Congestion avoidance and control
SIGCOMM '88 Symposium proceedings on Communications architectures and protocols
TCP/IP illustrated (vol. 2): the implementation
TCP/IP illustrated (vol. 2): the implementation
Comparative performance analysis of versions of TCP in a local network with a lossy link
IEEE/ACM Transactions on Networking (TON)
Modeling TCP Reno performance: a simple model and its empirical validation
IEEE/ACM Transactions on Networking (TON)
Comparative study of various TCP versions over a wireless link with correlated losses
IEEE/ACM Transactions on Networking (TON)
Throughput analysis of TCP on channels with memory
IEEE Journal on Selected Areas in Communications
Web traffic modeling exploiting TCP connections' temporal clustering through HTML-REDUCE
IEEE Network: The Magazine of Global Internetworking
MSWiM '05 Proceedings of the 8th ACM international symposium on Modeling, analysis and simulation of wireless and mobile systems
Data-unit-size distribution model when message segmentations occur
Performance Evaluation
| Hi-index | 0.01 |
In this paper, we investigate the effect of the retransmitted packet size preservation (RPSP) property on TCP performance over links with Bernoulli bit-errors, where RPSP-means that the sizes of all packets transferred over datalinks (referred to as transferred packets) with the same sequence-number at retransmissions are equal to the size of a packet generated from a message at the original transmission (namely, generated packet). To achieve this, we present a Markov model for size-sequences of the generated and transferred packets, in which the TCP message-segmentation and error-recovery functions with RPSP are taken into account, respectively. Furthermore, we derive the round-trip time and the transferred packet corruption probability, which are the main parameters of TCP performance. Applying them to the conventional TCP-Reno send-rate formula, we discuss the effect of RPSP on TCP-Reno goodput. The key findings from the numerical results include that RPSP cannot be negligible for networks that suffer from a high bit-error rate (e.g., 1 脳 10^-4, which is the mean bit-error rate of a wireless link in an industrial environment), when the message-segmentation occurrence probability is relatively small.