High track densities and event rates in the LHC tracking detectors put stringent demands on the track reconstruction software in terms of performance, speed and maintainability over the experiment's lifetime. The ATLAS experiment is therefore commissioning a new software for the track and event reconstruction in the Inner Detector, which will allow to use common code for the off-line and high-level trigger reconstruction, as well as for different detector set-ups. This is achieved by breaking the algorithmic structure down to small, independent modules which communicate by common interfaces and use a new, global event data model. The Inner Detector has just completed the new track reconstruction within this framework and has successfully applied it to all kinds of simulated LHC events as well as test tracks from the real detectors. The preparation for the first real LHC data is now continuing at a fast pace, for example by integrating the algorithms to calibrate and align the detector further into the new Inner Detector software, and by implementing new tracking techniques from the literature.