Heartbeat has allowed us to provide decreases in both planned and unplanned outages. To accomplish this, we created a fairly simple configuration involving switching the gateway IP address between the cluster nodes, and dynamic routing daemons for the links with our upstream ISPs.
Heartbeat and spare hardware allowed us to create a prototype pair of redundant routers at no cost beyond our time. Based on my experience with other commercial high availability solutions, I would estimate that heartbeat took no more time to learn and deploy than other high availability solutions.
After the initial test environment was set up and thoroughly tested, we were able to deploy the redundant routers with no additional cost outlay. We were able to re-purpose a recently released server to run as the standby node, bring that up in our network for a week to ensure there were no problems with it, and then re-load our regular router with the new configuration, and bring it up as the primary.
While we can now survive a number of networking and hardware failures, we're finding that one of the primary benefits of this solution is that we can do hardware and software maintenance on the routers while experiencing no downtime. A clean shutdown of the primary router results in a seamless fail-over to the secondary machine with no lost packets.
Heartbeat has allowed us to provide the benefits of a very expensive Cisco solution for the expense of a couple of dual-port gigabit cards. I haven't priced it out, but based on other hardware we've purchased I would estimate that a similar Cisco solution would have cost at least $20,000 and possibly up to $100,000. Heartbeat has allowed us to provide world-class service to our hosting clients at reasonable costs. Additionally, we were able to prototype the deployment and provide a proof-of-concept at no cost (more info).
Heartbeat is a great addition to the Linux software stack. Thanks Alan.