In this paper, we present an analytical tool called RETINA for analyzing real-time constraints in complex networks with stochastic traffic. RETINA is based on a statistical method called Real-Time Queueing Theory and can predict the fraction of late tasks in networks of queues given traffic conditions and queue protocol specifications. Upon submitting a profile of the network structure and data flow descriptions, the user is able to query certain performance metrics. Currently RETINA supports two types of real-time scheduling algorithms (EDF adn FIFO) and three types of performance queries: the deadline miss rate of a designated flow, the minimum feasible deadline for a particular flow, and whether all QoS requirements of all flows will be met when a new flow is added. Simulation results show excellent agreement with the results predicted by our tool. RETINA has made it possible to analyze larger and more realistic networks and can be used to address network design tradeoffs. This use of RETINA to identify and quantify network bottlenecks is presented in the paper.