Abstract
TCP Hybla is increasingly being recognized for yielding high throughput over satellite IP links. However, due to the difficulty in measuring the expected performance over such links, the deployment decision is not trivial. A good analytical model of TCP Hybla can help spread its deployment by allowing easy estimation of its performance using network parameters like packet loss probability and round trip time (RTT) over the target satellite connection. Also, it can lead to design of congestion control mechanisms based on Hybla friendly rate control. In this paper, we therefore propose new analytical models for (i) steady-state throughput and (ii) latency for TCP Hybla over satellite IP links. The new model for steady-state throughput results from a major overhaul of our previously devised model for the same. The latency model has been devised newly. Using the latency model, the throughput and the average transmission time can be estimated. We verify the accuracy of our model by comparing with simulation and emulation results. It is observed that for packet loss probability of 10−4 or below that correspond to real life packet loss probability over satellite link, our analytical models for throughput and latency closely match both the simulation and emulation outcomes. That is, our analytical models correctly represent the behavior of satellite links. Our new model leads to up to 3% more accuracy than our previously proposed model while calculating steady-state throughput.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have