The one problem we have with measuring uptime is that the server uptime doesn't necessarily mean the SQL Server was accessible at all times. We made a custom check to run from a centralized server to check connectivity every 9 minutes to make sure a simple select runs against master. However, it's hard to measure the pct of uptime strictly of SQL Server.

The logic I currently try to use it 100-((num_of_connectivity_failures/160)*100) = pct_uptime_for_connectivty_to_sql

1440 minutes in a day / 9 minutes = 160 connectivity checks per day.

For our missions critical systems, we check connectivity to the individual databases every minute, so the checks would be more 1440 checks per day and the percentage becomes more accurate.

Any other thoughts / suggestions are always welcomed though.

Steve