Skip to content
Book Open access

Faastion: Elastic and Scalable Native Library Isolation for High-Density Serverless Platforms

Nov 2026 · 0 citations · 63 references

Abstract

Serverless platforms increasingly co-locate concurrent function invocations to maximize elasticity and density. This co-location is supported by either software or hardware isolation techniques. While software isolation provided by existing sandboxes is flexible and lightweight, it does not uphold its security guarantees across native code. Conversely, hardware-based approaches, such as PKU-based sandboxing, enforce stronger boundaries for native code but are constrained by the limited number of available protection domains. We propose Faastion, a serverless runtime that combines Language Based Isolation (LBI) with a novel PKU domain multiplexing architecture for highly concurrent serverless environments. Faastion maintains the high concurrency, low latency, and low memory footprint of LBI while isolating native libraries. Compared to OpenWhisk, which enforces process-based isolation, Faastion exhibits a \(47\%\) lower memory footprint and eliminates cold starts, reducing 99th percentile end-to-end request latency by 6.9 × in a real-world trace execution. Furthermore, Faastion efficiently secures native libraries and prevents performance interference, increasing throughput by up to 3.7 × compared to alternative co-location-aware runtimes like Knative and Hydra.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.