An initial approximation to the resource-optimal checkpoint interval

Abstract

In this paper, some basic considerations are made about cost-effective checkpointing in a cloud-like environment. We deduce a cost model of a task considering expected costs of computational and storage resources and time-to-solution costs. With the simple analytic expression, we calculate efficient checkpoint intervals relying on cost parameters, taken from real-world computational clouds. Results demonstrate reasonability of the model proposed and we discuss its potential implications in a high-performance computing environment. © 2011 Springer-Verlag Berlin Heidelberg.