The performance of TCP/IP for networks with high bandwidth-delay products and random loss
IEEE/ACM Transactions on Networking (TON)
Modeling TCP Reno performance: a simple model and its empirical validation
IEEE/ACM Transactions on Networking (TON)
Fluid-based analysis of a network of AQM routers supporting TCP flows with an application to RED
Proceedings of the conference on Applications, Technologies, Architectures, and Protocols for Computer Communication
TCP is max-plus linear and what it tells us on its throughput
Proceedings of the conference on Applications, Technologies, Architectures, and Protocols for Computer Communication
Efficient solution of multiple server queues with application to the modeling of ATM concentrators
INFOCOM'96 Proceedings of the Fifteenth annual joint conference of the IEEE computer and communications societies conference on The conference on computer communications - Volume 1
Hi-index | 0.00 |
The purpose of this paper is to introduce an accurate performance model for the TCP Reno version in the presence of a bottlenecked router which uses the Random Early Detection (RED) active queue The paper makes two contributions:• It introduces an accurate model of a target source operating according to the TCP Reno mechanism in the presence of background traffic. The background traffic is represented by a general discrete batch Markov arrival process (D-BMAP), which is modified in order to make the phase transitions dependent on packet losses. It can therefore be used to model a collection of UDP and/or TCP sources. Under this model, packets are dropped only when the router is congested, or when the RED protocol is invoked, i.e., the buffer occupancy is taken into account.• The paper also introduces an accurate model of the RED mechanism, which tracks the evolution of the difference between the instantaneous and average queue sizes. This representation is chosen since the average queue size tends to follow the instantaneous queue size, and therefore the difference between them is usually limited.This model is different from the models presented in the literature for TCP in a number of ways. Unlike [1, 2] where packet losses are random, and independent of the actual buffer occupancy, our model captures the buffer occupancy, and the actual packet losses due to buffer overflow. This paper also models the cross traffic using a general process. Reference [3] considered the effect of cross traffic only by modeling the service times using a random process.