Job scheduling on a hypercube
Thesis/Dissertation
·
OSTI ID:6087363
The author studies the scheduling of independent jobs on hypercube multiprocessors. He assumes that the hypercube system supports space-sharing for multiprogramming, i.e., a hypercube is partitioned into subcubes and each job is assigned to a dedicated subcube and many jobs can be running simultaneously without interfering with each other. Then the problem of how to schedule a set of jobs so that they can be finished as early as possible becomes important. He investigates two kinds of scheduling algorithms for the problem. The first one is nonpreemptive scheduling, i.e., no job is allowed to be interrupted during its execution. In this case, the problem is NP-Complete. He proposes an approximation algorithm called LDF, which generates a schedule with a finish time less than twice that of an optimal schedule. Compared with the earlier proposed algorithm, his algorithm is simpler and has almost the same performance. More importantly, his LDF algorithm can achieve this performance without knowing the job processing times, which may be hard to obtain in practice. Also he proves a lower bound result which implies that it is unlikely to find simple heuristic algorithms that can perform much better than the existing algorithms including LDF. The second kind is preemptive scheduling, i.e., a job can be preempted during its execution and rescheduled later. He develops a feasibility algorithm that runs in O (n log n) time and generates a schedule with at most min{l brace}n-2, 2{sup m}-1{r brace} preemptions. It can generate a feasible schedule for the given job set if there exists one. This improvement is important because many scheduling algorithms depend on a feasibility algorithm as a building block. Furthermore, based on an advanced search technique, he presents an algorithm that can find the optimal schedule in O(n{sup 2} log {sup 2}n) time.
- Research Organization:
- Ohio State Univ., Columbus, OH (United States)
- OSTI ID:
- 6087363
- Country of Publication:
- United States
- Language:
- English
Similar Records
Strategies for job allocation in a hypercube concurrent computer
Parallel-sorting algorithms for hypercube multiprocessors
Scheduling jobs that arrive over time
Thesis/Dissertation
·
Sat Dec 31 23:00:00 EST 1988
·
OSTI ID:5923068
Parallel-sorting algorithms for hypercube multiprocessors
Thesis/Dissertation
·
Sat Dec 31 23:00:00 EST 1988
·
OSTI ID:6155985
Scheduling jobs that arrive over time
Conference
·
Thu Apr 06 00:00:00 EDT 1995
·
OSTI ID:64945