An Adaptable Optimal Network Topology Model for Efficient Data Centre Design in Storage Area Networks

Abstract

In this research, we look at how different network topologies affect the energy consumption of modular data centre (DC) setups. We use a combined-input directed approach to assess the benefits of rack-scale and pod-scale fragmentation across a variety of electrical, optoelectronic, and composite network architectures in comparison to a conventional DC. When the optical transport architecture is implemented and the appropriate resource components are distributed, the findings reveal fragmentation at the layer level is adequate, even compared to a pod-scale DC. Composable DCs can operate at peak efficiency because of the optical network topology. Logical separation of conventional DC servers across an optical network architecture is also investigated in this article. When compared to physical decentralisation at the rack size, logical decomposition of data centers inside each rack offers a small decrease in the overall DC energy usage thanks to better resource needs allocation. This allows for a flexible, composable architecture that can accommodate performance based in-memory applications. Moreover, we look at the state of fundamentalmodel and its use in both static and dynamic data centres. According to our findings, typical DCs become more energy efficient when workload modularity increases, although excessive resource use still exists. By enabling optimal resource use and energy savings, disaggregation and micro-services were able to reduce the typical DC's up to 30%. Furthermore, we offer a heuristic to duplicate the Mixed integer model's output trends for energy-efficient allocation of caseloads in modularized DCs.