This paper presents an algorithm for evaluating the probability that connectivity can be maintained between two given nodes in a physical network affected by a disaster. Nodes and links in a disaster area are probabilistically broken, and the disaster area is modeled using a half plane. This paper also proves that this probability of connectivity increases for a generic network topology when the perimeter length of the convex hull of a physical link route decreases, and that it becomes maximum when these physical link routes become straight line segments. In addition, this paper proposes an optimal server placement method by considering robustness against disaster and an optimal link/node replacement strategy determining which nodes or links should be replaced with those robust against disaster. Intuitive node (link) replacement strategies are also suggested based on the analysis of this paper.