A Flow-level Monitoring Middleware for Automatic Flow Categorization

Wang Xuliang (1451211)


Current network systems are not completely optimized for all applications, which lead to inefficient utilization of the network. Actually, each network application has different requirements of bandwidth and latency for network. However, the current network systems determine routing paths based only on topological information of network. In order to allocate more efficient network paths for each application, it is necessary to understand the communication patterns of each application. In this paper, I propose a flow-level monitoring and categorizing middleware called AppMon, which enables to understand the communication patterns of each application by leveraging OpenFlow technology that allows me to monitor each applications behavior in a flow-level manner. I have also figured out appropriate categorization thresholds and monitoring parameter sets based on the analysis of pubic packet capture datasets.