Analytical Study of Pre-congestion notification (PCN) techniques

TL;DR

This paper analyzes various PCN algorithms to determine their effectiveness under different network congestion scenarios, providing recommendations and a benchmark for selecting suitable algorithms.

Contribution

It investigates the analytical behavior of known PCN algorithms, modifies them for specific scenarios, and develops a benchmark to aid in choosing the right algorithm based on network conditions.

Findings

Different PCN algorithms perform optimally under specific congestion levels.

Simulation results guide the selection of algorithms for realistic network scenarios.

A benchmark tool is proposed to assist researchers in algorithm selection.

Abstract

Maintaining the quality of service (QOS) and controlling the network congestion are quite complicated tasks. They cause degrading the performance of the network, and disturbing the continuous communication process. To overcome these issues, one step towards this dilemma has been taken in form of Pre-congestion notification (PCN) technique. PCN uses a packet marking technique within a PCN domain over IP networks. It is notified by egress node that works as guard at entry point of network. Egress node gives feedback to communicating servers whether rate on the link is exceeded than configured admissible threshold or within the limit. Based on this feedback, admission decisions are taken to determine whether to allow/block new coming flows or terminate already accepted. The actual question is about selection of right algorithm for PCN domain. In this paper, we investigate the analytical behavior of some known PCN algorithms. We make slide modifications in originality of PCN algorithms without disquieting working process in order to employ those within similar types of scenarios. Our goal is to simulate them either in highly congested or less congested realistic scenarios. On the basis of simulation done in ns2, we are able to recommend each PCN algorithm for specific conditions. Finally, we develop a benchmark that helps researchers and scientific communities to pick the right algorithm. Furthermore, the benchmark is designed to achieve specific objectives according to the users' requirements without congesting the network.