Online Optimization in Internet Data Centers

Internet data centers (IDCs) perform multi-customer hosting on a virtual-ized collection of hardware resources. These systems give a new answer to website hosting by delegating all the worry of server management on the IDC provider side. These computing farms have to cope with important issues. Besides management and security considerations we find the important problem of resource allocations. This problem despite its combi-natorial nature is hard to solve since hosted customers increasingly require support for peak loads that are orders of magnitude larger than what they experience in their normal state. Thus, a hosting environment needs a fast turnaround time in adjusting the resources (bandwidth, servers, and storage), assigned to each customer. In this work we present an autonomous system for online resource allocations in IDCs. Our system takes advantage of monitoring informations upcoming from the infrastructure to reconsider its mathematical modelling of the components. Combined to the versatility of Constraint Programming (CP) it performs a continuous adaptation of the allocated resources and ensures a smart hosting of the applications.