Congestion control for RDMA

Fast congestion control for RDMA-based datacenters

This project addresses the challenge of achieving extreme low latency in datacenter networks for Online Big Data (OLBD) applications, which are critical workloads in datacenter computing. Remote Direct Memory Access (RDMA), which is a promising alternative to traditional TCP, significantly reduces unloaded network latencies by about an order-of-magnitude. However, in practice, existing congestion control schemes for RDMA require several tens of round trips to converge to their optimal rates with typical workloads. Our project aims to close this gap between unloaded and loaded network latencies. We collaborate with researchers from Purdue University in this NSF-funded project.

SIGCOM 2018 Fast Congestion Control in RDMA-based Datacenter Networks