Message to Conveners

To: kpitts@fnal.gov, savard@fnal.gov, willis@fnal.gov,
    chlebana@fnal.gov, reb@fnal.gov, rlc@fnal.gov, rolli@fnal.gov,
    mschmidt@fnal.gov, bed@fnal.gov, goshaw@fnal.gov

Dear physics group conveners,

 as you probably know Rob and Kevin put together a CAF computing review
 (CAF = central analysis facility).

 The review committee would like to hear from the physics groups about
 your expected analysis needs, as best you know it.

 To learn more about this, please take a look at:

 http://www-cdf.fnal.gov/upgrades/computing/projects/central/

 Here's how you can help. We need feedback before September 5th
 if you want to influence the recommendations regarding charge 1, the
 purchase of another SUN SMP for $400,000 .

 (1) How many people in your group will be using the central analysis
     facility (CAF) for summer conferences 2002? To 0th order this includes
     anybody analyzing any data next May/June/July/August .

     Please feel free to give a rough estimate of people doing
     most of their work offsite, versus those who will have to rely
     heavily on datasets that are served via CAF.

 (2) How many events will an average person in your group
     spin through on the central facilities? What's the largest
     number of events a person in your group is likely to
     access via the central facilities? How many people will
     need to access those largest datasets? Do you expect them to
     do this more than once/twice ? I.e. at what level do you expect
     people to copy them to their desk tops/offsite?

     Please assume an optimistic 300pb^-1 of data for next summer.

 (3) Do you have (epect to have) a "standard" PAD skim procedure to 
     feed secondary data sets to the group? 

     If yes:
     (3.1) What type of format do you (expect to) use? ntuple?
           PAD-like requiring AC++ to read it?
     (3.2) What is the expected data reduction factor?
     (3.3) Do you expect to co-ordinate across physics groups?
     (3.4) Do you intend to have these secondary datasets served
           by the CAF? Or are most people going to copy them offsite?
     (3.5) What fraction of (1) and (2) above is going to be satisfied
           by running on the secondary sets only?

     For your info,
     we are assuming that PADs will be around 50-100kB per event and
     some form of standard ntuples will come to 5-10kB per event.
     In addition, PADs require AC++ and allow redoing reconstruction
     while ntuples don't. If that sounds unreasonable, please let us
     know why.

 (4) Is there anything else you think the review committee should be
     aware of?


It's ok if you don't have all the answers. Any input whatsoever would
be appreciated. 

Furthermore, it would help if you could formulate a group response
rather than every person in the collaboration bombarding us with
their opinions.

Many thanks for your help.
CAF computing review committee
http://www-cdf.fnal.gov/upgrades/computing/projects/central/reviewers.html