SchedBPF - Scheduling BPF programs
Year: 2025
Type: conference-paper
Abstract: The Linux BPF framework enables the execution of verified custom bytecode in the critical path of various Linux kernel routines, allowing for efficient in-kernel extensions. The safety properties and low execution overhead of BPF programs have led to advancements in kernel extension use-cases that can be broadly categorized into tracing, custom kernel policies, and application acceleration. However, BPF is fundamentally event-driven and lacks native support for periodic or continuous tasks such as background tracing, metric aggregation, or kernel housekeeping. Existing approaches such as kernel modules with kthreads, userspace daemons, or BPF timers fail to satisfy all the essential requirements for periodic kernel extensions such as fine-grained CPU control, kernel safety, and minimal overhead.
Authors: Kavya Shekar, Dan Williams
Institution: Virginia Tech
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Funder: Universitas Brawijaya
Award: G2355803430
OpenAlex ID: W4413311662