 |
 |
• at least 100 MB storage
• at least 3GB of bandwidth
|
 |
 |
• remote reboot capability
• 100Mbps burstable access
|
 |
 |
• unlimited IPs
• secured cabinet
|
 |
|
 |
|
|
|
 |

The 95th percentile is the smallest number that is greater than 95% of the
numbers in a given set. This reason this number is so useful in measuring
your data throughput is that it gives a very accurate picture of the cost of
the bandwidth. The advantage to the customer is that you have the ability to
burst at speeds of up to 100Mbps, while only paying for your actual usage.
This means that in a 30-day month your top 36 hours of usage is thrown away.
We compile traffic data samples every 5 minutes which consists of the average usage of that 5 minute period in
bits per second value. At the end
of the month these samples are added up, and the top 5% thrown away. The
number that is after that is your 95th percentile and what you would be billed for. In a 30-day month you would have 8,640 5-minute samples, and
your top 432 samples would be thrown away. Below is an example of a 95th
percentile figure.

The above graph is shown as 2-hour averages. So as you can see this customer
reached peaks of over 2Mbps for at least 2 hours, and peaks higher than the
95th calculation of 1272Kbps for much longer. If the customer was setup with
flat-rate billing the customer would have had to either risk losing business
by denying, or slowing business during peak times, or, choose to pay for a
larger amount of bandwidth than that really needed. The 95th percentile calculation is a fair system where you pay for what you get. Another example
of the benefit of the 95th percentile would be if a customer runs an automated off-site backup very early in the morning when the bandwidth is
very low. However, this will cause a brief spurt in usage. If you were charged for actual bandwidth usage you would be charged a large amount
although the sustained data rate is very low.
|
|
|
|
|
|
|