Simulating the scheduling of parallel supercomputer applications
Conference
·
OSTI ID:5342872
An Event Driven Simulator for Evaluating Multiprocessing Scheduling (EDSEMS) disciplines is presented. The simulator is made up of three components: machine model; parallel workload characterization ; and scheduling disciplines for mapping parallel applications (many processes cooperating on the same computation) onto processors. A detailed description of how the simulator is constructed, how to use it and how to interpret the output is also given. Initial results are presented from the simulation of parallel supercomputer workloads using Dog-Eat-Dog,'' Family'' and Gang'' scheduling disciplines. These results indicate that Gang scheduling is far better at giving the number of processors that a job requests than Dog-Eat-Dog or Family scheduling. In addition, the system throughput and turnaround time are not adversely affected by this strategy. 10 refs., 8 figs., 1 tab.
- Research Organization:
- Lawrence Livermore National Lab., CA (USA)
- Sponsoring Organization:
- DOE/DP
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 5342872
- Report Number(s):
- UCRL-102059; CONF-8911132--1; ON: DE90003604
- Country of Publication:
- United States
- Language:
- English
Similar Records
Adaptive Parallel Job Scheduling with Flexible CoScheduling
Nonpreemptive run-time scheduling issues on a multitasked, multiprogrammed multiprocessor with dependencies, bidimensional tasks, folding, and dynamic graphs
Performance analysis of job scheduling policies in parallel supercomputing environments
Journal Article
·
Mon Oct 31 23:00:00 EST 2005
· IEEE Transactions on Parallel and Distributed Systems, 16(11):1066-1077
·
OSTI ID:918866
Nonpreemptive run-time scheduling issues on a multitasked, multiprogrammed multiprocessor with dependencies, bidimensional tasks, folding, and dynamic graphs
Thesis/Dissertation
·
Wed Dec 31 23:00:00 EST 1986
·
OSTI ID:7106903
Performance analysis of job scheduling policies in parallel supercomputing environments
Conference
·
Thu Dec 30 23:00:00 EST 1993
·
OSTI ID:46276