Article Preview
Top1. Introduction
With the rise of wireless networking, performance expectations have steadily increased as technology has consistently improved. Similarly, internet applications that depend on streaming timeliness centric data in the midst of congestion, e.g., streaming movies, internet radio, online gaming, etc., have taken the forefront in pushing the limits of wireless and other networking technologies. Congestion based limitations can result from the finite bandwidth and erratic loss due to the characteristics of the physical medium. The protocol layer that deals with these limitations and transporting data in the midst of these challenges is the transport layer.
The most popular protocol in this layer, TCP, has been thoroughly investigated in this context. As a result, several new protocols have been developed and standardized to implement new strategies and features for dealing with the changing demands of these internet applications. The protocols considered in this article include DCCP and SCTP which approach this task in different ways, but which both attempt to provide the necessary functionality to facilitate further growth in performance.
DCCP and SCTP
The Datagram Congestion Control Protocol (DCCP) was designed for applications that need connection based congestion control without reliability (Kohler et al., 2006). The most significant feature of DCCP is its modular configurability of the congestion control algorithm. Depending on the network environment and the requirements of the application, it can handle congestion differently depending on which standardized congestion control algorithm it uses. These algorithms are denoted by their Congestion Control Identifiers (CCID). The CCID’s standardized so far include CCID2, which is designed to behave TCP-like with several of the same mechanisms as TCP, and CCID3, which is designed to be TCP-friendly but to control congestion based on an equation so that traffic is affected less rapidly but more smoothly (Kohler et al., 2006). The Stream Control Transport Protocol (SCTP) has similar congestion control mechanisms as TCP including the SACK extension, slow start, congestion avoidance, and fast retransmit (Ye et al., 2002). Though the congestion control mechanisms of these protocols are all related to TCP and focus on interaction with TCP traffic, they each vary in specific ways that can affect how they interact with wireless congestion. It is these variations that will provide different behaviors and performance in different congestion environments.
Several authors have investigated the individual performance of the transport layer protocols in question. Holland and Vaidya (1999) focused on fairness, a congestion control metric considered in this article. The fairness of TCP in particular was proven to be lacking in wireless networks. Takeuchi et al. (2005), however, points out DCCP’s own limitations in fairness when sharing a bottleneck with TCP. Simulation has also been performed using DCCP focusing on throughput smoothness, another metric adopted by this article, where DCCP was shown to be strong in this arena by Navaratnam et al. (2006). Rhee and Xu (2007) analyzes the TFRC congestion control algorithm to determine its limitations. They then verify their findings using ns2 simulations. Active wireless experimentation has been performed with TCP (Bruno et al., 2007) and DCCP (De Sales et al., 2008) where these transport protocols are evaluated in wireless networks similarly to the methodology adopted in this article. SCTP has also been experimented with using heterogeneous network environments using a network evaluation tool designed for this purpose by Emma et al. (2006).