IP header compression: a study of context establishment

The use of bandwidth constrained links in wireless networks necessitates the use of bandwidth saving header compression schemes. In these schemes, a compressor and a decompressor collaborate to encode bulky IP headers into streamlined compressed headers. Intuitively, the gain from compression is the average compressed header length divided by the uncompressed header size. In this paper, we show that this is not true in many situations: this does not account for the cost introduced by the variability of the header sizes during compression context initialization. First, we provide analytical basis for cost of compression context creation and compare the relative costs of establishing contexts with an optimistic approach and with an acknowledgement-based approach. We conclude that the optimistic approach can be used as an upper bound in our analytical model. Second, we evaluate the impact of header compression on traffic patterns in terms of bandwidth allocation, in order to accommodate the burst introduced by compression context initialization. We compute the actual compression ratio, that is the actual number of users of IP header compression that can be multiplexed onto a given link. We show this number is significantly different from the one computed by dividing the link bandwidth by the average rate of a compressed flow. Our results provide a formal basis for context management especially during handovers in mobile networks. For example, our result indicates that there is significant benefit to relocate compression contexts (from one network node to another) rather than to re-establish them each time during handovers.