Supporting Efficient Execution of Workflows on Everest Platform

Abstract

Workflows is an important class of parallel applications consisting of multiple tasks with control or data dependencies between them. Such applications are widely used for automation of computational and data processing pipelines in science and technology. The paper considers the execution of workflows on Everest, a web-based distributed computing platform. Two different approaches for workflow description and execution are presented and compared. The proposed solutions and optimizations address common problems related to the efficient execution of workflows on distributed computing resources. The results of experimental study demonstrate the effectiveness of proposed optimizations.