Fine-tuning distributed rate limiter: a multi-scenario performance evaluation

Abstract

This dissertation focuses on investigating and assessing different distributed system techniques, especially when it comes to rate-limiting resource protection. The Token Bucket Algorithm, Fixed Window Counter Algorithm, and Sliding Window Log Algorithm are a few of these algorithms. External libraries like Redis for quick value retrieval and Apache Zookeeper for synchronization are integrated into this study. To guarantee accurate data gathering, the benchmarking tests will mainly evaluate the latency, response time, and CPU utilization of each method in a containerized environment. The results show that the best way to deal with the difficulties of distributed rate restriction is to use the Sliding Window Log Algorithm enhanced with LuaScript. Moreover, the application of this rate limiter to a practical scenario yields anticipated outcomes: a decrease in acceptance rates accompanied by reductions in latency, response time, and CPU usage.