OpenFlow implements flow-based control over switches with improved network management performance. However, a packet may match more than one flow entry due to the intra-table dependency phenomenon among flow entries. Moreover, different packets may incur different conflicting flow entries under the intra-table dependency. Forwarding packets by the first-match scheme for prioritized flow entries may not always produce the best outcome. Thus, an online conflict detection procedure executed for each incoming packet is needed to flag the conflicts to network administrators. In addition, the SDN controller may frequently update the service provisioning policies that are specified in the flow entries and deliver them to the switches in a large OpenFlow-based environment. This needs a high-performance conflict detection mechanism to support real-time updating. However, performing conflict detection within a large flow table will be very time consuming. This paper first develops a graph-based multilevel redundancy reduction scheme to construct highly compact matching trees that will be used in conflict detection for a large flow table. Then, a conflict detection algorithm with higher performance and lower cost, the Compact Bit Vector algorithm (CBV), is proposed. The performance of the CBV has been validated through an extensive mathematical performance analysis followed by simulations, with good results in terms of requiring less time for the search, lower memory requirement and lower incremental updating time. Obviously, the CBV is very suitable for the conflict detection task of a large and frequently updated flow table.

OpenFlow通过改进的网络管理性能对交换机实施基于流的控制。然而，由于流条目之间的表内依赖性现象，一个分组可能匹配一个以上的流条目。此外，在表内依赖性下，不同的分组可能招致不同的冲突流条目。对于优先流条目，通过优先匹配方案转发数据包可能不会始终产生最佳结果。因此，需要为每个传入数据包执行的在线冲突检测过程，以将冲突标记给网络管理员。此外，SDN控制器可能会频繁更新流条目中指定的服务供应策略，并将其交付给大型基于OpenFlow的环境中的交换机。这需要一种高性能的冲突检测机制来支持实时更新。但是，在大型流表中执行冲突检测将非常耗时。本文首先开发了一种基于图的多级冗余减少方案，以构建高度紧凑的匹配树，该树将用于大型流表的冲突检测。然后，提出了一种性能更高，成本更低的冲突检测算法，即紧凑位向量算法（Compact Bit Vector Algorithm，简称CBV）。CBV的性能已通过广泛的数学性能分析以及随后的仿真得到验证，其结果是，所需的搜索时间更少，内存需求更低，增量更新时间更少。明显，