Packet-Level Network Compression: Realization and Scaling of the Network-Wide Benefits

Abstract: The existence of considerable amount of redundancy in the Internet traffic at the packet level has stimulated the deployment of packet-level redundancy elimination techniques within the network by enabling network nodes to memorize data packets. Redundancy elimination results in traffic reduction which in turn improves the efficiency of network links. In this paper, the concept of network compression is introduced that aspires to exploit the statistical correlation beyond removing large duplicate strings from the flow to better suppress redundancy. In the first part of the paper, we introduce "memory-assisted compression", which utilizes the memorized content within the network to learn the statistics of the information source generating the packets which can then be used toward reducing the length of codewords describing the packets emitted by the source. Using simulations on data gathered from real network traces, we show that memory-assisted compression can result in significant traffic reduction. In the second part of the paper, we study the scaling of the average network-wide benefits of memory-assisted compression. We discuss routing and memory placement problems in network for the reduction of overall traffic. We derive a closed-form expression for the scaling of the gain in Erdos-Renyi random network graphs, where obtain a threshold value for the number of memories deployed in a random graph beyond which network-wide benefits start to shine. Finally, the network-wide benefits are studied on Internet-like scale-free networks. We show that non-vanishing network compression gain is obtained even when only a tiny fraction of the total number of nodes in the network are memory-enabled.