DRS: a fault tolerant network routing system for mission critical distributed applications

We present a novel proactive routing algorithm (dynamic routing system, DRS) that consistently searches for failures via frequent ICMP (Internet control message protocol) echo requests. Our algorithm differs from its predecessors in that it is proactive instead of reactive by looking for failures before they affect message transmissions. When a failure is detected, an alternative route is identified and used. Based on our actual implementation, we developed an analytical model of the DRS to evaluate its potential use for large networks. Using this model, we computed for various network sizes the fault identification times given a percentage of network usage. Calculations are carried out for a typical 10 Mbs Ethernet.