Toward an Optimized Servers Placement for Mobile Applications Offloading

Application's offloading is considered as a promising approach to overcome the two major limitations of mobile handsets, namely their lack of computing capacity and limited energy autonomy. In a preliminary recent paper, we have proposed the Mobile Application's Offloading (MAO) algorithm enabling to shift applicative jobs from mobile terminals to close remote servers. Compared to other proposals from the literature, the originality of this algorithm consists in considering the Quality of Experience (QoE) perceived by the end-users as an additional decision criterion before proceeding to any application offloading. By sake of simplicity, in order to specify the MAO algorithm, we have assumed that the remote server was systematically located at the antenna's site, that is at the center of the radio cell in which is located the mobile user. In the present paper, we extend our analysis to the more realistic context of Mobile Cloud Computing (MCC). For that purpose, we consider point- tomultipointWavelength Division Multiplexing Passive Optical Network (WDM-PON) infrastructures exploiting the Cloud Radio Access Network (CRAN) principles. In such configurations, each leaf of the optical tree is equipped with a radio-antenna. Unlike our previous study, the remote offloading servers can be placed at different splitting points of such optical infrastructures. We assess by means of closed-forms fitting functions the performance of our algorithm in terms of rejected jobs. Rejected jobs are those that cannot meet the expected users QoE or/and the energy efficiency requirements. We end this paper with proposing an optimized applications servers placement in a typical CRAN configuration.