Data Transfer in a Science DMZ using SDN with Applications for Precision Medicine in Cloud and High-performance Computing

As the difficulty and cost of generating data has decreased, there have been emerging problems of data movement between where it was collected to where it can most effectively be processed. The Science DMZ model is an established solution to help maximize the throughput from any given network infrastructure between sites but there still exists room for exploration and improvement, including the use of Software-Defined Networking (SDN). We aim to explore how the use of SDN-enabled flows in a Science DMZ may impact its main trait: performance. In this paper we have successfully demonstrated improvements in data transfer between certain sites and under certain conditions. Though we have also seen sites where there is no difference or even a performance penalty, all the results are consistent and reproducible over hundreds of test transfers suggesting there are unique properties of each site configuration leading to real and sustained performance increases to a Science DMZ.