Profiling-as-a-service: Adaptive scalable resource profiling for the cloud in the cloud

Abstract

Runtime profiling of Web-based applications and services is an effective method to aid in the provisioning of required resources, for monitoring service-level objectives, and for detecting implementation defects. Unfortunately, it is difficult to obtain accurate profile data on live client workloads due to the high overhead of instrumentation. This paper describes a cloud-based profiling service for managing the tradeoffs between: (i) profiling accuracy, (ii) performance overhead, and (iii) costs incurred for cloud computing platform usage. We validate our cloud-based profiling service by applying it to an open-source e-commerce Web application. © 2011 Springer-Verlag.