Run-time resource management in SOA virtualized environments

Service Oriented Architecture (SOA) and virtualization of physical resources are key emerging technologies which are driving the interest of research both from industry and academia. The combination of the two is leading to a new paradigm - the Service Oriented Infrastructure - (SOI) whose goal is to provide a flexible solution for accessing component based service applications on demand. SOI environments are characterized by high workload fluctuations which cannot be accommodated by separating design and run-time point of view as traditionally done in Software Engineering practice. Hence, the design of SOA applications has to be complemented with issues related with the run-time resource provisioning. In this paper the problem of determining the optimum capacity allocation for multiple Virtual Machines which share the same hosting environment is addressed. The overall goal is to maximize the Service Provider profits associated with multiple classes of Service Level Agreements. The capacity allocation problem is modeled as a non-linear problem which is optimally solved. The effectiveness of our solution is assessed by performing real experiments in a prototype environment.