Analysis of the metrics that influence the performance of SOAP/XML message routing in ESB Applications

Abstract

RESEARCH QUESTION: What are the key performance metrics that influence the performance of SOAP/XML message routing tasks in ESB Applications?\

TASKS: Finding ESB performance evaluation set up and method from existing literature. Deciding on the possible performance metrics to be used in this research from existing literature Designing test scenarios and typical ESB tasks according to the existing literature.

ABSTRACT: Enterprise Service Bus (ESB), as the code component of Service-orientated Architecture (SOA) implementation, has a significant impact on whole system performance. This thesis zooms in to evaluate the performance of different ESB applications in different SOAP/XML message routing tasks, and gain an understanding of the influence of each performance metric on ESB performance.

The research goal is to find out the key ESB performance metrics which influence the most in ESB performance from different ESB tasks and application settings. Compared with existing ESB performance evaluations, federated ESB evaluations, as well as performance metrics correlation analysis, are the main contributions of this study. In terms of the research scope, f our ESB applications – Tibco, BizTalk, Fuse, and Mule – were tested; bothsingle ESB experiments and federated ESB experiments have been conducted; nine test cases were included in single ESB experiments, 16 test cases were involved in the federated ESB experiments, and three ESB tasks have been performed during experiments.

The single ESB experiment results show that Mule ESB applications had the best performance and Fuse ESB applications had the lowest performance. The federated experiment results show three good performance cases with Mule ESB involved, which is consistent with the single ESB experiment results. The federated experiment results also suggest that the combination of two well-performing single ESBs is not necessary to result in good performance ESB applications

Overall, response time was the performance metrics that gave the best indication of ESB performance and influence the most to ESB performance. The findings of this research not only allow architecture to choose suitable ESB applications but also enable more research on federated ESB evaluation. Apart from this, the research contribution to ESB performance metrics correlation analysis is very useful to the ESB performance testing and monitoring community. Nevertheless, the research could have included more test cases, and further statistical analysis and result verification would bring a more comprehensive understanding of ESB performance.