On the Analysis of Spatially Constrained Power of Two Choice Policies

We consider a class of power two choice based assignment policies for allocating users to servers, where both and servers are located on two-dimensional Euclidean plane. In this framework, we investigate the inherent tradeoff between communication cost, load balancing performance different allocation policies. To end, first design evaluate Spatial Power (sPOT) policy in which each user is allocated least loaded server among its geographically nearest sequentially. When placed square grid, sPOT maps classical (POT) Delaunay graph associated with Voronoi tessellation set servers. show that 4-regular provide expressions asymptotic maximum using results from literature. For uniform placement map balls bins corresponding regions second order diagram lower bound expected prove does not achieve POT benefits. However, experimental suggest efficacy respect cost. Finally, propose non-uniform sampling best metrics. Experimental validate effectiveness our proposed