A performance comparison of dynamic Web technologies

Abstract

Today, many Web sites dynamically generate responses "on the fly" when user requests are received. In this paper, we experimentally evaluate the impact of three different dynamic content technologies (Perl, PHP, and Java) on Web server performance. We quantify achievable performance first for static content serving, and then for dynamic content generation, considering cases both with and without database access. The results show that the overheads of dynamic content generation reduce the peak request rate supported by a Web server up to a factor of 8, depending on the workload characteristics and the technologies used. In general, our results show that Java server technologies typically outperform both Perl and PHP for dynamic content generation, though performance under overload conditions can be erratic for some implementations.