Tuning Kernel Network Parameters With Reinforcement Learning

The Linux kernel has multiple parameters that can be used to fine-tune network performance. Given the number of parameters and the multiple values each one can take makes deciding the optimal values a non-trivial search problem. The goal of the project is to use more advanced search techniques from discrete optimization and reinforcement learning to guide the search for optimal parameters in an efficient manner.

Tracked here in Taiga

Ulrich Drepper

Team: Office of the CTO
Location: Remote DE
Boston Metro RIG