Model Checking Probabilistic Real-Time Properties for Service-Oriented Systems with Service Level Agreements

TL;DR

This paper introduces a novel model checking approach for probabilistic real-time properties in service-oriented systems, using Interval Probabilistic Timed Automata (IPTA) to verify SLA guarantees efficiently.

Contribution

It presents the first implementation of an IPTA model checker, extending PRISM, capable of directly verifying SLA specifications in service-oriented systems.

Findings

IPTA can directly capture SLA guarantees.

Model checking IPTA is only slightly more expensive than PTA.

Symbolic analysis makes IPTA verification efficient.

Abstract

The assurance of quality of service properties is an important aspect of service-oriented software engineering. Notations for so-called service level agreements (SLAs), such as the Web Service Level Agreement (WSLA) language, provide a formal syntax to specify such assurances in terms of (legally binding) contracts between a service provider and a customer. On the other hand, formal methods for verification of probabilistic real-time behavior have reached a level of expressiveness and efficiency which allows to apply them in real-world scenarios. In this paper, we suggest to employ the recently introduced model of Interval Probabilistic Timed Automata (IPTA) for formal verification of QoS properties of service-oriented systems. Specifically, we show that IPTA in contrast to Probabilistic Timed Automata (PTA) are able to capture the guarantees specified in SLAs directly. A particular challenge in the analysis of IPTA is the fact that their naive semantics usually yields an infinite set of states and infinitely-branching transitions. However, using symbolic representations, IPTA can be analyzed rather efficiently. We have developed the first implementation of an IPTA model checker by extending the PRISM tool and show that model checking IPTA is only slightly more expensive than model checking comparable PTA.