Efficient Monitoring of Web Service SLAs∗

Web services are increasingly used in inter-organizational settings. If an organization depends on the service quality provided by another organization it often enters into a bilateral service level agreement (SLA) to precisely determine service quality and permitted service use. SLAs then also determine penalty payments as risk mitigation against poor service quality and overuse of the service. Once these agreements are entered into, it becomes necessary to monitor for both poor service quality and also abuse of the provision beyond the agreed limits. We address the question of how service level agreements can be monitored efficiently and automatically. We show how timeliness constraints, such as latency, throughput, availability and reliability, in formal service level agreements can be translated into timed automata. We attach time stamps to SOAP messages and consider these messages as timed letters. We are then able to reduce the question of detecting SLA violations to acceptance of timed words by the timed automata that have been derived from the SLA. Acceptance of a timed word by a timed automaton can be decided in polynomial time and because the timed automata can operate while SOAP messages are exchanged at run-time there effectively is only a linear run-time overhead. We evaluate the efficiency and scalability of this approach using a large-scale case study in a service-oriented computational grid.