Date of Award
Angela C. Sodan
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
In parallel computing, jobs have different runtimes and required computation resources. With runtimes correlated with resources, scheduling these jobs would be a packing problem getting the utilization and total execution time varies. Sometimes, resources are idle while jobs are preempted or have resource conflict with no chance to take use of them. This greatly wastes system resource at certain degree.
Here we propose an approach which takes periodic checkpoints of running jobs with the chance to take advantage of migration to optimize our scheduler during long term scheduling. We improve our original Scojo-PECT preemptive scheduler which does not have checkpoint support before. We evaluate the gained execution time minus overhead of checkpointing/migration, to make comparison with original execution time.
Shi, Jiaying, "Extending Scojo-PECT by migration based on application level checkpointing" (2009). Electronic Theses and Dissertations. 7923.