Answers are grounded in this paper and cite their sources.
With the widespread deployment of cloud services, data center networks are developing toward large-scale, multi- path networks. To support the new trend, some novel network architectures have been proposed, and Fat-tree is one of the most promising architecture and gets a lot of attention because it has good performance on aggregate bandwidth with a simple topology. This paper presents Diamond, an improved Fat-tree architecture for large-scale data centers. Diamond replaces all the aggregation switches of Fat-tree with edge switches and connects directly edge switches to core switches. By this alteration, the average route path and End-to-End (ETE) delay in Diamond are 10% less than that in Fat-tree. We design FAR, a simple and high-efficient routing method for Diamond. In FAR, each switch requires only hundreds of route entries to support a large-scale network with tens of thousands servers, and to build FAR routing tables, switches exchange very few messages. Diamond and its FAR routing are verified through OPNET simulations.
No takes yet. Share an insight, caveat, or question.
Sun et al. (2014) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: