KIT Research Student Awards

Daniel Funke and Fabian Gorsler were granted KIT Research Student Awards and will be funded to work on individual research projects at the Descartes Research Group.

Daniel Funke will work on a game theoretic approach towards a truthful resource reservation mechanism tailored at reservation/pricing models in Cloud Computing:

Prudent capacity planning to meet their clients future computational needs is one of the major issues cloud computing providers face today. By offering resource reservations in advance, providers gain insight into the projected demand of their customers and can act accordingly. However, customers need to be given an incentive, e.g., discounts granted, to commit early to a provider and to honestly reserve their predicted future resource requirements. Customers may reserve capacity deviating from their truly predicted demand, in order to exploit the system for their own benefit, thereby causing futile costs for the provider. Therefore, truthfulness is an essential quality of such a mechanism. The idea is to use general reservation/pricing models as they can be found in economic literature and to apply them to reservation scenarios specific for the cloud computing market.

Fabian Gorsler will work on an approach to estimate resource demands in virtualized environments:

Hosting enterprise services on virtualized platforms requires an efficient performance management strategy at the application level. A precondition is the ability to predict how the performance of running services would be affected if the system configuration or the workload changes. For describing performance-relevant behavior, information is needed about which resources a service demands and to what extent the resources are used. Such information is crucial for accurate performance predictions. To the best of our knowledge, there is no resource demand estimation approach which specifically takes the characteristics of the virtualization platform into account. The idea is to use monitoring data obtained at system run-time to estimate VM-specific resource demands that include VM-specific hypervisor overhead.