BatchUp: Achieve fast TCAM update with batch processing optimization in SDN

Abstract

The parallel access mechanism supported by TCAM has made it popular with commercial switches. However, due to the ordered arrangement of rules, TCAM’s update speed cannot keep up with policy update requests, especially for software-defined networks, which expect to achieve fine-grained control over the network. Meanwhile, the legacy single-rule update solutions cause redundant rule moves due to weak global view. In this paper, we present a new two-stage batch processing solution BatchUp to alleviate the high time cost of the rule update process. In stage 1, we propose the costless rule insertion algorithm CRIA. CRIA models the rules to be inserted with the available locations in TCAM as a bipartite graph. By solving for maximum matching, CRIA obtains an upper bound on the number of rules that can be inserted without TCAM movement. However, the high utilization of TCAM makes it unrealizable to find available locations for all rules to be inserted directly. Therefore, in stage 2, we model the remaining rules insertion as an optimization problem for solving the minimum movement cost and prove that it has NP-hard complexity. A heuristic updating algorithm HRIA based on simulated annealing and greedy algorithm is proposed to search for approximate optimal solution. The evaluation results present that BatchUp has advantages over existing state-of-the-art solutions in terms of the number of TCAM moves and overall time cost, and has applicability and robustness in different scenarios.