A High Performance Design, Implementation, Deployment, and Evaluation of The Slim Fly Network
Novel low-diameter network topologies such as Slim Fly (SF) offer significant cost and power advantages over the established Fat Tree, Clos, or Dragonfly. To spearhead the adoption of low-diameter networks, we design, implement, deploy, and evaluate the first real-world SF installation. We focus on deployment, management, and operational aspects of a cluster with 200 servers deployed at CSCS, and carefully analyze performance. We demonstrate techniques for simple cabling and cabling validation as well as a novel high-performance routing architecture. Our real-world benchmarks show SF’s strong performance for many modern workloads such as deep neural network training, graph analytics, or linear algebra kernels, as well as cost/performance advantages over non-blocking Fat Trees. We also overview our current ongoing and future projects and illustrate the relevance of a low-diameter cluster for their success and impact.