In noncooperative networks, resources are shared among selfish users, which
optimize their individual performance measure. Traditional design methods
tend to perform poorly in such net networks, as they do not take into accou
nt the inherent noncooperative nature of the network users, Such networks r
equire specialized design techniques in order to achieve efficient utilizat
ion of resources. We consider the generic and practically important class o
f atomic resource sharing networks, in which traffic bifurcation is not imp
lemented, hence each user allocates its whole traffic to one of the network
resources. We investigate topologies of parallel resources within a game-t
heoretic framework and establish the foundations of a design and management
methodology that enables operation of such networks efficiently, despite b
oth the lack of cooperation among users and the restrictions imposed by ato
mic resource sharing. We study various problems pertaining to capacity allo
cation, pricing, and admission control, and show that their solutions are s
ubstantially different from those corresponding to traditional networks.