Design of efficient packet marking-based congestion management techniques for cluster interconnects

Resumen

The growth of parallel computers based on high-performance networks has increased the interest and effort of the research community in developing new techniques to achieve the maximum performance from these networks. In particular, the development of new techniques for efficient routing to reduce packet latency and increase network throughput. However, high utilization rates of the network could result in what is known as network congestion, which could cause a degradation of the network performance because all, or a part of the network has exceeded the maximum utilization, which is imposed by the saturation point of the network. Congestion management in multistage interconnection networks is a serious problem not completely solved. In order to avoid the degradation of network performance when congestion appears, several congestion management mechanisms have been proposed. Most of these mechanisms are based on explicit congestion notification. For this purpose, switches detect congestion and depending on the applied strategy, packets are marked to warn the source hosts. In response, source hosts apply some corrective actions to adjust their packet injection rate. Although these proposals seem quite effective, they either exhibit some drawbacks or are partial solutions. Some of them introduce some penalties over the flows not responsible for congestion, whereas others can cope only with congestion situations that last for a short period of time. The aim of this dissertation is to analyze the different strategies to detect and correct congestion in multistage interconnection networks and propose new congestion management mechanisms targeted to this kind of lossless networks. The new approaches will be based on a more refined packet marking strategy combined with a air set of corrective actions in order to make the mechanisms capable of effectively managing congestion regardless of the congestion degree and traffic conditions.