Decision as a service: Separating decision-making from application process logic

In a service-oriented architecture, adaptive and evolvable applications should be able to select, configure and compose different existing application services to deal with the changes which can arise from runtime contextual changes or the change of user requirements and preferences. To support this, hybrid service composition approaches have been proposed, in which the core of the application logic is specified in terms of processes, while rules are employed to specify the conditions and constraints to adapt the application behaviour. The rules are then exposed as a decision service which can be employed by the process to make adaptation decision with respect to runtime circumstances. The interaction between processes and decision services are generally performed in a synchronous request-response manner. We argue that such an interaction is not efficient to support different types of adaptation at runtime and therefore asynchronous interaction should also be supported. In this paper, we present an adaptive service provisioning architecture and a decision service template allowing both synchronous request-response interaction and asynchronous notification. To motivate the proposed architecture and the decision service template, we use a blood pressure monitoring scenario from the homecare domain. We also explain the implementation of the proposed approach based on commercially available rule and process engines. Finally, we discuss: 1) what is the efficient way (synchronous request-response interaction vs. asynchronous notification) of calling decision service to execute different types of decision rules; and 2) to what extent the use of a decision service facilitates dealing with unforeseen changes.