Minutes of the Computing meeting held on 20 May 1999

Present: P.Mato,  M.Cattaneo, M.Franck, P.Maley, P.Colrain, G.Corti, E.Van Herwijnen, J.Harvey, R.Forty, 
              P.Binko, R.Chytracek, M.Needham, R.Van Der Eijk, G.Gracia, J.Closier, T.Cass, D.Steele, F.Ranjard

Apologies:  F.Harris, I.Videau, V.Breton

Minutes : Florence Ranjard

      1.  Computing meeting day (everybody)

• It seems that Thursday is NOT the right day for our Computing meeting because it clashes with Trigger meeting
  and it is not close enough to the Physics meeting or Calorimeter meeting (Tuesday afternoon) for outside people.
• ==> Computing meetings will be held on Wenesday  morning 9.00-11.00

2. a mini NT farm (E. Van Herwijnen)

Eric has set a NT mini farm to run simulation. He asks for some guinea-pigs who accepts to put their PCs into the farm.
A 10Gbytes disk is needed. It is used at night with LSF.

the charge for LSF licence is not clear yet.

Users are used to reboot at any time, what will happen to running jobs? We have to learn how to use such a farm.

3. a LHCb Linux cluster (T.Cass)

Tony presented the proposal for an LHCb batch facility at CERN based on 4 dual processor PCs running Linux.
The proposal will be presented to COCOTIME on Friday 21 May.

• Why?
  • rsplus does not fulfill our needs,
  • we will have to move to a new compiler,
  • ROOT does not work properly on AIX,
  • AIX is less and less supported.
• What?
  • the peak load on rsplus represents 4 batch PCs.
  • with 4 dual processor PCs we double the required batch capacity.
  • if it runs Linux the LHCb cluster would be equivalent to other clusters such as NA49 and Nomad.
  • we need good I/O capability to match analysis work requirements..
• How to access?
  • through Lxtest a Linux interactive cluster such as rsplus/rsbatch.
• When?
  • end of June, early July.
  • Lxtest should come earlier.
• Who?
  • Catherine Charbonnier will be in charge of the cluster.
• Questions / Remarks
  • it is a positive move, but are the resources similar to rsplus/rsbatch?
    • we have requested about 1/16 rsbatch , we will get about 1/4 rsbatch.
  • turnround on rsplus is quite good in the evening, would it be as good on a 4 dual processors?
    • it should, or IT would have to improve it.
  • which security?
    • same as any other cluster.
    • no web server.
  • CERN Linux installation does not have rfio, need sue features to install it

          4. Detector Description : News (R.Chytracek)

             Rado presented a status report of the Detector Description project (ppt)

• detector description transient mode
  • logical detector description
    • description based on a detector logical organization,
    • used for navigation
    • set a unique identifier for individual detector element.
  • geometry detector description
    • defines structure, hierarchy of volumes, shapes, materials, dimensions,ect..
• Prototype
  • test prototype is done
  • release prototype being prepared.

          5. A OO Reconstruction program: How to start? (M.Cattaneo)

             Marco presented its first ideas: (ppt)

• discussion during next Software week.
• we should define the scope of the reconstruction program.
• short term goals:
  • build a team of physicists,
  • define reconstruction strategy,
  • identify and define external inputs.
• medium term goals:
  • build a reconstruction skeleton within Gaudi independent of SICB
  • wrap SICB reconstruction algorithms,
  • integrate OO code.

 

======

next meeting:  Wenesday 26 May 1999 9:00-11:00  blg 32 /sc-22
                       P.Binko: Histogram Template Library (HTL)