Model and Analysis of the Checkpointing Aided Parallel Execution Technique

摘要

Checkpointing techniques have been used for years to secure the execution of sequential and parallel programs. However, it has been demonstrated recently that they can also be used to generate a parallel program from a sequential source code automatically. As a result, this technique allows any programs to be executed on any kind of distributed parallel system. This article presents the analysis of this innovative technique called CAPE for Checkpointing Aided Parallel Execution, and a model for performance evaluation. As an example, the presented model is used to provide some hints to determine the optimal number of processors to run the automatically generated parallel code on a distributed system.