1Flexible QoS-Based Service Selection and Provisioning in Large-Scale Grids

Grid computing is a key enabling technology for the field of e-Science. As scientists rely increasingly on computational support for their work, Grids allow research organisations to pool their resources and spread the cost of expensive hard-ware. Often, such Grids span across organisational boundaries, thereby giving users access to a vast range of services — from running complex simulations to analysing large data sets [1]. However, as Grids become larger, more open and dis-tributed, new challenges need to be addressed [2]. In particular, the behaviour of services may be highly uncertain, as mani-fested by frequent failures and execution delays. This uncer-tainty may be due to network delays, power cuts, competition for resources or machine failures. Furthermore, services will typically be owned by agents that act autonomously and cannot be assumed to behave cooperatively or deterministically [3]