25-08-2017, 09:32 PM
This article is presented by:
Mohammad Nayeem Teli
Endrit Thanasi
Colorado State University
Analysis of Datagram Congestion Control Protocol
ABSTRACT
In the last decade, there has been an enormous increase in the need to stream line data. Some of the primary applications like Voice over IP, Video Conferencing, music, and video streaming etc, need continuous flow, and transmission in real time, and consume a large amount of bandwidth. Data streaming usually employs one of the most common transport layer protocols, either UDP1, or TCP2, primarily UDP. DCCP3 is a new transport protocol in the UDP/TCP family that provides an unreliable transmission with congestion control meant for applications that prefer timeliness to reliability[2]. TFRC4 for VoIP is one of the choices of congestion control mechanisms provided by DCCP when used to carry voice[5]. In this paper we are aiming to improve TFRC variant for VoIP congestion control algorithm by reducing the variation of transmission rate, hence providing VoIP sessions with better quality.
INTRODUCTION
UDP has been the favorite protocol for applications that are sensitive to delay, but it lacks congestioncontrol. UDP chokes TCP traffic since TCP responds to congestion and varies its dataflow while UDP is unresponsive. The problem with TCP in carrying such data is delay, due to retransmissions. A better transport layer protocol needs to assimilate the best of both TCP and UDP , inorder to provide a viable solution. DCCP has been designed to incorporate the best features of both UDP and TCP, in addition to selective congestion control algorithms. These selective algorithms can be chosen to provide better performance based on the requirements of the data being transmitted. For exclusive transmission of voice packets, DCCP has a congestion control mechanism called TFRC variant for VoIP. However, there is a grey area. Whenever there is a silent or a pause moment in a conversation, DCCP is designed to drop to a slow start rate, which affects the quality of voice, bandwidth utilization, and the equilibrium rate. To counter these problems the authors of DCCP introduced a faster restart technique, in which the rate quadruples before it reaches the pre-pause sending rate. However, this technique does not help the sending rate to either reach an equilibrium point or to stay at equilibrium for long once it is achieved. The reason being, in a typical phone conversation, 35-47% of the time is a pause or a mutual silence.
For more information about this article,please follow the link:
http://www.cs.colostate.edu/~nayeem/papers/dccp.pdf