INTOP: Improving Nginx Performance by Tuning OS Parameters

The performance of web servers is extremely important to the user experience of internet users. However, we find that the performance of Nginx, the most popular web server, under its default configuration, is far from optimal when the traffic load is heavy. This in turn makes Nginx can not satisfy users' requirements from time to time. To address this issue, we introduce INTOP, a novel approach to Improve Nginx's performance by Tuning OS Parameters (INTOP). First, INTOP excels at pinpointing the best multi-dimensional configuration parameters, ensuring that Nginx performs optimally under heavy loads. Second, INTOP determines the most influential parameter combinations and identifies the ideal values of them, streamlining the automatic tuning of OS parameters using the synergy of machine learning and genetic algorithms. To evaluate INTOP, we use two typical load test programs. The results show that INTOP can reduce the 99th percentile latency of Nginx with variant loads by 5.1 × on average and up to 36.2× compared to default configuration.