Modeling variable bit rate video on wired and wireless networks using discrete-time self-similar systems

In recent years, wireless networks have gained a lot of interest due to the increasing demand for Internet based wireless applications. Due to bandwidth constraints video traffic over the Internet is compressed. This type of compressed traffic shows burstiness on different time scales. In particular, the phenomena of long-range dependence (LRD), self-similarity and heavy-tailed distributions appear to play a prominent role in modeling VBR video traffic. In this paper we present novel discrete-time models based on self-similar considerations to characterize the statistical properties of VBR video traffic. We analyze MPEG VBR video traces that were collected on a wired network as well as multimedia traffic obtained at the access points (AP) on a wireless LAN based on IEEE 802.11b implementation. The results show that these traffic traces display long-range dependence and heavy-tailed behavior and the discrete-time statistically self-similar model discussed here can not only generate traces with the measured Hurst parameter but also provide better fits to the autocovariance function of the time-domain data than the existing models such as Markovian, LRD and M/G/∞.