A full-scale railway axle made of medium strength steel EA4T and adopted for high-speed applications, is deep rolled. The induced residual stresses were experimentally characterized by X-ray diffraction and hole drilling. A realistic finite element model is proposed to overcome some of the existing shortcomings in simulation of deep rolling. Deep rolling coverage is defined, formulated and incorporated into the simulation. The model is validated by the experimental measurements. A parametric study is performed to investigate the effect of rolling force (4-19 kN), rolling feed (0.1-0.7 mm/rev) and roll geometry (1.5-10 mm roll tip radius) on the distribution of residual stresses and the induced hardening. A fatigue crack propagation algorithm is used to analyze the influence of technological parameters on the lifetime of railway axles. Lower feeds, higher loads and thicker rolls, all resulting in higher deep coverage, can result in higher improvement against fatigue crack propagation. However, extremely high coverage can deteriorate the performance of deep rolled components. As a general rule of thumb, adopting deep rolling feed to get a coverage level of 500-900%, while avoiding too high rolling loads and too thin rolls, can induce a suitable compressive residual stress distribution; and effectively prevent/retard fatigue crack propagation