Unfair by design: eBPF-based scheduling of mixed database workloads

2026-05-04

AI Summary

This work addresses the challenge of interference from background tasks on latency-sensitive workloads in hybrid database environments, a problem inadequately mitigated by existing operating system schedulers due to ineffective isolation and susceptibility to priority inversion. The paper proposes UFS, a selective unfair scheduler built upon eBPF and the sched_ext framework, which introduces application-aware scheduling into the kernel for the first time. UFS dynamically restricts background tasks to idle CPU cycles via eBPF and enables immediate preemption upon the arrival of high-priority tasks. It further incorporates an application-level hinting mechanism to alleviate priority inversion. Evaluated under PostgreSQL mixed workloads, UFS achieves up to a 2× improvement in throughput for latency-sensitive tasks and reduces tail latency by 50% compared to the default Linux scheduler.

Abstract

Modern database systems increasingly co-schedule time-sensitive and background tasks. In such mixed workloads, background tasks should ideally utilize only spare CPU capacity without interfering with latency-critical requests. While some database-level solutions address this challenge, many database systems still rely on operating system (OS) schedulers, which, despite supporting priorities, do not reliably isolate high-priority tasks. Furthermore, they remain vulnerable to priority inversion, where preempted background tasks can delay other work. We present UFS, a selectively unfair scheduler implemented as an eBPF-based sched_ext scheduler in the Linux kernel. UFS restricts background tasks to idle CPU capacity and preempts them immediately when time-sensitive tasks arrive. To address priority inversion, UFS incorporates application-level hints via eBPF maps, ensuring that background tasks are not unnecessarily delayed should time-sensitive tasks wait for them to release locks. Our integration of UFS into PostgreSQL demonstrates that, under mixed workloads, UFS improves throughput for time-sensitive tasks by up to 2X, while reducing tail latency by half, compared to existing scheduling options in Linux.

Problem

• mixed workloads
• priority inversion
• CPU scheduling
• latency-critical tasks
• background tasks

Innovation

• eBPF
• scheduler
• priority inversion
• mixed workloads
• database