We address the problem of resource placement in general networking applications, in particular cloud computing. We consider a large-scale service faced by regionally distributed demands for various resources. The service aims at placing the resources across regions to maximize profit, accounting for demand granting revenues minus resource placement costs. Cloud computing and online services, utilizing regional datacenters and facing the problem of where and how much to place various servers, naturally fall under this paradigm. The main challenge posed by this setting is the need to deal with arbitrary multi-dimensional stochastic demands. We show that, despite the challenging stochastic combinatorial complexity, one can optimize the system operation using fairly efficient algorithms.
|Number of pages||3|
|Journal||Performance Evaluation Review|
|State||Published - 4 Sep 2014|
|Event||32nd International Symposium on Computer Performance, Modeling, Measurement, and Evaluation, IFIP WG 7.3 Performance 2014 - Turin, Italy|
Duration: 7 Oct 2014 → 9 Oct 2014