Software-defined Wide-area Networks in the Coevolution of Network and Computational Bandwidth

Abstract

Software-Defined Wide-Area Networks (SD-WAN) are a technology that utilizes programmatic configuration to maximize network throughput. The main problem to achieve efficient use of network bandwidth with SD-WAN is the network delay (latency) in the communication. A network could be further slowed down by not enough memory or CPU resources. Here we conduct experiments, investigating the capability of an SD-WAN to maximize the network throughput at increasing levels of latency. We find that to a limited extent the computational optimization at the software side could offset the network limitations. At higher network latency the SD-WAN adds little value. Further increase of the capacity can only be achieved by hardware upgrades of the network bandwidth. Such solutions may be considered drivers of a coevolutional relationship between computational and network bandwidths, whereby the network use becomes more efficient.