RESOURCE ALLOCATION METHOD IN DISTRIBUTED CLOUDS

摘要

A method for allocating resources in a distributed cloud computing including a plurality (N) of interconnected datacenters (V) for performing a user task requiring a plurality (NVM) of virtual machines (VM), the method comprising:
a) generating a plurality of sets (F) of datacenters, each set (F n ) comprising a different respective first datacenter (V) and one or several additional datacenters, the additional datacenter(s) being iteratively selected among datacenter candidates (V i ) chosen among not-yet selected datacenters of said plurality of datacenters, until covering the needs in virtual machines (NVM) for performing the user task, at least by comparing communication cost values (C i ) computed for each datacenter candidate (V i ) based at least on the communication cost ( g i ) from already-selected datacenter(s) of the set (F n ) to the current candidate (V i ) and on an estimated communication cost ( h i ) induced by adding the remaining datacenters needed after selecting the current datacenter candidate, and
b) selecting a set (F B ) of datacenters (V) for performing the user task based at least on a comparison of the total communication costs of the sets (F) of datacenters generated at step a).