A Solution for Instant Response of Cloud Platform Based on Nginx + Keepalived

Abstract

How to fulfill load balancing in high concurrency status and thereby to achieve instant response of cloud platform is always a focus of attention to the industry. This document presents an instant response solution for cloud platform based on Nginx + Keepalived, which, by means of Nginx, enables selection of proper load balancing algorithm based on the scenario requested by client, and detection of server working status by using Keepalived to solve single point of failure. This solution is implemented fully with open-source software, and will allow high-concurrency requests of tens of thousand per second ensure good user experience and achieve the goal of low cost and high availability, making it particularly suitable for small-/medium-sized cloud platforms with economic applicability requirements. Test results show that, with respect to analysis on 50% or 80% of the requests, the average response time for 10,000 concurrent requests is only 101 ms and 152 ms, with memory usage up to 22% and CPU peak power up to 80%, which meets the expected response time shorter than 3 seconds in high-concurrency scenario of tens of thousand visits per second.