Exploiting graph compression techniques for coding and monitoring of networks

In order to monitor large enterprise networks, network monitoring data is often gathered from a wide range of sensor locations across the network. However, such activities create bottlenecks for existing network applications traffic flows. Network data collected at remote sensor sites must be transmitted to the central network monitoring station efficiently. The solution presented in this paper involves encoding network information via compressed network graphs. We devise a graph compression scheme termed Variable-Interval-Coding, which encodes intervals of multiple network graphs together. The interval-coding scheme offers two key benefits: (i) network information monitored at remote sites is communicated back to the central site only when significant network deviation events occur and, (ii) the network change information is captured and delivered using compressed graphs. The proposed solution is implemented within a `remote-to-central-site´ monitoring system for analyzing network changes detected by remote sensors in the network. Besides data storage and transmit bandwidth savings, we show that equivalent network change analysis can be achieved using compressed network graphs, instead of having to examine the full collection of network data from these remote sites.