SDN enabled BDSP in public cloud for resource optimization

Abstract

Novel computing paradigm realized by cloud computing and virtualization technologies paved the way for commoditization of computing resources. Clouds and their federation made inexhaustible computing resources leveraging ample scope in producing opportunities and productivity with provision for on-demand resources in pay as you go fashion. With the wealth of resources, big data and big data stream processing (BDSP), big data analytics became a reality now. Two stakeholders such as cloud service providers and cloud users are mainly affected if the cloud infrastructure fails to deliver intended services to the satisfaction end users. Resource optimization has been an active research topic in cloud computing to overcome this problem. It is more so with the emergence of Software Defined Networking (SDN). Resource reservation and dynamic resource allocation are two approaches found in the literature. Dynamic resource allocation is highly preferred optimization problem considered in this paper. BDSP needs highly reliable and automated resource optimization in the context of increased big data streaming workloads to be processed by real-world applications. In this paper, we proposed a methodology for SDN enabled BDSP in public cloud for resource optimization. We defined a mathematical model and proposed an algorithm to achieve it. CloudSimSDN is used to build a prototype application that demonstrates proof of the concept. Our experimental results reveal the utility of SDN based approach for resource optimization in a cloud in the presence of BDSP by decoupling data forwarding and network controlling.