Adaptive and cost-effective service placement

International audienceIn this paper we will discuss the problem of placing service replicas in the cloud network to fulfil customers' demand. Such a problem hass been well studied in "Facility Location" topic; however, in this area opening a new facility is costly and its migration is an expensive decision which makes the solution static. In cloud network infrastructures, services are pieces of software that can be migrated from one server to another in an easier way and therefore the dynamic service placement should be part of the solution. This paper aims to propose a new scheme of optimal dynamic service placement in the context of changing network environment. This scheme reinforces the existing solutions proposed in "Facility Location", allowing at each period of time to find the best locations for services given the network topology, resource availability and customers' demand. The scheme considers the efficiency of the new locations versus the necessary modifications that have to be made to obtain them. The solution has been evaluated by simulation and has shown that small modification in service deployment can achieve high performance close to optimal solutions