The rapid growth of telecommunication capacity, driven in part by the wide-ranging deployment of fiber-optic technology has led to increasing concern regarding the survivability of such networks. Ln communication networks, survivability is usually defined as the percentage of total traffic surviving some network failures in the worst case. Most of the survivable network design models proposed to date indirectly ensure network survivability by invoking a connectivity constraint, which calls for a prespecified number of paths between every distinct pair of nodes in the network. In this paper, we introduce a new network design model which directly addresses survivability in terms of a survivability constraint which specifies the allowable level of lost traffic during a network failure under prescribed conditions. The new model enables a network designer to consider a richer set of alternative network topologies than the existing connectivity models, and encompasses the connectivity models as special cases. The paper presents a procedure to compute link survivability, develops an integer programming formulation of the proposed survivability model, and discusses a special case of practical interest and its associated heuristic procedure. The proposed heuristic is tested on data from real-world problems as well as randomly generated problems.