Efficient implementation of multi-level Dragonfly networks with Hamming graph for future optical networks

With the emergence of cloud computing and virtualized infrastructure in the datacenters, the use of high-radix routers is the most cost-effective alternative for interconnection networks. They are typically utilized as a part of datacenters for High-Performance Computing (HPC). Software Defined Networking (SDN) consolidates the benefits of datacenter virtualization, increasing resource flexibility and utilization and reducing infrastructure costs and overhead. Datacenter networks should be able to ensure high throughput and resiliency. For such reasons, Hamming graphs and Dragonfly networks are suitable for use with high-radix routers. Multilevel Dragonfly networks are used for lower-radix routers to increase the maximum achievable system size with the same router design. This paper introduces Hamming graphs and Dragonfly topologies based on SDN basis. It also presents a novel addressing scheme for Dragonfly topology with simulation experiments. The proposed model will be used for minimal, non-minimal, and adaptive routing in Dragonfly networks to extract a Python code using Mininet, which includes MiniEdit that is used to create and run network simulations. Evaluations show that with global trunking, systems are built with fewer groups than the maximum allowed. Therefore, there is no compelling reason for an additional cost. The proposed recommendations will be useful in the implementation of optical networks.