I don't work for a huge company and we don't have Tera-byte sized databases but we have multiple mission critical 120GB databases and an awesome NetOps group. Instead of buying everything brand spanking new, they bought some awesome hardware that had been refurbished. What that allowed us to do was to afford not just 2 but 3 identical systems. 2 of them are clustered onsite. We regularly test them by forcing a failover. The "outage" is usually something less than 5 seconds. The third system is in another state about 500 miles away. It's the DR system. I don't know what tools they're using for all of this but the fail over to the DR site is also measured in very few seconds and it's all automatic. They've made my job as a DBA a proverbial cake walk when it comes to HA and DR.

During the "failover", most users lose no work and most don't even know the failover occurred.

Living on the bleeding edge is expensive. Instead of buying the latest and greatest which also commands the most expense, they bought the latest and greatest of the refurb world. For what most people would have paid a little more for just 1 system, we have 3. And, I have to tell you, these systems aren't some beatup ol' relics. They might not beat "state of the art" but they do a damn fine job of keeping up.