgrades/maintenance