Benchmark Data

For our performance evaluation of the Linux TSS system, a subset of HP disk traces [10] have been used. It contains disk accesses generated by HP-UX file systems on a system not running scientific applications.

Further, in order to reduce the time taken to execute a trace through the system, Trace Compaction (as in [3]) has been done. Trace compaction reduces the time between the accesses by a factor of 1000. It was found in [3] that the trace compaction has little impact on the results but speeds up the performance evaluation process. Also, a part of the actual trace containing about 70,000 accesses was used for the trace simulation.