A reliable and scalable service bus based on Amazon SQS

Abstract

Cloud computing infrastructures are becoming a very powerful mean for the implementation of reliable and extensible computing systems. In this paper, we evaluate the viability of migrating a framework for the execution of (scientific) workflows from a cluster-based to a cloud-supported implementation. As a first step, we focus on the viability of adapting the framework message bus (which has a Linda semantics) to the use of the Amazon Simple Queue Service (Amazon SQS). The paper evaluates the performance of the cloud-based bus and studies the influence of the network latency, depending on different geographical locations and configurations. It also compares the cloud-based bus with DRLinda, our former implementation, in terms of economic cost and performance. This comparison allows us to conclude that, under the same conditions, the cloud-based message bus is faster, more scalable and more reliable. © 2013 Springer-Verlag.