Load server plans base 'server hours' on an Amazon Web Services Large instance. As with AWS, a server hour starts when the server starts and ends when the server is terminated, and is rounded up to the nearest hour. So, a server that is running for 90 minutes consumes 2 server hours.
The amount of concurrent virtual users, page views, or transactions that you can get out of single server is going to be dependent on the complexity of the test. For more on calibration and factors that impact VUs per server click here.
There is an option in CloudTest to choose a different sized server. After calibrating, you may find that a high memory or high CPU option will yield much more load than a Large. For purposes of counting hours, larger instances will consume hours faster than a Large. For example, each hour of an Extra Large counts as 2 load server hours against your plan. So, if you find you can get more than twice the amount of load out of an XL than a Large, it may be worth it to launch XLs using your plan instead.
Here are the relative values
Large: 1 hour = 1 load server hour (1x)
Extra Large: 1 hour = 2 load server hours (2x)
High Memory Extra Large: 2x
High Memory Double Extra Large: 3x
High Memory Quadruple Extra Large: 6x
High CPU Extra Large: 3x
Cluster Compute Quadruple Extra Large: 5x