High-speed network demonstrated at SuperComputing 2018

March 1, 2019

Bill Fink presenting work of the SESDA HECN task at the SuperComputing 2018 (SC18) event in Dallas, TX in November 2018.

SESDA’s HECN team collaborated with various partners (Mid-Atlantic Crossroads (MAX), Naval Research Laboratory (NRL), StarLight optical network exchange facility, CenturyLink, SCinet) to create a multi-100-GigE network topology for live demos at SuperComputing 2018 (SC18) in Dallas, TX. This included 4×100-GigE network paths between NASA Goddard and SC18, 2×100-GigE network paths between StarLight and SC18, and 2×100-GigE network paths between NASA Goddard and StarLight. At SC18 they demonstrated the use of NVMe over Fabric (NVMe-oF) technology using internet Wide Area RDMA Protocol (iWARP) across a 4×100-GigE network infrastructure as a SCinet Network Research Experiment (NRE).
This showcased very high performance disk-to-disk network data transfers between a single pair of high performance NVMe servers, one at NASA Goddard and one at SC18, with a very low level of system CPU utilization and achieved an aggregate throughput of about 130 Gigabits per second (Gbps). They also performed more traditional network data transfers using the normal Linux TCP/IP network stack, which achieved an aggregate throughput of almost 200 Gbps while totally consuming all of the system CPU resources.

Comments are closed.