Emulating AQM from end hosts

Abstract

In this paper, we show that end-host based congestion prediction is more accurate than previously characterized. However, it may not be possible to entirely eliminate the uncertainties in congestion prediction. To address these uncertainties, we propose Probabilistic Early Response TCP (PERT). PERT emulates the behavior of AQM/ECN, in the congestion response function of end-hosts. We present fluid-flow analysis of PERT/RED and PERT/PI, versions of PERT that emulate router-based RED and PI controllers. Our analysis shows that PERT/RED has better stability behavior than router-based RED. We also present results from ns-2 simulations to show the practical feasibility of PERT. The scheme presented here is general and can be used for emulating other AQM algorithms.

Less-than-Best-Effort Service: A Survey of End-to-End Approaches

“…Some of these proposals, like Compound TCP [21], TCP Illinois [22] and RACC [23], are hybrid loss-and delay-based mechanisms, whereas others (e.g., FAST TCP [24], NewVegas [25], or CODE TCP [26]) are variants of Vegas based primarily on delays. 4 Though not specific to Vegas, there are also proposals, such as [27], [28], aiming at improving the coexistence of loss-based and delay-based congestion-controlled flows.…”

“…Bhandarkar et al [27] claim that it is possible to significantly improve the accuracy of delay-based congestion prediction by adopting some simple techniques (smoothing of RTT samples, increasing the RTT sampling frequency). Nonetheless, they acknowledge that even with such techniques, it is not possible to eradicate detection errors.…”

Improved coexistence and loss tolerance for delay based TCP congestion control

“…This idea was enhanced (in a similar manner to PERT [15,19]) by Budzisz et al [1] to improve fair coexistence with loss based TCP algorithms 1 . On receipt of every ack, cwnd (w) is evaluated as follows:…”

“…Multiplicative decrease on loss. Compound TCP (CTCP) [14] smoothed RTT similar to Vegas Dual mode: When delay conditions are favorable it is Reno AMID plus MIMD (two windows are used (delay and loss), otherwise it is Reno AIMD Probabilistic Early Response TCP (PERT) [15] RTT and smoothed RTT dynamic thresholds based on inferred queueing delay (q j = τ j − τ min ) Random Early Discard (RED) inspired probabilistic reaction to queueing delay, with loss probability matching when q j ≥ 0.5qmax. Hamilton Delay [1,16] RTT threshold based on inferred queueing delay Probabilistic reaction based on queueing delay and a back off function (see III) …”

A strategy for fair coexistence of loss and delay-based congestion control algorithms

“…Motivated by these and other potential benefits, delay-based congestion control remains an active area of research and new algorithms continue to be developed. Recent examples include: Fast TCP [4], [5]; Microsoft Compound [6] (partially based on delay); more recent delay-based AIMD variants [7]- [9]; and this present work which is one outcome of a Cisco-funded project to investigate delay-based congestion control in harsh network environments. Traditionally, a number of arguments are usually put forward that question the use of delay in congestion control applications.…”