Distance Based Server Selection in 5G Networks

Abstract

Distance aware selection of servers and other net-worked resources is necessary to take full advantage of the 5G ecosystem, where the new radio and edge computing both contribute to significantly reduced latencies, but where propagation times to distant servers may add noticeable delays. The literature on server selection, and in particular on distance aware server selection, is, however, surprisingly thin. The few approaches we have come across are either relatively complex, based on continuous measurements, or both and not reflected in current standards. The present work therefore aims at simple but accurate mechanisms which build on and are backward compatible with present standards. We briefly describe the existing alternatives for distance aware selection as reflected in the standards, and propose include location information in server URLs such that distances can be explicitly calculated and evaluated by selecting entities. Moreover, we present three distinct and simple approaches to encode locations, two of which advantageously can be combined. The performance of the methods are then subject to a comprehensive performance evaluation where a large number of varied scenarios are considered. The results reveal that simple geographical coordinates should be restricted to very dense networks (typically networks which form full meshes) whereas a combination of the other approaches work well in other networks (in our case networks based on trees, stars or loops). Contrary to earlier proposals, our proposals consistently provide upper bounds on distances which eliminates the uncertainties that surrounds approximate solutions.