Provisioning Algorithms for Service Differentiation in Middleware Appliance Clusters

Abstract

Service oriented architectures (SOA) and XML-based Web Services have become the technology of choice in enterprise networks. These networks support multiple services and are typically architected in multiple computing tiers, with a main service tier for the business logic and a separate, offload" tier, for, say, the CPU-intensive XML processing. The offload tier is typically populated by clusters of middleware appliances, usually hardware-assisted devices that are optimized for their tasks. Service differentiation refers to the generic problem of managing the enterprise network resources in order to achieve desired performance objectives on a per service basis. In our research, we defined SAA/SDA (Service Activation Algorithm/Service Deactivation Algorithm) and its variations that manages the CPU allocation in the appliance tier, in order to provide service differentiation. The main design objective of SAA/SDA is to overcome the disadvantages of the present known, static solutions. We analyze the performance of SAA/SDA via simulations.