The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Workshop:INDIS'21: 8th Workshop on Innovating the Network for Data-Intensive Science
Authors: Brian Tierney, Eli Dart, and Ezra Kissel (Energy Sciences Network (ESnet)) and Eashan Adhikarla (Lehigh University)
Abstract: It is well known that loss-based TCP congestion control algorithms are problematic for high-speed high-latency flows that are common in Big Science. In 2016 Google released a new congestion control algorithm called ’BBR’ (Bottleneck Bandwidth and Round-trip time) that uses a model-based approach, and the design has since been refined in an alpha release of BBRv2. In this paper, we describe and perform a set of experiments that assess the suitability of BBRv2 for use on Data Transfer Nodes (DTNs). The experiments were run on both production R&E networks as well as within a controlled testbed environment. Our analysis of the results show that BBRv2 improves upon BBRv1for common Big Science transfer scenarios and is a promising option in high-speed short-queue networking environments.
Back to INDIS'21: 8th Workshop on Innovating the Network for Data-Intensive Science Archive Listing