DDSoR: A Dependency Aware Dynamic Service Replication Strategy for Efficient Execution of Service-Oriented Applications in the Cloud

Abstract

With the attractive features of service oriented architecture (SOA), such as the high re-usability, the fast development and the reduced cost, many large scale distributed applications are currently delivered within cloud data centers following the SOA style (namely service-oriented applications). A service-oriented application is composed of a set of reusable services, each of which may be used concurrently by multiple applications. A key challenge in this context is how to guarantee an efficient execution of these applications to quickly and effectively respond to the ever-increasing incoming requests. Service replication has been touted as an efficient technique that supports the non-functional requirement of services, specified in Service- Level-Agreement (SLA) between cloud service providers and consumers, such as availability, reliability and response time, through providing multiple replicas of a given service. Several service replication strategies for cloud computing context were proposed in the literature. The major shortcoming of these attempts is their neglect of service dependencies. In view of this, we propose a Dependency aware Dynamic Service Replication Strategy called DDSoR. The main goal of the proposed strategy is to minimize execution time of distributed service oriented applications by clustering dependent services and replicating them in the same server. The service replica cluster placement problem is taken as a three dimensional multiple knapsack problem.