Parallel Execution of For Loops Using Checkpointing Techniques

Abstract

Checkpointing techniques have been used to secure sequential and parallel programs for many years. In this article, we show how checkpointing techniques can be used in order to transform and execute for loops of sequential programs on a parallel machine, especially on a cluster or on a grid. This article also presents the implementation we have developed to valid the concept and some performance measurements.

Improving Performance of CAPE Using Discontinuous Incremental Checkpointing

“…This transformation uses two parallel sections for the example. However, it can be generalized to an as-large-aspossible set of sections [16] and to loops [17]. In the present example, the parent, i.e.…”

“…BOINC [18], used in the scope of the Seti@Home project, is probably one of the most famous tool aiming at distributing works among a set of computing resources. For an in-depth description of CAPE, refer to [16] and [17].…”

Design and performance analysis of CAPE based on discontinuous incremental checkpoints

“…Apart from the above two categories, we have been developing CAPE (which stands for Checkpointing Aided Parallel Execution) [14] [15], a method that uses checkpointing technique to execute OpenMP programs on distributed architectures. The first prototype of CAPE proved the feasibility of the approach but the use of a complete checkpointer as the base tool strongly decreased the global performance.…”

Discontinuous incremental: A new approach towards extremely lightweight checkpoints

“…CAPE [1], [2] stands for Checkpointing Aided Parallel Execution. It consists in modifying a sequential program so that instead of executing each part the one after the other on a single machine, parts are distributed over a set of machines to be executed in parallel.…”

“…Some attempts have tried to port OpenMP on distributed memory architectures with various success. In order 1 viet hai.ha@it-sudparis.eu 2 eric.renault@it-sudparis.eu to achieve this goal, we have developed a new parallel computing paradigm called CAPE which aims at using checkpoints to distribute the execution of a program on a distributed-memory environment [1], [2]. However, both existing checkpointing techniques cannot be used for an effective implementation.…”