Planning Databases Service Level Agreements through Stochastic Petri Nets

Abstract

The Service Oriented Architecture (SOA) has become a pattern for managing business transactions through
distributed services, usually supported by third-party providers. In the SOA world, Service Level Agreements (SLA) are
used to establish the requirements between customers and providers. Among the clauses agreed in SLA are those related
with databases execution time, which have direct relationship with the overall web services performance. However, the
high variability, typical of the SOA environments, makes di?cult the negotiation of an appropriate SLA that could be
guaranteed in practice. Thus, e?orts to predict the quality of SOA-based transactions are justi?ed by anticipate possible
problems that tend to arise at run-time, disturbing the SLA clauses. In this paper we propose a simulation modeling
approach, based on stochastic Petri nets, for performance evaluation of databases requests in data-intensive business
applications. Through our proposal it is possible to predict the resources consumption and performance degradation of
databases, upon the variation of the workload levels, both at design-time and run-time. A case study was conducted in
order to illustrate our contributions.