Generalized TCP congestion avoidance and its effect on bandwidth sharing and variability

Abstract

Abstract-To model possible suggested changes in TCP window adaptation in response to randomized feedback, such as ECN, we formulate a generalized version of the TCP congestion avoidance algorithm. We first consider multiple such generalized TCP flows sharing a bottleneck buffer under the Assured Service model and use a fixed point technique to obtain the mean window sizes and throughputs for the TCP flows. To further study how changes in the adaptation algorithm affect the variability in the throughput, we use an analytical-cum-numerical technique to derive the window distribution (and related statistics) of a single generalized flow under state-dependent randomized congestion feedback.

Predicting bottleneck bandwidth sharing by generalized TCP flows

“…The fixed-point method for analyzing the sharing of a bottleneck buffer was first presented in [8,9] later presented a similar analysis based on a control-theoretic model. The extension of the fixed-point technique for generalized TCP flows was investigated in [10]. This paper combines the results in [8,10] into a common framework.…”

“…The extension of the fixed-point technique for generalized TCP flows was investigated in [10]. This paper combines the results in [8,10] into a common framework. Ref.…”

“…Ref. [8] shows how such a negative correlation can be exploited to obtain more accurate estimates of the stationary distribution of the congestion windows, while [10] discusses how such negative correlation is exploited by intelligent Ôdrop-biasingÕ strategies to reduce the variability of the queue occupancy. Fig.…”

“…Ref. [10] also uses this approach to consider how changes to the adaptation parameters affect the bandwidth sharing paradigm between multiple TCP flows. Such studies are useful in evaluating the possible effects of suggested changes to TCPÕs current congestion control algorithm.…”

“…In packets are provided preferential treatment in network buffers via the RIO (RED with In/Out) discard algorithm; RIO is similar to RED except that it uses different thresholds for in and out packets to ensure that out (opportunistic) packets are dropped before in packets. For simulation-based studies involving the generalized congestion avoidance algorithm, we assume that our bottleneck queue uses the ORED [10] buffer management algorithm; ORED is similar to RIO but differs in two respects:…”