Per-packet load-balancing approaches are increasingly deployed in datacenter networks. However, their combination with existing congestion control algorithms (CCAs) may lead to poor performance and even throughput collapse.
In this talk, I first model the throughput collapse of a wide array of CCAs when some of the paths are congested. I explain that since CCAs are typically designed for single-path routing, their estimation function focuses on the latest feedback and mishandles feedback that reflects multiple paths. I propose to use a median feedback that is more robust to the varying signals that come with multiple paths. I introduce MSwift, which applies this principle to make Google’s Swift CCA robust to multi-path routing while keeping its incast tolerance and single-path performance. Finally, I demonstrate that MSwift improves the 99th-percentile FCT by up to 25%, both with random packet spraying and adaptive routing.
M.Sc. student under the supervision of Prof. Isaac Keslassy and Prof. Mark Silberstein.
