An efficient flow monitoring algorithm using a flexible match structure
Ze Yang, Kwan Lawrence Yeung · 2016
We focus on designing efficient flow monitoring algorithms in SDN/OpenFlow by fully exploiting all three polling mechanisms, poll-single, poll-some and poll-all. Notably, the poll-some mechanism has not been adopted by any existing flow monitoring algorithm due to the inflexible match structure standardized in the early version of the OpenFlow specification. To enable the poll-some mechanism, we need to find out the minimum number of match structures required to exactly match all not-yet-covered flows at a switch. An efficient heuristic called Critical Column First (CCF) is then proposed for solving the Minimum Match Structure (MMS) problem. The idea is to consider the column of the flow matrix that can potentially identify the largest number of not-yet-covered flows first. With CCF, an existing flow monitoring algorithm called LFF [1] is extended to support all three polling mechanisms. We call it LFF+ algorithm. As compared with LFF, simulation results show that LFF+ can cut down the communication cost by about 54%.