On Cost Efficient Dataflow Computing Program Deployment in SDN Managed Distributed Computing Environment

Abstract

Dataflow computing has been regarded one of the most promising computing paradigms in the big data era. With the vast distribution of data sources, it is significant to deploy the dataflow based applications in distributed environment to digest these data. In dataflow computing, the data flows shall be transferred between different processing units to the accomplish the predefined semantics. Software-defined networking (SDN) has emerged as an effective network management technology to orchestrate the data flows among these processing units. For each data flow, a forwarding rule shall be inserted into the forwarding table of each switch on the routing path. However, the number of rules that can be inserted in one forwarding table is limited. We are motivated to take such constraints into the consideration of dataflow applications deployment in distributed computing environment managed by SDN. An efficient deployment algrotithm is proposed and evaluated in this paper.