A reusable architecture for dependability and performance benchmarking of cloud services

Abstract

With the increasing demand and benefits of cloud computing services, new solutions are needed to benchmark the dependability and performance of these services. Designing a dependability and performance benchmark that covers a variety of fault and execution scenarios, poses various architectural challenges. In this paper, we present a generic software architecture for dependability and performance benchmarking for cloud computing services. We provide the details of this generic architecture i.e. various components and modules, that are responsible for injecting faults in cloud services in addition to the components responsible for measuring the performance and dependability. We make use of this architecture to build two software prototypes: MRBS and MemDB. These prototypes are used to benchmark two popular cloud services: MapReduce and Memcached. The case studies with the use of software prototypes demonstrates the benefits of building a generic architecture.