System performance modelling WG meeting

→ Europe/Zurich
600/R-001 (CERN)

600/R-001

CERN

15
Show room on map
  • Local: Andrea Sciabà, Markus Schulz, Dirk Duellmann
  • Remote: Michele Michelotto, Johannes Elmsheuser, Gareth Roy

Andrea presented an outline of the CHEP talk on the cost model WG. No particular comments received.

Markus presented his ideas on areas where efficiency improvements could be gained, trying to quantify them. In some cases the gains are theoretical maxima that cannot possibly be achieved, but it still gives an idea of the existing gap. For example, a simple estimate produces a saving in manpower for operating the storage in WLCG of up to 40%.

Markus mentioned a study by Xavier Espinal on the reliability of storage at the Tier-0. Related to this is the idea is to keep only one copy of the data on disk and rely more on tape as a backup. The impact could be estimated by looking at popularity data or even extrapolating from the incident at CNAF.

Johannes mentioned that ATLAS is working on improving fast simulation, namely FastCaloSim and FastChain. This will considerably improve simulation time, but as the "expense" of many more MC events being produced and hence more storage utilisation.

During the round table, Dirk reported that they are introducing on-the-fly popularity metrics on all EOS instances.

Markus reported that there is at CERN a student from Dubna who worked on a dynamical model of a computer centre for NICA. Her work might very well be useful for us. In addition, there is a summer student who will work on xrootd caching together with David Smith. Gareth reported that also at Edinburgh they are working on xrootd caches (see slides).

As an action for those who could not attend the meeting, please read (and possibly comment, on the mailing list) Markus' slides on possible gains.

 

 

There are minutes attached to this event. Show them.
    • 16:30 → 16:50
      Ideas on potential gains 20m
      Speaker: Markus Schulz (CERN)
    • 16:50 → 17:05
      CHEP talk (and paper) 15m
      Speaker: Andrea Sciaba (CERN)
    • 17:05 → 17:25
      Round table 20m
    • 17:25 → 17:30
      AOB 5m