A lightweight, per-queue Quality of Service (QoS) IO Scheduler for cloud environments and multi-tenant systems running Linux. The latency-sensitive applications on these systems suffer degraded performance (99th percentile latencies increasing by upwards of 500%) when competing with throughput-intensive workloads, such as backups, batch processing, and archiving.
A machine using this scheduler will be able to manually fine-tune the ratio of IO requests between latency-sensitive and through-put intensive applications, as well as dynamically turn on and off the scheduler depending on the system's needs. Additionally, the CPU usage will only increase by around 10-15% on average. As a drawback, however, there will be notably degraded performance for the throughput-intensive applications and workloads as they give up their IO slots.
See CONTRIBUTING.md for setup, build instructions, and
contribution guidelines.
@BrandonPacewic, @phannawich, @TSrirama2026, @Benjamin-Anderson-II, @CameronDilworth
Project Partner: @godhanipayal
When encountering an issue or requesting a feature please make a post in the discussion tab of this repository.
This project is licensed under GPL-2.0 with the Linux-syscall-note exception,
the same license as the Linux kernel. See COPYING for details.
All contributions are subject to this license.