Applying a Stochastic Model to a Dynamic, QoS Enabled Web Services Hosting Environment

Data centres which host Web services for other organisations and users in a Grid environment must provide for Quality of Service (QoS) requirements to be specified to ensure deployed services perform as desired. As services hosted by a data centre receive unpredictable rates of demand, servers must be allocated dynamically to service pools that are over utilised to avoid breaking QoS requirements. This work describes how a cost based stochastic model for resource allocation is used in data centre middleware to balance server utilisation, and how the model was used to enable a data centre to meet QoS requirements. The stochastic QoS model is compared to two other QoS models and is shown to be the most effective in a number of experiments.