SC19 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

INDIS Keynote: The Relevance of Software Defined Exchange Points in the Path of Scientific Workflows

Workshop: INDIS Keynote: The Relevance of Software Defined Exchange Points in the Path of Scientific Workflows

Abstract: International research network connections are increasing, providing new paths for scientific workflows to move data and work across continents: In the north Atlantic between North America and Europe, the Advanced North Atlantic (ANA) collaboration operates 740G of distributed bandwidth capacity; between the U.S. and South America, the AmLight Express and Protect (AmLight-ExP) project added three 200G optical waves between Florida and Brazil, for a total capacity of 630Gbps; the AmLight-SACS project will be activating a 100G optical wave between Brazil and Angola, establishing a new south Atlantic route between the Americas, Africa and Europe; and Bella, a project led by RedCLARA and GEANT, will provide new high-capacity network paths between South America and Europe. While this phenomenon provides the opportunity for network operators to add resiliency to the global R&E fabric, it also adds complexity to network management and traffic engineering. Software Defined Exchange points (SDX) provide enhanced capabilities to simplify flow-management and enable new network services. Unlike traditional exchange points, an SDX controller has global visibility of the network topology, making it possible to maintain related states for the full path of a flow. This presentation will describe what an SDX is and why SDXs are relevant to scientific workflows in the global R&E network fabric. The AtlanticWave-SDX project and its SDX controller will be presented as a novel distributed programmable controller along with use cases for two science drivers. Results from extensive experiments in a testbed will be presented that demonstrate the AtlanticWave-SDX controller’s high performance in terms of scalability, response time, and resource utilization, which are extremely important to support advanced scientific applications. Finally, future work will be described to clarify where SDXs can have an impact on the global R&E network fabric.

Back to Innovating the Network for Data Intensive Science (INDIS) Archive Listing

Back to Full Workshop Archive Listing