Research On Network Resource Allocation Based On Software Defined Network

SDN (Software Defined Network) is a new type of network architecture, and NFV (Network Function Virtualization) is a new concept about realizing complex network function by virtualization, both concepts were proposed in recent years. Implementing NFV using SDN has an advantage of totally deployment automation, undoubtedly, it has some progressive meaning.However, current network environment is very complex, implementing NFV using SDN must lead to fine-grained rules, and the very fine-grained rules must put a high pressure on physical hosts which are the carriers of implementing NFV. For solving the problem above, the subject of the paper is allocating network resources, especially processing resources (like CPU) optimally, making network resources evenly shared in the whole network.Our main work includes:Firstly, we model the processing resources allocation problem with linear programming, the object function of the model is maximizing the residual processing resource of the host which is allocated the most ratio of the network flow, the object function is used for realizing the load balancing object; The model’s constraint conditions can be divided into three categories:bandwidth constraint, flow conservation constraint and processing resource limition constraint. We solve and analysize the optimal solution of specific topologies using linear optimation tools.Secondly, we design the unsplittable flow approximation algorithms. For the two kinds of topology:symmetrical topology, such as Fat-Tree, and asymmetrical topology—modeled by GT-ITM tool, the name of the algorithms are respectively allocation the network flow along the shortest path according to the scale coefficient and allocation the network flow along the optimal simple path. In order to evaluate the efficient of the algorithm, we test two parameters which were load balancing and maximum available flow. At last, we compare our algorithms with linear programming model by evaluating the above two parameters.Thirdly, we design the flow table rules based on multi-level flow table form, and we discuss further about how rules change with network dynamics.Finally, we end the paper and point out the direction of future research in this subject.