A viscosity solution approach to the asymptotic analysis of queueing systems (Q751705)

Consider a queuing system, the ``state'' \(X_ t\) of which at time t is an n-dimensional vector \(\in \{{\mathbb{Z}}^+\}^ n\), the set of non- negative integers. The problem of interest is the estimation of large deviation behavior of the rescaled system \(X^{\epsilon}_ t=\epsilon X_{t/\epsilon}\), corresponding to large time and large excursions of the original system, under the assumption that the original process \(X_ t\) is a jump Markov process. For example, for \(n=2\), if \(A^{\epsilon}\) is the event that \(x_ t+y_ t\geq M/\epsilon\) for some \(0\leq t\leq T/\epsilon\), given \(x_ 0=x/\epsilon\), \(y_ 0=y/\epsilon\), where M, T are positive reals and \(X_ t=(x_ t,y_ t)\), then via results of the paper it has been claimed that as \(\epsilon\downarrow 0\), \[ P(A^{\epsilon})=\exp [\{-I(x,y)+O(1)\}/\epsilon], \] where the O(1) term converges to zero uniformly in (x,y) in compact subsets of \(\{(x,y):\;x\geq 0,\quad y\geq 0,\quad x+y<M\}\) and \(I(x,y)=u(x,y,0)\), u being the value function of a non-standard deterministic control problem.