Measurement-based performance analysis of e-commerce applications with Web services components

Web services are increasingly used to enable interoperability and flexible integration of e-business systems. In this paper we focus on measurement-based performance analysis of an e-commerce application which uses Web services components to execute business operations. In our experiments we use a session-oriented workload generated by a tool developed accordingly to the TPC-W specification. The empirical results are obtained for two different user profiles, Browsing and Ordering, under different workload intensities. Unlike the previous work which was focused on the overall server response time and throughput, we present Web interaction, software architecture, and hardware resource level analysis of the system performance. In particular, we propose a method for extracting component level response times from the application server logs and study the impact of Web services and other components on the server performance. The results show that the response times of Web services components increase significantly under higher workload intensities when compared to other components. From the hardware resource measurements it is obvious that the higher response times of Web services components are due to parsing XML messages and contention for database resources. The results of our study identify software components and hardware resources which are potential bottlenecks in the system and thus provide valuable information for capacity planning of e-commerce applications.