Poisson convergence, in large deviations, for the superposition of independent point processes

For a small buffer queueing system fed by many flows of a large class of traffic processes we show the single server queue and associated sample paths behave as if fed by marked Poisson traffic in a large deviations limit.The timescale of events of interest tends to zero, so we study the log moment generating function as time tends to zero. The associated rate function depends only on the mean arrival rate and the moment generating function of the arrivals. These results are useful in estimating drop probabilities while studying the effect of small buffers on communication protocols.

[1]  Jin Cao,et al.  A Poisson limit for buffer overflow probabilities , 2002, Proceedings.Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies.

[2]  Damon Wischik,et al.  Big Queues, Springer Lecture Notes in Mathematics , 2004 .

[3]  Jin Cao,et al.  Internet Traffic Tends Toward Poisson and Independent as the Load Increases , 2003 .

[4]  G. Raina,et al.  Buffer sizes for large multiplexers: TCP queueing theory and instability analysis , 2005, Next Generation Internet Networks, 2005.

[5]  Amir Dembo,et al.  Large Deviations Techniques and Applications , 1998 .

[6]  Nick McKeown,et al.  Part I: buffer sizes for core routers , 2005, CCRV.

[7]  Derong Liu.  The Mathematics of Internet Congestion Control , 2005, IEEE Transactions on Automatic Control.

[8]  Damon Wischik,et al.  Big queues , 2004, Lecture notes in mathematics.

[9]  A. Weiss.  A new technique for analyzing large traffic systems , 1986, Advances in Applied Probability.