|
|
|
| e-Pub |
Previous | 
Home
Section: Research Program
Transparent Resource Management for Clouds
Given the extremely large offer of resources by public or private clouds, users need software assistance to make provisioning decisions. Our goal is to design a cloud resource broker which handles the workload of a user or of a community of users as a multi-criteria optimization problem. The notions of resource usage, scheduling, provisioning and task management have been adapted to this new context. For example, to minimize the makespan of a DAG of tasks, usually a fixed number of resources is assumed. On IaaS clouds, the amount of resources can be provisioned at any time, and hence the scheduling problem must be redefined using one new prevalent optimization criterion: the financial cost of the computation.
Provisioning strategies
The provisioning strategies are hence central to the broker. They are designed after heuristics which aim to fit execution constraints and satisfy user preferences. For instance, lowering the costs can be achieved with strategies aiming at reusing already leased resources, or switch to less powerful and cheaper resources. However, some economic models proposed by cloud providers involve a complex cost-benefit analysis which we plan to address. Moreover, these economic models incur additional costs, e.g. for data storage or transfer, which have to be taken into account to design a comprehensive broker.
User workload analysis
Another possible extension of the capability of such a broker is the analysis of user workloads. Characterizing the workload might help to anticipate the behavior of each alternative provisioning strategy. The objective is to allow the user to select the suitable provisioning solution thanks to concrete information, such as completion time and financial cost.
Simulation of cloud platforms
Providing concrete information about provisioning solutions can also be achieved through simulation. Although predicting the behavior of applicative cases in real grid environment is made very difficult by the shared (e.g. multi-tenant), heterogeneous and dynamic nature of the resources, cloud resources (i.e. VMs) are perceived as reserved and homogeneous and stable by the end-user. Therefore, proposing an accurate prediction of the different strategies through an accurate simulation process would be a strong decision support for the user.