VieDAME for WS-BPEL : a non-intrusive dynamic adaptation and monitoring solution for WS-BPEL based on aspect oriented programming

Abstract

The service-oriented computing approach provides the basis for various research projects from industry and academia that aim to build large scale, manageable and extensible software systems. By combining independent, self-contained and coarse grained entities, called services, a loosely coupled but structured network of computational elements emerges. Consequently, these services can be used to model processes that align to certain business models to automate enterprise business processes. Web services are the most common technology to build the aforementioned computational elements. They rely on proven technologies such as HTTP and are the building blocks of information systems for many enter- prises. In the majority of cases, the orchestration layer for combining services into structured processes is provided by BPEL, which is the de-facto standard for Web service composition. While being developed by big players such as IBM and Microsoft, the BPEL standard does not define any requirements for neither runtime process monitoring nor dynamic runtime exchange of partner services.<br />The lack of those capabilities results in reduced applicability of BPEL in high- availability scenarios. This thesis introduces the Vienna Dynamic Adaptation and Monitoring Extension for BPEL (VieDAME for short), which addresses the issues mentioned above. VieDAME can be described best as an extension to an existing BPEL execution environment. It transparently adds capabilities that allow runtime monitoring of partner service communication and runtime part- ner service exchange by introducing an adaptation and monitoring layer based on Aspect Oriented Programming. No modifications to the BPEL engine, the BPEL process or the related services are required. Additionally, the proof-of- concept implementation of the VieDAME approach provides a comprehensive user interface that can be utilized to access QoS statistics and configure runtime adaptation. The thesis reviews underlying technologies and thoroughly discusses frameworks and tools incorporated into VieDAME, and finally describes an in- depth evaluation that proves the applicability in high-load situations.<br />