Geo-distributed data centers present a better robustness to failures than monolithic ones, since they have less single points of failure. In this work we tackle the trade-off between robustness and latency caused by the distance between DC sites in geo-distributed data centers. This trade-off is addressed through the formulation of an optimization problem. The results show that the distance between DC sites increases significantly the latency only in case of very strong robustness requirements. On the other hand, the latency increase by the distance is negligible for moderate robustness requirements.