Fengyuan Ren

The Linux kernel is permeated with constant values that are critical to system performance. Many of these constants, referred to as _perf-consts_, are magic numbers with brittle assumptions on hardware and workloads. Unfortunately, there is no capability of _in-situ_ tuning of perf-const values o...

Fast-evolving machine learning (ML) workloads have increasing requirements for networking. However, host network transport on RDMA NICs is hard to evolve, causing problems for ML workloads. For example, single-path RDMA traffic is prone to flow collisions that severely degrade collective communic...