Reproduction and Extension of CASR: A Cache-Based Adaptive Scheduler for Serverless Computing with Novel K=4 Queue Granularity Analysis

This paper presents an independent reproduction and experimental extension of CASR (Cache-Based Adaptive Scheduler for Serverless Runtime), originally proposed by Chen et al. in Future Generation Computer Systems 2025. We implement CASR from scratch in Python including W-TinyLFU caching and PPO reinforcement learning agent.We evaluate on the Microsoft Azure Functions 2019 dataset containing 1,332,032 daily invocations against five baseline algorithms across three workload types. Key findings:1. CASR eliminates wasted memory time across all evaluated workloads2. CASR reduces cold start rate by 3.848 to 14.929 percentage points compared to FaaSCache3. Novel K=4 experiment reduces cold start rate by up to 5.900 percentage points over original K=3 design All code available at:https://github.com/Krishn4nmol/CASR_Project