Studies have indicated several situations in which network congestion is not generally caused by the sending source. Rajaie et al [1] shows that, in a shared best-effort network with a high level of statistical multiplexing, the observed loss pattern has a near random behavior that is determined by the aggregate traffic. Similarly,the experimental results
by Bolot [2] show that the losses of probe packets are essentially random unless the probe traffic shares a large fraction of the available bandwidth. These studies suggest that network congestion as indicated by packet loss is not necessarily self-induced, but can
also be cross-induced. This is particularly true for
media streaming in which its intrinsic bandwidth
requirement could be a small fraction of the available
bandwidth.