In this paper we investigate the problem of delivering quality of service (QoS) guarantees in the context of communication networks, studied through stochastic fluid models. The paper follows the approach developed in (C. G. Cassandras, et al., 2002), (C. Panayiottou, et al., 2004). Its main contribution is that it develops infinitesimal perturbation analysis (IPA) estimators of a cost function that also includes delay violation constraints, rather than the average delay that is typically investigated in [1], [2] and other similar work.