A New Approach to Developing High-Availability Server

This paper presents a new approach to developing high availability (HA) server using Rapid Spanning Tree Algorithm and Protocol (RSTP) as specified in the IEEE 802.1w standard. The traditional approach requires the use of multiple IP addresses and complex management software to monitor multiple links. If there is a link failure, the management software requires to perform complex fail-over functions, including IP address binding, maintenance of stability and integrity of active applications, and synchronization of ARP tables on all clients. This new approach, built on RSTP, is 100% transparent to applications and clients, and does not require any management software. Most importantly, the failover time and fall-back time is less than one second. We conducted extensive STP and RSTP experiments in the lab to prove the concept and demonstrate the functionality, and ran an RSTP simulation to demonstrate its performance (measured in fail-over time and fall-back time) for several failure scenarios.