CS Document 523-v2
Testing the CDF Distributed Computing Framework
- Public document
- Same as Viewable by
- A major source of CPU power for CDF (Collider Detector at Fermilab) is the CAF (Central Analysis Farm at Fermilab. The CAF is a farm of computers running Linux with access to the CDF data handling system and databases to allow CDF collaborators to run batch analysis jobs. Beside providing CPU power it has a good monitoring tool. The CAF software is a wrapper around a batch system, either fbsng or condor, to submit jobs in a uniform way. So the submission to the CAF clusters inside and outside Fermilab from many computers with kerberos authentification is possible. It is mainly used to access datasets which comprise a large amount of files and analyze the data. Up to now the DCache system is used to access the files. In autumn 2004 some of the important datasets will only be readable with the help of the data handling system SAM (Sequential Access to data via Metadata). This will be done in order to switch to use only one data handling system at Fermilab and on the remote sites. SAM has been used in run II to store, manage, deliver and track the processing of all data. It is designed to copy data to remote sites with remote analysis in mind. To prove CAF and SAM could provide the required CPU power and Data Handling, stress tests of the combined system were carried out.
A second goal of CDF is to distribute computing. In 2005 50% of the computing shall be located outside of Fermilab. For this purpose CDF will use the DCAF (Decentralized CDF Analysis Farms) in combination with SAM. To achieve user friendliness the SAM station environment has to be common to all stations and adaptations to the environment have to be made.
- Files in Document:
- V_Bartsch.pdf (21.4 kB)
- Notes and Changes:
- Fermilab Publication number CONF-04-492-CD
- Associated with Events:
- CHEP2004 held from 27 Sep 2004 to 01 Oct 2004 in Interlaken, Switzerland