I often hear HA at work. There are some HA projects on-going, or done, I am not sure. I am not involved in the projects, yet. I hope. :D
As a cloud service provider, high availability design is one of the vital architecture considerations. With HA architecture, the service would run at optimal performance even it is running at high load or one of the server node is down. Anyway, it is normally measured as the percentage of uptime in a year. Scheduled downtime most of the time does not count in the HA measurement.
I quote this from Wiki : By doing this, they can claim to have phenomenally high availability, which might give the illusion of continuous availability.
There's another numbered system for this HA measurement. One nine, refers to 90%; two nines is 99%; three nines is 99.9%; four nines is 99.99% and so on. The more nines, the better HA the system is.
There are a lot of design principle to ensure HA. For example, redundancies, load balance, failover mechanism, etc.
Some articles for further read.
Update: Availability refers to how long your service is up and running without interruption.