Modelling TCP Reno with spurious timeouts in wireless mobile environments

Abstract

Abstract-TCP has been found to perform poorly in the presence of spurious timeouts (ST) caused by delay spikes which are especially more frequent in today's wireless mobile networks than in traditional wired network. Because STs do not frequently occur in wired networks, and are generally considered to represent a transient state, previous research did not consider the effects of ST on the steady state performance of TCP. However, ST is more frequent in wireless mobile environments, and must be considered explicitly to accurately model the steady state sending rate and throughput of TCP. In this paper, we propose an analytical model for the sending rate and throughput of TCP Reno as a function of packet error rate and characteristics of spurious timeouts. The accuracy of the proposed model has been validated against simulation results. The accuracy of the model has also been compared with previous models, and has been found to be more accurate than previous models in the presence of spurious timeouts.

An algorithm to detect TCP spurious timeouts and its application to operational UMTS/GPRS networks

“…This makes difficult to measure its occurrence accurately, because it can only be observed when it is effectively used 4. Some little discrepancies in the table results are due to numerical approximations.…”

“…Refs. [3][4][5] propose analytical models of TCP SRTOs and validate the models by simulation. Additionally, several proposals exist to upgrade TCP such that it is better able to avoid SRTOs.…”

Performance of TCP on low‐bandwidth wireless links with delay spikes

“…Work related to modeling spurious timeouts has been done in [8], where a model for TCP Reno experiencing spurious timeouts has been given which very closely follows the operation of the actual TCP Reno protocol. The model is based on an extension of the well known model in [22] and considers a large bandwidth delay product, with no sending rate limitations (we include the sending rate limit and congestion losses).…”

“…In comparison to [8] and [1], using a more abstract representation of how the TCP's goodput accumulates (and not attempting to model the protocol itself), we obtain a model that is straightforward to utilize and which allows quite general RTT spike processes. Finally, in comparison to our earlier paper [16], the present paper includes the impact of congestion losses (i.e., a lossless scenario is not assumed anymore).…”

“…Numerically we have observed that a very good approximation of E(A | T ) is given by taking the average of the upper and lower bounds in (8),…”