cUpdate: Updating Software-Defined Networks with the Least Traffic Migration

摘要

Dynamic flows may lead to congestion, deadlock, packet loss and forwarding loops in a dedicated network. The update process is sensitive to the steps it took. Given the initial and final states of traffic distribution, diverse update strategies will generate unpredictable results, not to mention the unknown final state situations. Currently, most methods only consider about migration process. However, no one care whether the target state is the optimized state. To tackle this essential issue, we propose cUpdate, an optimized correlation update strategy which will migrate the least traffic and suffers less packet loss. We further propose two efficient algorithms to generate the proper final state and calculate the sequence of update steps. We implement a prototype of cUpdate in a Software Defined Network (SDN) by using OpenFlow switches. The comprehensive experiments show that cUpdate can cause less number of traffic migrations than prior methods and is capable of avoiding deadlock and selecting a better final state with least packet loss(0:28%).