Congestion Control Researches Over Heterogeneous Networks

During the last decade years, with the globalization of the telecommunication industry and the development of the Internet, new classes of hosts such as mobile devices are gaining popularity, while the transmission media becomes more heterogeneous. In order to provide end-to-end communication services for mobile hosts, wireless networks are quickly becoming an interior part of the Internet, and the reliable protocols such as TCP must be supported over them. So challenges are met in providing reliable transport services to all hosts regardless of the type of network connectivity used. Over the past couple of years, TCP performance over wireless links has been an active area of research and the researchers have been heavily involved in coordinating transport control research on heterogeneous network.TCP is the prevalent reliable transport protocol and has an important role in the success of the Internet today. Although TCP has served well the wired Internet for almost 20 years, it is not ready for the wired-cum-wireless environments at present. TCP has been designed and turned to perform well under the assumption that all losses are an indication of congestion. However, when a TCP connection traverses a wired-cum-wireless link, packets may be lost due to wireless transmission errors, user mobility and channel asymmetry. In addition to congestion losses, TCP implicitly assumes that all packet losses are due to congestion, and triggers congestion control mechanism as a packet loss is detected. It has been previously demonstrated that this causes severe performance degradation when non-congestion loss happens. To avoid the performance degradation, researchers have been questing to distinguish different packet loss behaviors, and it is becoming a topic of intense research.In chapter one, a general introduction to the TCP congestion control mechanism is provided and the limitation of the traditional TCP/IP is studied with respect to a wireless network. After analyzing the issues essentially, we propose the existing problems in current TCP congestion control.Researchers have proposed various modified TCP congestion control algorithms to eliminate the drawbacks of Reno in a heterogeneous network. In Chapter two, recently proposed solutions are surveyed along with their main advantages and disadvantages. After an exhaustive analysis of the key problems within TCP/IP, we raise the corresponding solutions in this field . Finally, some constructive solutions arp given for further research.Based on the problem of the performance degradation of TCP in the presence of bandwidth asymmetry, we propose a novel algorithm and evaluate it in Chapter three, which is called ESACK. The main contribution of ESACK is as the following: Firstly, with the limited 40 bytes available for the TCP options, it informs the sender more information about the incoming packets than SACK does within-one byte. Secondly, with the significant two bits of "ST", each octet of the ESACK option outlines the state of the data packets that received by then, informing the sender of the packet losses explicitly. Finally, an "Asym Ratio" option can be piggybacked on the packets sent by the source, which carries out a "dynamic" ACK congestion control scheme on reverse path. The results of simulation show that the ESACK has better advantages in asymmetric environment and does well in solving the problems raised from burst losses. Compared with SACK algorithm, the ESACK has utilized available bandwidth of the network effectively, while kept the fair quality of service and less buffer memory.In order to improve the TCP performance in wireless environments, we put forward a window-based explicit loss notification scheme in chapter four. It is implemented in the cooperation of multiple network protocol layers and makes full use of the advantages of them. At first, in contrast with the retransmission mechanism in data link layer, the routes in this scheme only pass the header of the corrupted packet to the receiver, which does not interfere with the TCP congestion...