Analysing dependencies in service compositions
ICSOC/ServiceWave'09 Proceedings of the 2009 international conference on Service-oriented computing
Using vector clocks to monitor dependencies among services at runtime
Proceedings of the International Workshop on Quality Assurance for Service-Based Applications
ECSA'11 Proceedings of the 5th European conference on Software architecture
Revealing hidden relations among web services using business process knowledge
ICSOC'11 Proceedings of the 9th international conference on Service-Oriented Computing
Capturing and Analyzing Service Network Models
Proceedings of International Conference on Information Integration and Web-based Applications & Services
| Hi-index | 0.00 |
The Service-Oriented Architecture (SOA) has become today's reference architecture for modern distributed systems. As SOA concepts and technologies become more and more widespread and the number of services in operation within enterprises increases, the problem of managing these services becomes manifest. One of the most pressing needs we hear from customers is the ability to "discover", within a maze of services each offering functionality to (and in turn using functionality offered by) other services, which are the actual dependencies between such services. Understanding dependencies is essential to performing two functions: impact analysis (understanding which other services are affected when a service becomes unavailable) and service-level root-cause analysis (which is the opposite problem: under-standing the causes of a service failure by looking at the other services it relies on). Discovering dependencies is essential as the hope that the enterprise maintains documentation that describe these dependencies (on top of a complex maze and evolving implementations) is vane. Hence, we have to look for dependencies by observing and analyzing the interactions among services. In this paper we identify the importance of the problem of discovering dynamic dependencies among Web services and we propose a solution for the automatic identification of traces of dependent messages, based on the correlation of messages exchanged among services. We also discuss our lessons learned and results from applying the techniques to data related to HP processes and services.