Heterogeneous Flow Table Distribution in Software-defined Networks

• Main Authors: Chun-Feng Wang, 王竣鋒
• Other Authors: Guey-Yun Chang
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/yzd7s5

碩士 === 國立中央大學 === 資訊工程學系 === 103 === Recently, Software-defined Network (SDN) has become an important and popular technology which provides for the flexibility of developing new protocols and the policies of real networks. The controller in SDN translates network policies into rules which are installed in the flow tables (Flow tables are usually stored in ternary content addressable memory (TCAM)) of switches in the networks. Hence, TCAM usually has some critical disadvantages (e.g., high costs, power consumption and high heat generation). Flow tables cannot scale beyond a few hundred entries. Therefore, switches may need to cache rules reactively (i.e., installing rules on demand). However, when cache misses happen, switches will send the packet-in message to the controller and reactively cache the rules, which causes packet delay and large buffers. In this thesis, we propose a rule partition and allocation algorithm that distributes rules across network switches to improve the performance. Our algorithm not only is applicable to small TCAM switch scenario, but also guarantees semantically-invariant (i.e., the global action of the network is unchanged). We implement our algorithm into the real world SDN scenario and the experiment result shows that our algorithm have obviously reduced TCAM usage.