Modeling and Analysis to Estimate the EndSystem Performance Bottleneck Rate for High-Speed Data Transfer

Abstract — The bandwidth available in a high-speed backbone network today is much greater than the capacity at the edge. As an example, when the transmission line rate is 10 Gbps, the end computing machine (end-system) may not be able to keep up with the arriving data rate and is hence often a bottleneck for data transfer. A key deficiency in most transport protocols is the lack of an accurate estimate of this bottleneck rate. In the absence of the same, the sender often overshoots the receiver’s capacity resulting in packet losses. When the sender receives feedback about the packet losses, it throttles the sending rate. Repeat occurrences of the same limits the performance of endto-end data transfer. We propose a mechanism to determine the end-system bottleneck rate. We employ a Stochastic Reward Net (SRN) queuing network model which captures different system components such as the processor, disk, etc., as well as the current workload in terms of the number of CPU-bound and I/O-bound tasks. The sending rate which would yield the best performance for data transfer to the end-system is estimated from the above model. We define this rate as the effective bottleneck rate. Finally, we validate our analytical model on an experimental testbed and demonstrate that matching the sending rate with the effective bottleneck rate leads to improved performance of end-to-end data transfer. I