This extensive log contains information on the user, account, and application, requested and used nodes and time, CPU time, submit, wait and run times. Note that in the first third of the log the utilization is about 10% lower than in the later two thirds. this could indicate that the system's configuration was different during this period. The original log, with all the available information, was available directly from the NPACI JOBLOG repository, in a special format described on that site. The version in the Standard Workload Format loses some data as specified below. The workload log from the SDSC SP2 was graciously provided by Victor Hazlewood, who also helped with background information and interpretation. If you use this log in your work, please use a similar acknowledgment.
Downloads:
|
|
The differences between conversion 4 (reflected in SDSC-SP2-1998-4.swf) and conversion 3 (SDSC-SP2-1998-3.swf) are
Note that the distinction between cancelation and failure is not clear. Jobs with a code of 1 (completed successfully) may include jobs that failed to perform their function but exited cleanly.
A flurry is a burst of very high activity by a single user. The filters used to remove the four flurries that were identified are
user=21 and job>13716 and job<16208 (944 jobs)Removing the first 10 jobs was added in the second cleaned version, as they seem to represent activity from long before the actual logging started. Note that the filters were applied to the original log, and unfiltered jobs remain untouched. As a result, in the filtered logs job numbering is not consecutive.
user=374 and job>14968 and job<28553 (11740 jobs)
user=197 and job>31766 and job<33203 (635 jobs)
user=328 and job>66552 and job<68107 (452 jobs)
Further information on flurries and the justification for removing them can be found in:
File SDSC-SP2-1998-4.2-cln.swf (cleaned)
