Network design for OSPF routing Luciana S. Buriol 1 , Paulo M. Franc ¸a, 1 Mauricio G.C. Resende 2 , and Mikkel Thorup 2 1 Faculdade de Engenharia El´ etrica e de Computac ¸˜ ao, UNICAMP, SP, Brazil. buriol,franca @densis.fee.unicamp.br, 2 Internet and Network Systems Research, AT&T Labs Research, Florham Park, NJ, USA. mgcr,mthorup @research.att.com Abstract. Internet protocol (IP) traffic follows rules established by routing pro- tocols, such as Open Shortest Path First (OSPF). Each router computes shortest paths using weights assigned by the network operator, and creates destination tables used to direct each IP packet to the next router on the path to its final des- tination. Furthermore, the routing protocol is used to establish procedures to be taken in case of a failure in the network. In this extended abstract, we describe a new genetic algorithm for designing a network with minimal total link capacity necessary to route demand without overload in case of any single edge or node failure. The Internet is made up of many routing domains, called autonomous systems (ASes). Internet protocol (IP) traffi c flows follows rules established by routing pro- tocols. Shortest path fi rst protocols, such as Open Shortest Path First (OSPF), are the most commonly used Interior Gateway Protocols (IGPs). These routing protocols direct traffi c based on link weights assigned by the network operator. Each router in the AS computes shortest paths and creates destination tables used to direct each IP packet to an outgoing link on shortest paths to its fi nal destination. If a router has multiple out- going links on shortest paths to a given destination, it splits traffi c evenly over these links. It is also the role of the routing protocol to specify how the network should quickly react to changes in the topology of the AS, such as when transmission lines go out of service or come back on, routers crash, or network policies change [5]. In such situa- tions, IP traffi c is re-routed through the shortest paths not traversing the affected part of the network. If such links have insuffi cient bandwidth capacity, overload occurs. To satisfy requirements for Quality of Service (QoS) in IP routing, it is desirable to design a network to easily handle a single link or router failure without causing overload. One possible solution is to maintain part of the link bandwidth free in the eventuality of failures. This extended abstract addresses the issue of designing a OSPF-routed network with minimum total link capacity needed to route the required demand and handle any single (link or router) failure. We assume the topology is given but link capacities must be determined. The capacity is limited to a discrete set of values 0 m 2m 3m , i.e. multiples of a unit value m . In fact, the capacity of a link corresponds, in this context, to its multiplicity. For simplicity we consider m 1. Our aim is to design an effi cient