In these last few years we observed a tremendous raise in the field of personal communication and location- context-based services. Unfortunately, in regard to these kind of services, standard connectivity, such as 3G, is limited by scalability issues and its strictly pull-based service model. As an alternate solution, Delay-Tolerant Networks (DTN) have already been proposed as a way to obtain a scalable and efficient urban backbone by leveraging an already available public transportation system. To bound delay of DTNs when deploying in an urban environment the most cost-effective solution is the partial involvement of road-side infrastructure. This paper addresses the problem of infrastructure allocation in a real city environment by analyzing the topology and the timetable of a real urban setting to propose an algorithm suitable for reducing deployed equipment. By means of extensive simulations we show that, a very limited number of exchange points leads to a significant performance improvement and helps planning for a bounded delivery delay.