Enabling SLA Monitoring for Component-Based SOA Applications – A Component-Based Approach

In service provisioning relationships, contracts re-lating to Quality of Service (QoS) are agreed in the form of a Service Level Agreement (SLA), which establishes conditions that must be met by the provider of the service. SLAs are often described as a set of Service Level Objectives (SLO). To watch the fulfil-ment of this contract, these SLOs must be monitored at runtime performing SLA monitoring [1]. The ability to collect QoS parameters is particularly important in Service Oriented Architectures (SOA), where the situation is that of a set of loosely cou-pled interacting heterogenous services from different providers. The monitoring and management task is thus intrinsically a distributed matter. Component-based ap-proaches have emerged as a suitable alternative to traditional, purely service oriented ones (e.g. Web Ser-vices based orchestrations) to design and build SOAs, much strongly enforcing principles like reusability, encapsulation, composability, and interoperability. The Service Component Architecture (SCA) has emerged as a standard specification, technologically agnostic, for designing and building SOA applications, by mod-eling services as components which can be reached by other services or by external applications through the use of appropriate bindings. Runtime managements concerns, however, are not addressed by this specifi-cation and are left as a platform specific issue. Monitoring the runtime behaviour of such appli-cations and watch the compliance to an SLA is not an easy task. Online data must be extracted and QoS metrics must be kept up to date; the composing and bindings can change and new components may need to be monitored; the SLOs can also evolve at runtime. There is an overhead cost and a tradeoff must be made with respect to the amount of data to collect. Several existing solutions are oriented to specific technologies like Web Services [2] or J2EE [3], or rely on measurements at the network level, and detec