In mobile ad hoc networks, researches have been concentrated on the implementations and performance comparisons of routing protocols. Little attention was paid to the performance of the end-to-end TCP connection. We study the route breakages caused by mobility and their effects on the performance in different TCP implementations (e.g. Tahoe, Reno, NewReno and Sack). Our performance simulation shows that current TCP implementations can not guarantee a stable performance because they are impossible to distinguish a packet loss caused by congestion or by route breakage. A new metric, weighted route length is introduced in the comparison of TCP performance in the mobile ad hoc network. We demonstrate that TCP Sack generally performs better than Reno, NewReno and Tahoe. The effects of interactions among routing, MAC and TCP protocol are also discussed. This paper presents the simulation results of four TCP implementations and provides comparisons based on some important performance metrics (throughput, goodput, transfer time) and multimedia QoS measurement (average packet delay).
[1]
Srinivasan Seshan,et al.
A comparison of mechanisms for improving TCP performance over wireless links
,
1996,
SIGCOMM 1996.
[2]
Nitin H. Vaidya,et al.
Analysis of TCP Performance over Mobile Ad Hoc Networks
,
1999,
Wirel. Networks.
[3]
Bartosz Mielczarek,et al.
Scenario-based performance analysis of routing protocols for mobile ad-hoc networks
,
1999,
MobiCom.
[4]
Charles E. Perkins,et al.
Ad-hoc on-demand distance vector routing
,
1999,
Proceedings WMCSA'99. Second IEEE Workshop on Mobile Computing Systems and Applications.
[5]
David A. Maltz,et al.
A performance comparison of multi-hop wireless ad hoc network routing protocols
,
1998,
MobiCom '98.
[6]
Theodore S. Rappaport,et al.
Wireless communications - principles and practice
,
1996
.