The data used in this book comes from the following sources.

NASA Ames iPSC/860

Data about 42,264 jobs submitted in the fourth quarter of 1993, collected by Bill Nitzberg. More than half were invocations of pwd by system personnel to verify that the system was responsive. The system is a 128-node hypercube manufactured by Intel.

Available from the Parallel Workloads Archive: http://www.cs.huji.ac.il/labs/parallel/workload/l-nasa-ipsc/.

LANL CM-5

Data about 201,387 jobs submitted from October 1994 to September 1996, collected by Curt Canada and including three large flurries. The system is a 1,024-node Connection Machine 5 from Thinking Machines Corp., the biggest of its kind at the time, which reached second rank in the Top500 list of 1993.

Available from the Parallel Workloads Archive: http://www.cs.huji.ac.il/labs/parallel/workload/l-lanl-cm5/.

SDSC Paragon

Data about 115,591 jobs submitted from January 1995 to December 1996, collected by Reagan Moore and Allen Downey and including several flurries. The system is a 416-node Intel Paragon machine. Originally this data was provided as two logs for separate years. Due to anonymization, user IDs may be inconsistent in the two years. Therefore we only use the 1995 data when user data is important.

Available from the Parallel Workloads Archive: http://www.cs.huji.ac.il/labs/parallel/workload/l-sdsc-par/.

CTC SP2

Data about 79,302 jobs submitted from June 1996 to May 1997, collected by Dan Dwyer and Steve Hotovy, with one small flurry. The system is a 512-node IBM SP2 machine, the biggest of its kind at the time, and ranked 6 in the 1995 Top500 list.

Available from the Parallel Workloads Archive: http://www.cs.huji.ac.il/labs/parallel/workload/l-ctc-sp2/.

KTH SP2

Data about 28,490 jobs submitted from September 1996 to August 1997, collected by Lars Malinowsky. The system is a 100-node IBM SP2.

Available from the Parallel Workloads Archive: http://www.cs.huji.ac.il/labs/parallel/workload/l-kth sp2/.

SDSC SP2

Data about 73,496 jobs submitted from April 1998 to April 2000, collected by Victor Hazlewood, and including one large job flurry and one large process flurry. The system is a 128-node IBM SP2.

Available from the Parallel Workloads Archive: http://www.cs.huji.ac.il/labs/parallel/workload/l-sdsc-sp2/.