Queuing models of synchronous compressors

Summary form only given. In synchronous compression, a lossless data compressor attempts to equalize the rates of two synchronous communication channels. Synchronous compression is of broad applicability in improving the efficiency of internetwork links over public digital networks. The most notable features of the synchronous compression application are the mixed traffic it must tolerate and the rate buffering role played by the compression processor. The resulting system can be modeled in the time domain by queuing methods. The performance of a compression algorithm in this application is governed by the interplay of its ultimate compression ratio, its computational efficiency, and the distribution function of its instantaneous consumption rate of the source. The queuing model for synchronous compression represents the compressor as the server fed by a single queue. We describe the basic model, develop the required basic queuing theory, look at service time statistics, and compare to simulation. We develop the queuing model for synchronous compression and relate it to theoretical and empirical properties of queuing systems and Lempel-Ziv compression algorithm performance. We illustrate that synchronous compression simulations are in agreement with the predictions of queuing theory. In addition, we observe various interesting properties of match length distributions and their impact on compression in the time-domain.