I have been looking around. Different provider uses different billing method. I completely understand what 95th percentile is.
However I am still a little bit confused with the following terms.

1. When someone mention MRTG Average Based, Is that 50th percentile or half of the highest MRTG graph.

2. I think there are some providers look at the actual usage. How is it being calcualted?

3. Let say if bandwidth is charged at the same price, which method would be cheapest? Acutal usage or MRTG average based?

I'm not really going to answer your question... but I always choose Actual measurement, as it's the most accurate, and there's not arguing about it. It's also the probably the lower cost comparing Average and Actual. What do others think?

Alan

Personally I like acutal measurement. I never get billed with 95th percentile, but I heard it is around 2.5x the acutal usage. However, I am not sure if 50th percentile or average measurement would be cheaper or more expensive than acutal usage. That is my main question with this post.

Originally posted by Eiv

2. I think there are some providers look at the actual usage. How is it being calcualted?

3. Let say if bandwidth is charged at the same price, which method would be cheapest? Acutal usage or MRTG average based?

To calculate actual usage, providers count traffic from and to your server, then they divide it by the time range.

Definitely go with actual usage! It is definitely fairer both for you and your provider.

LOL, I certainly agree!

Is there a list anywhere that shows what method of calculating bandwidth a dedicated hosting company uses?

Alan

Originally posted by Eiv
I have been looking around. Different provider uses different billing method. I completely understand what 95th percentile is.
However I am still a little bit confused with the following terms.

1. When someone mention MRTG Average Based, Is that 50th percentile or half of the highest MRTG graph.
It's neither... because 50th percentile would be the median of the data set. The average usage figure it is actually the average of the data set (add up all the numbers and divide by the number that are in the set).

I.E. You have a set of numbers: { 1,3,5,6,8,9,10,10,15,45 }
To get what would be called "average usage" more or less, you simply add those numbers together and divide by 10 (because that's how many there are). In this case, the average is 11.2.

50th percentile would be the term that exists exactly in the middle of the set. So if this were a set of 9, it would be the 5th number. However, we have an even number of terms (10), so there is no "middle number". The middle point of the set exists exactly halfway between the 5th and 6th terms, so we simply average the 5th and 6th terms together (8 and 9) and get 8.5 as the 50th percentile, or median number.

In this case, the 50th percentile worked out smaller than the average. The reason being is because it is not affected by the large outlier to the right (the 45). However, the 95th percentile figure would have been higher than both of these numbers, so both methods are almost always going to get you a lower figure than 95th (which I believe is 30 in this case).

Unfortunately, I don't think you will find many providers measuring bandwidth in 50th percentile. That essentially means you could spike for more than two weeks of each month and not be charged for it. Good if you're the customer, bad if you're the company selling the service. ;)

2. I think there are some providers look at the actual usage. How is it being calcualted?

This method usually uses the same figures that MRTG uses, except instead of converting them into a "rate" figure (I.E. XX kbits/second), it just sores the data and at the end of the month, all of the numbers are added together and you get the "actual usage" for that port for the month.

3. Let say if bandwidth is charged at the same price, which method would be cheapest? Acutal usage or MRTG average based?
From my experience, they are usually very close. Unfortunately I don't have any data to back that up. I think you will find that the difference between these two methods is negligible when compared to the difference between either of them and 95th percentile-based billing.

That's my $0.02 anyway. Anyone feel free to let me know if I'm mistaken on anything I've said here.

Have a happy holidays.

Best Regards,
Matt Lightner
mlightner@site5.com