Concurrent Management of Composite Services According to Response Time SLAs

The ongoing aim behind the work described here is to investigate support for a data centre type environment where an application can be implemented as a composition of components, or workflow. For instance, an application might reuse established components. As is typical in data centre operation, hosting of an application is governed by a Service Level Agreement (SLA) between provider and application owner. Such an SLA describes levels of service and corresponding monetary charges paid to and or by the provider. A wrapper has been built around an existing workflow engine (ActiveBpel) to support investigation of such a facility. Previous work has illustrated how a range of control loops can be implemented in this wrapper. The work described here focusses on the implementation and evaluation in this wrapper architecture of a policy which determines machine allocation and admission queue size for a collection of applications, seeking to maximize provider profit, particularly during overload conditions.