rekoil saved me typing this:
The industry standard for usage-based bandwidth billing is a formula called "95p". What happens is that the circuit load is constantly measured (typically in 5-minute averages) throughout the billing period. The highest 5% of those measurements are then thrown out, and the highest measurement remaining is what the month's bill is based on.
This has two implications: First, a customer doesn't get bent over if they get a traffic spike or DoS that is short-lived...you can push you pipe to capacity for almost a day and a half before you actually get billed for running the connection at that level.
Second, given that most pipes see their highest usage during business hours, weekend traffic rarely exceeds the current 95p value, and as such is effectively "free". That's why most companies do their network-based backups at night and on weekends.
|