We propose distributed algorithms for real-time monitoring and admission control that allow base stations in heterogeneous wireless cellular networks to dynamically serve mobile users under the constraint of: 1) accommodating all active transmissions in a single shared channel; and 2) guaranteeing a minimum signal-to-interference-plus-noise ratio (SINR) to each served user. In particular, we develop distributed techniques for iterative real-time computation of the spectral radius of an unknown network matrix (often the Perron root of the matrix) that indicates the time-varying limits of power control stability, i.e., the limits of network capacity. Solely locally available information is used as algorithmic input. By drawing a formal analogy with the Google PageRank algorithm, the computations are shown analytically to be exponentially fast and sufficiently accurate for optimal (error-free) stability detection. Numerical simulations of an existing office building demonstrate the applicability of the proposed algorithms to actual UMTS W-CDMA systems characterized by discrete power control with limited step-size.