A Dynamic Multipath File Transfer Engine for Software-Defined Networking

Abstract

With the fast development of the internet, it is widely expected that data traffic will grow exponentially. To fulfill the internet users’ demand of Quality of Experience (QoE), the way we deal with “the tsunami of data” becomes a big problem. How to transfer data, especially those very big files, as fast as possible can be a very important part of that problem. One promising solution could be using multiple paths instead of a single path transmission. Different conditions on different paths make it difficult to handle multipath transmission well. However, Software-Defined Networking (SDN), a centralized network model, provides new solutions to make multipath transmission possible. In a SDN network, the network status can be monitored. With such monitoring information, an optimal multipath solution can be found to transfer a file as fast as possible.
In this thesis, the optimal multipath transmission problem is studied. The objective is to minimize file transfer time between two end systems in a given SDN network. First, we model this as a linear programming problem on a time-expanded network. The time-expanded network used makes that the complexity of this solution is pseudo-polynomial. From this solution, we are able to derive a Fully Polynomial-Time Approximation Scheme (FPTAS) where we can accurately control the speed versus accuracy tradeoff. Lastly, we have build an SDN-based proof-of-concept implementation through which we have evaluated our two algorithms.