Bounds on Gang Scheduling Algorithms

Abstract : We define the bounds of two gang service algorithms, namely time balanced and instruction balanced, using interactivity and throughput based metrics. The workload is composed of V-RAM jobs, and the architectural model chosen is LogP. We also analyze the effect of workload and space sharing strategies on the bounds found. The following results are demonstrated: when interactivity is considered, instruction balanced gang service is the best possible non-clairvoyant algorithm. When the objective is to increase throughput Instruction balanced and time balanced gang service are equivalent algorithms. In this case the preemption strategy is not a determinant factor in performance
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01574110
Contributor : Lip6 Publications <>
Submitted on : Friday, August 11, 2017 - 3:45:13 PM
Last modification on : Thursday, March 21, 2019 - 1:01:00 PM

Identifiers

  • HAL Id : hal-01574110, version 1

Citation

Fabricio Alves Barbosa da Silva, Isaac Scherson. Bounds on Gang Scheduling Algorithms. 2nd International Conference on Parallel Computing Systems, Aug 1999, Ensenada, Mexico. ⟨hal-01574110⟩

Share

Metrics

Record views

16