Flow Entry Redundancy Reduction for Software-Defined Networking Multi-Flow Tables

  • 巫 郁槤

Student thesis: Master's Thesis


Software-Defined Networking (SDN) has been proposed as a promising paradigm in the next generation networks SDN addresses the network management problem occurred in the static architectures of conventional networking environment It decouples the software-based control plane from the hardware-based data forwarding plane and utilize centralize remote controller to manage the entire network environment With the programmable controller the costs of operating a network is reduced and network administrators are able to manage network resources in a more flexible fashion to satisfy the ever-changing networking requirements OpenFlow is the first standard communication protocol implemented for SDN Data packets transmitted in OpenFlow is treated as a flow and flow-based control is utilized over OpenFlow-enabled switches to determine how data packets should be forwarded across the network For a more agile and flexible management over the network multi-flow table pipeline has been proposed in the later version of OpenFlow specifications to improve the problems when handling a large set of diverse flow entries in a single flow table Therefore the effectiveness of the packet classifier becomes the key factor when searching for corresponding flow entries in high traffic loads Ternary Content-Addressable Memory (TCAM) is the most commonly used component in both industry and academia to address the packet classi?cation performance issues This is because the processing time of routing and flow entry lookup is much lower than random access memory However it is not suitable for the network with large number of flow entries due to its limitation in capacity and its high cost Furthermore when multi-flow table pipeline being used in large number of flow entries sometime redundant flow entries is unavoidable due to unintentional mistake or repeated flow entries made by different network administrators while adding new flow entries In order to address these problems a mechanism is needed not only to decrease flow entry lookup time but also to identify the redundant flow entries in the single or multi-flow tables Thus this thesis proposes a multi-flow table reduction algorithm to overcome the redundant flow entry problem in OpenFlow to reduce the number of redundant entries in single or multi-flow tables The algorithm can be divided into two parts multi-flow tables to a single flow table conversion and redundant flow entry reduction by constructing an all-match tree Two optimization methods were used to reduce the process time and memory usage requirement while constructing this tree From the experiment results the multi-flow tables reduction algorithm can effectively reduce the number of flow entries in single and multi-flow tables with tolerable time As a result the proposed algorithm reduces the processing time of flow table lookup and the memory requirement in multi-flow tables after performing flow entry redundancy reduction mechanism
Date of Award2016 Sep 6
Original languageEnglish
SupervisorYau-Hwang Kuo (Supervisor)

Cite this