Journal of High Speed Networks

We investigate the performance of TCP under three representatives of packet scheduling algorithms at the router. Our main focus is to investigate how fair service can be provided for elastic applications sharing the link. Packet scheduling algorithms that we consider are FIFO (First In First Out), RED (Random Early Detection), and DRR (Deficit Round Robin). Through simulation and analysis results, we discuss the degree of achieved fairness in those scheduling algorithms. Furthermore, we propose a new algorithm which combines RED and DRR algorithms in order to prevent the unfairness property of the original DRR algorithm, which appears in some circumstances where we want to resolve the scalability problem of the DRR algorithm. In addition to TCP Reno version, we consider TCP Vegas to investigate its capability of providing the fairness. The results show that the principle of TCP Vegas conforms to DRR, but it cannot help improving the fairness among connections in FIFO and RED cases, which seems to be a substantial obstacle for the deployment of TCP Vegas.

In this paper, we first investigate the fairness between TCP Reno and TCP Vegas by focusing on the situation where Reno and Vegas connections share the bottleneck link. From the analysis and the simulation results, we find that the performance of TCP Vegas is much smaller than that of TCP Reno as opposed to an expectation on TCP Vegas. The RED algorithm improves the fairness to some degree, but there still be an inevitable trade‐off between fairness and throughput. Accordingly, we propose a ZL‐RED (Zombie Listed RED) algorithm, which enhances the RED algorithm to provide fair service for many flows at the bottleneck router. ZL‐RED uses the Zombie List, which has been originally proposed by SRED, to detect mis‐behaving flows which send packets at higher rate than others. Then, ZL‐RED sets higher packet discarding probabilities for those mis‐behaving flows. We evaluate an effectiveness of ZL‐RED by simulation experiments, and show that ZL‐RED can actually improve fairness among TCP connections.

The robustness of optical networks with respect to the accumulation of in‐band crosstalks was statistically studied in this paper, with emphases on the effect of network topology, identified by the network scale and physical connectivity. The in‐band crosstalks, a major and irreparable signal quality degrading factor, are induced by non‐ideal optical switches and demux/mux pairs at optical cross‐connect (OXC) in optical wavelength division multiplexing (WDM) networks. For the first time, our numerical simulations revealed that in a given network scale, there exists an optimum physical connectivity: the resultant overall accumulation of in‐band crosstalks can be minimized. Then the network scale was taken into account and the optimum network topology with respect to the accumulation of in‐band crosstalks was discussed for practical networks, with the limit of crosstalk ratio of the state‐of‐the‐art optical switches and demultiplexers. It was shown that proper network design can greatly relax the requirements posed on the optical switches and demultiplexers.

A computer network consists of a set of computing nodes interconnected via communication channels. It is commonly assumed that, for each pair of network nodes u and v, there is at most one channel from u to v. However, it is often desirable to have multiple channels between nodes. That is, for every pair of network nodes u and v, there may be multiple channels from u to v. In this paper, we consider the problem of providing deterministic quality of service guarantees when there are multiple channels between nodes. We show that any packet scheduling protocol that operates over a single channel can be modified to operate over multiple channels. In addition, this transformation increases the packet delay through the node by only a small amount. However, having multiple channels between nodes may cause packet reorder. This reorder significantly increases the upper bound on end‐to‐end delay. We show how this increase in delay is avoided through the use of efficient sorting techniques.