The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Abstract: With the advent of new applications such as Cloud Computing, Big Data, and Machine Learning, modern data center network (DCN) architecture has been evolving to meet numerous challenging requirements such as scalability, agility, energy efficiency and high performance. The above new applications are expediting the convergence of High-Performance Computing and Data Centers. Inspired by hypermeshes, this paper presents HyperOXN, a novel cost-efficient topology for exascale Data Center Networks (DCNs). HyperOXN takes advantage of high-radix switch components leveraging state-of-the-art colorless dense wavelength division multiplexing (DWDM) technologies, effectively supports -cast traffic and at the same time meets the demands on high throughput and low latency. It’s shown that HyperOXN outperforms a comparable Fat-Tree topology in cost, throughput and power consumption under a variety of workload conditions.