Predicting service request rates for adaptive resource allocation in SOA

Service orientation is rapidly becoming the common practice in the IT world. A price one often has to pay for the advantages of service oriented architectures (SOA) is performance deterioration. SOA performance heavily depends on the allocation of computational resources to services. The needs of services in computational resources are however changing, depending e.g. on the environmental factors and changes in business processes (and hence service orchestrations). To ensure good performance results, the resource allocation should respond to the changes in the SOA environment. In this paper we focus on the detection of the changes in the environment and the prediction of the expected service requests rates. For this purpose we first discover a stochastic model of the service request rates. Then we monitor the system to detect changes in the environment behaviour and signal the necessity to reconsider the resource allocation, providing a prediction of the service request rates for the coming period. Moreover, we monitor whether the model is still a fair reflection of the behaviour, and when necessary, we adapt the model appropriately.