No. 132 (00309)

Family name : van Lingen
Given name : Frank 
Affiliation : California Institute of Technology 
Abbreviation : Caltech 
E-mail address : Frank.van.Lingen@cern.ch

Title :  Failure and dependencies of processes within models for the CMS
         DC04 using the MONARC simulation framework and monitoring  data 

Authors : Frank van Lingen, Kathrin Paschen, Iosif Legrand, Ciprian
          Dobre, Corina Stratan 

Abstract : 
Within the next months the CMS experiment at CERN will start its data
challenge 2004 (DC04). DC04 involves the production, analysis and
distribution of large amounts of event data. Several computing centers
will act as tier 1 or tier 2 with CERN as tier 0. The data flow within
DC04 involves several cycles and dependencies. This paper discusses
several models that simulate the data flow of DC04.

In order to get realistic models,  DC04 models are based on monitoring
data that are extracted from running mulitple analysis jobs.

From this monitoring data duration distributions and other parameters
are generated, that are used to get a realistic estimate for parameters
in the DC04 models.

A focus of the model, are dependencies between the different processes
within DC04 and the impact of failure or delay of some of these process
for the whole data flow. The aim of the models is to provide insight in
what would be a suitable configuration and set up for DC04.