BPFComposer: Towards deterministic and safe composition of eBPF Programs (and their Shared state)

Read full text
Authors: Prankur Gupta, Takshak Chahande, Miao Xu, Prashanth Kannan, Tejas Birajdar, Lilian Liu, Rajiv Krishnamurthy, Theophilus Benson
Cited by: 1
References: 1

Abstract

Deploying eBPF at scale requires programs from independent teams to coexist safely 1 and share state efficiently, yet the kernel provides almost no primitives to support this. The result is silent functionality loss, network degradation, and fleet-wide outages often undetected during isolated testing. We present BPFComposer, which helps address this by interposing on the kernel at two levels: a managed program chainer layer enforcing deterministic execution order and return-value isolation, and a shared map registry providing versioned, provider-consumer coordination. Deployed across millions of hosts, BPFComposer has eliminated over 400GB of redundant map memory fleet-wide and prevented recurrence of the failure classes that motivated its design.

Fields of Study & Subjects

Advanced Software Engineering Methodologies Model-Driven Software Engineering Techniques Software Engineering Research

Publication details

Journal / source
Proceedings of the 20th ACM International Conference on Distributed and Event-based Systems
Publisher
ACM
Pages
150–156
Publication date
22 June 2026