Speaker
Description
URL for further information
http://twiki.cern.ch/twiki/bin/view/EGEE/SA3
Impact
Understanding the performance of CEs is very important for operators of production grid infrastructures. A common method for benchmarking enables a fair comparison to be made between different implementations. Establishing the performances limitations of such interfaces is vital in order to initially decide on the adoption of a specific implementation and to manage the growth of the infrastructure.
The need for a standard interface with respect to CEs has already been established and the BES interface from the OGF is emerging as a potential solution. This work endeavors to understand from a performance aspect if an implementation of BES is suitable for a production infrastructure.
Conclusions and Future Work
The conclusion of this work is the presentation of the results from the performances tests. Future work includes the addressing the limitations found.
Detailed analysis
This paper starts with a short overview on Grid interfaces for batch system more commonly known as " Computing Elements" (CE). The important metrics for understanding the performance of these interfaces are discussed and a benchmarking method for obtaining these results is presented. The evaluation criteria for CEs used in the LCG project are given as baseline for performance requirements in general. This benchmarking method is used to evaluate three different CE implementations; the LCG-CE, the CREAM CE and the BES interface from Platform Computing. The LCG-CE is the interface that has been successfully used on the LCG production infrastructure over the past four years, the CREAM CE is the new interface provider by gLite which is starting to have production exposure and the BES interface from platform is an implementation of the OGF standard.
Keywords
Grid, CE, Computing Interface, BES, Performance, OGF
