Ergodicity of a Lévy-driven SDE arising from multiclass many-server queues (Q1737963): Difference between revisions

The paper considers a \(d\)-dimensional stochastic differential equation (SDE) of the form \(dX(t) = b(X(t))dt + \sigma (X(t))dW(t) + dL(t),X(0) = x \in {\mathbb R^d}\). Here (A1) the function \(b:{\mathbb R^d} \to {\mathbb R^d}\) is given by \(b(x) = l - (M + (\Gamma - M)ve')x\) if \(e'x > 0\), and \(b(x) = l - Mx\) if \(e'x \le 0\), with \(l \in {\mathbb R^d},e = (1,\dots,1)' \in {\mathbb R^d},v \in {\mathbb R^d},e'v = 1\), \(M \in {\mathbb R^{d \times d}}\) is a nonsingular \(M\)-matrix such that the vector \(e'M\) has nonnegative components, and \(\Gamma = \mathrm{diag}({\gamma _1},\ldots,{\gamma _d})\)with \({\gamma _i} \in {\mathbb R_ + },i = 1,\ldots,d\); (A2) \({\{ W(t)\} _{t \ge 0}}\) is a standard \(n\)-dimensional Brownian motion, and the covariance function \(\sigma :{\mathbb R^d} \to {\mathbb R^{d \times n}}\) is locally Lipschitz and satisfies, for some constant \(k > 0\), \({\left\| {\sigma (x)} \right\|^2} \le k(1 + {\left| x \right|^2})\), \(x \in {\mathbb R^d}\); (A3) \({\{ L(t)\} _{t \ge 0}}\) is a \(d\)-dimensional pure-jump Lévy process specified by a drift \(\vartheta \in {\mathbb R^d}\)and Lévy measure \(\nu (dy)\). In (A1)-(A3), \(\left\| \cdot \right\|\)is the Hilbert-Schmidt norm, a \(d \times d\) matrix \(M\) is called an \(M\)-matrix if it can be expressed as \(M = sI - n\) for some \(s > 0\) and some nonnegative \(d \times d\) matrix \(N\) with the property that \(\rho (N) \le s\), where \(I\) and \(\rho (N)\)denote the \(d \times d\) identity matrix and spectral radius of \(N\), respectively. The paper studies the ergodic properties of the processes in question. The authors indicate that such SDE describe the limit of the queueing processes arising in multiclass many-server queues with heavy-tailed arrivals and/or asymptotically negligible service interruptions in the Halfin-Whitt regime as special cases. Beyond that, they also study the class of models driven by a subordinate Brownian motion, which contains an isotropic (or rotationally invariant) \(\alpha\)-stable Lévy process as a special case. The authors: \begin{itemize} \item[a)] identify conditions on the parameters in the drift, the Lévy measure and/or covariance function which result in subexponential and/or exponential ergodicity; \item[b)] show that these assumptions are sharp, and \item[c)] identify some key necessary conditions for the process to be ergodic. In addition, they show that for the queueing models described above with no abandonment, the rate of convergence is polynomial, and provide a sharp quantitative characterization of the rate via matching upper and lower bounds. \end{itemize}