A Novel Approach to Reduce Bandwidth Cost and Balance Network and Server Level Load in Intra Data Center Network

Abstract

These days, cloud computing is very popular because of its on demand utility like services. To mitigate the huge demand generated by the consumers of different level from individual to enterprise, the operational cost is also increasing. Furthermore, to retain delay within tolerable limit for sensitive applications, particularly during sudden surge of traffic is a great challenge. Data Center Network (DCN) is the underlying physical infrastructure of cloud computing. Therefore, to reduce the operational cost and delay, a framework is required to optimally manage the resources within a data center. In this paper, we consider the requests from enterprise clients mapped to a virtual network that requires both bandwidth and processing assets by interfacing it from an entry point to at least one server of a DCN. We present an efficient traffic engineering scheme based on a Mixed Integer Linear Programming model. This scheme from the ingress entry point to egress server is used in DCN to minimize the cost of establishing links between an entry point and a server while maintaining load balancing in both the link and server level. Load balancing helps to reduce delay. The performance of our model is discussed by conducting a trade-off analysis between different objectives of the composite objective function. We also investigate how various network parameters like bandwidth cost, maximum link and server resources utilization changes by setting different level of priority on different objectives.