What question did this study set out to answer?

To develop a predictive system for managing resource demands in web services before overload occurs.

April 16, 2026Open Access

PFAC: Pre-Flight Admission Control System for High-Throughput Distributed Services

Key Points

To develop a predictive system for managing resource demands in web services before overload occurs.
Formalized a cost model for admission control
Implemented PFAC using Redis and Lua for capacity reservation
Conducted performance tests using k6 with various admission modes and controllers
PFAC showed improved resource allocation compared to traditional methods
Demonstrated effectiveness in high-throughput scenarios
Provided reproducible benchmarks within CI environments

Abstract

This document specifies PFAC (Pre-Flight Admission Control), a predictive, cost-aware admission layer for Rust web services. Rather than reacting to overload after resources have already been consumed, PFAC estimates per-request resource demand before handler execution, atomically reserves capacity in a shared store (Redis via Lua), and adapts an effective budget through a health-scaled safety factor. The specification formalizes the cost model, relates PFAC to classical and modern admission/overload work (RED, CoDel, SEDA, Google SRE adaptive throttling, Breakwater, Prequal), and provides implementation sketches for Tower/Axum, Redis/Lua, and batched MySQL writes. Twelve CI-reproducible k6 scenarios on GitHub Actions runners record end-to-end behavior against a Rust/Axum server with three admission modes (PFAC, plain token bucket, and disabled) and two controllers (three-band linear and PI). This repository contains the LaTeX source for the paper (pfac-spec.tex), the benchmark server with Redis-backed admission control, the device API implementation and examples, and the k6 load test scripts.

PFAC: Pre-Flight Admission Control System for High-Throughput Distributed Services

Key Points

Abstract

Cite This Study