Automatic real-time control of railway traffic perturbations has recently received the attention of practitioners. The aim is to make use of mathematical algorithms to maintain the required service availability during unplanned disturbances to operations. In the literature many tools for real-time traffic control are proposed, but their effects on traffic have never been studied neither in real life nor in realistic simulation environments. We can mention only a few pilot tests and a unique installation in the Lötschberg Base tunnel in Switzerland, which is in any case an ad-hoc implementation not extendible to other case studies. In this paper we present the ON-TIME framework for the real-time management of railway traffic perturbations. The main innovation is a standard web service-oriented architecture that ensures scalability and flexibility. A standard RailML interface is used for the input/output data of the modules, allowing immediate applicability of the framework to any network having a RailML representation. The scalability makes the framework independent from the number of modules and the amount of data exchanged. The flexibility permits any module to be replaced with others having similar features. The framework is tested in a closed-loop with the simulation environment HERMES for a perturbed traffic scenario on the Swedish Iron Ore line. Tests are performed for two different replanning algorithms (ROMA and RECIFE) used as conflict detection and resolution modules of the framework. The analysis represents a proof-of-concept to confirm the effectiveness of our framework in automatically solving conflicts and deadlocks during perturbed traffic conditions.