Presented is a system and method for distributing a network application across a plurality of geographically dispersed network sites. The system comprises a plurality of network sites connected by a shared network and interconnected by a dedicated non-blocking communication network. The system can use different interconnecting network topologies based on the number of sites to be interconnected. The method balances the network application load and resources across the interconnected network sites based on a distribution policy implemented without burdening the shared network. The method provides redundancy capabilities by detecting the loss of a network site and redistributing the network application load to the remaining network sites.