The Impact of Transient Loads on the Performance of Service Ecologies

Abstract

Well defined, loosely coupled services are the basic building blocks of the service-orientated design-integration paradigm. Services are computational elements that expose functionality in a platform independent manner and can be described, published, discovered, orchestrated and consumed across language, platform and organizational borders. Using service-orientation (SO) it is fairly easy to expose existing applications/resources and to aggregate them into novel services called composite services (CS). This aggregation is achieved by defining a workflow that orchestrates the underlying services in a manner consistent with the desired functionality. Since CS can aggregate atomic and other CS, they foster the development of service layers and reuse of already existing functionality. This gave rise to the idea of service networks and consequently service ecologies. In these service ecologies providers and consumers build a self-organizing system across platform and organizational borders, in which resources and business processes are shared. However, the ease with which resources are shared raises concerns in regards to the dependability (e.g. reliability, availability) of the resulting system and consequently the feasibility of service ecologies. This paper focuses on the performance issues that arise when composite services are exposed to transient loads and discusses the need for admission control and adaptive scheduling as a means for improving the performance in service ecologies.