Fair resource allocation for different scenarios of demands |
| |
Authors: | Emmanuel Medernach Eric Sanlaville |
| |
Institution: | a Université Blaise Pascal, Clermont-Ferrand, France b Université du Havre, Le Havre, France |
| |
Abstract: | This paper considers a resource allocation problem, which objective is to treat fairly all the system users. Usually the requests cannot be entirely predicted, but the manager can forecast the request evolution, this leading to a set of possible scenarios. Such a problem arises for instance in network bandwidth allocation as well as in storage space management. It also appears in the management of computer systems, such as computational grids or in cloud computing, when teams share a common pool of machines. Problems of fair resource sharing arise among users with equal access right but with different needs.Here the problem is tackled by a multi-criteria model, where one criterion is associated to one scenario. A solution is a policy, which provides an allocation for each scenario. An algorithm is proposed and analysed that lists all solutions which are Pareto optimal with regard to the different possible user request scenarios. The algorithm is used offline, but can be adapted, with some additional hypothesis, to be used online. |
| |
Keywords: | Resource allocation Fairness Uncertainty on demands Grid scheduling Multicriteria optimization |
本文献已被 ScienceDirect 等数据库收录! |
|