Autonomic Workload Management

Many computer systems manage workloads - for example, any web server receives numerous requests on an ongoing basis, and must allocate resources to the requests in a way that yields satisfactory performance. The trend towards the shared use of underlying infrastructures by diverse types of workload, for example in Cloud Computing, means that workload management is likely to become increasingly important and challenging. Where properties of a workload are unpredictable, there are often benefits to be derived from autonomic workload management - where things like the allocation of resources to different types of request, are automatically changed while requests are being executed. Although there has been significant effort directed at various aspects of workload management, several important open problems remain. For example, very different techniques have been applied for decision making in autonomic systems, but it is often less than clear when specific approaches are likely to be most effective. In addition, the interplay between different autonomic components is often difficult to anticipate. This studentship will investigate and evaluate different techniques for workload management, with a view to improving understanding both of the applicability of existing methods and with a view to developing new techniques for autonomic workload management. For example, control-based methodologies will be explored for this purpose, and the proposal of new models for the systems treated will be also pursued.