Talk:Burstable billing

Comment
I tried to extend a bit, clean up and also added the critics. I'll put it on my watchlist. --Illtillwillkillbill 17:23, 13 June 2006 (UTC)

critics
In this section the sentance "A good datacenter/upstream will allow you to upgrade your commitment and therefor benefit of other pricing options if they notice these type of peak patterns." seems to be against NPOV. I suggest removing it or strongly rephrasing.--Mattarata 16:31, 27 July 2006 (UTC)
 * I removed the section --Mattarata 02:52, 28 August 2006 (UTC)
 * Only saw your edit now (if this thingy would only generate emails ;-)), but why is this sentence against NPOV when it's simply the truth? Good DC will not take advantage of your sudden peak or rise in traffic, but will allow you to upgrade the commitment and stay within a better financial frame. --till 18:40, 5 October 2006 (UTC)
 * Claiming that something or someone is good or better than something else is a point of view, especially in this case. If you must present a proint of view in an article, then you need to present the opposing view and be totally netural on the whole situation. We are not here to pass judgement on business or industry practices, rather present facts as they are proven and agreed upon by the community.
 * Say you buy transit bandwidth from a major backbone such as AT&T or Verio. You sign a contract with them to buy bandwidth at a certain price. Sometimes they build tiers into the pricing and sometimes they don't. If they do not give you tier pricing, and simply sell you bandwidth at a fixed price for a fixed term, how are they a bad company when you start to use way more than what your originally predicted? By your logic, if they hold you to the terms in which you both agreed they are bad, but if they waive their contract and allow you to have the pricing changed then they are good? I would say that the customer that demands to have their contract changed when they start to use more than what they predicted might be the "bad" party where as the company that writes good fair legal contracts and then holds customers to those contracts might be the "good" party.
 * Now you could be referring to a company that will "upgrade" a commitment so that the customer pays more overall due to the commit but slightly less per Mbps/GB/whatever. This could be seen as good to the customer and also good to the company, but again its really POV and not for us to decide. --Mattarata 22:11, 5 October 2006 (UTC)

It seems that this page is very negative towards 95th percentile billing... Just my impression. One should remember that the physical infrastructure, the fiber optic lines and routers are priced on peak capacity. If I buy some fancy new cisco that can handle 10GiB/sec of traffic but I only run 10MiB/sec through it except on mondays, when I run my backups and use all 10GiB/sec, I don't get a discount on the router. Same goes for the fiber the data flows over. 95th percentile billing is an attempt to fairly price that in without the problems associated with putting a hard cap on someone's bandwidth. It's not always the best choice for all traffic patterns, but if you are buying bandwidth as an ISP, it is usually the cheapest option. (again, my opinion, as someone who does buy bandwidth as an ISP, thus I'm writing on the talk page rather than on the article itself.)LukeCrawford (talk) 10:42, 8 October 2009 (UTC)

Weasel Words
I am going to remove the weasel words template. Could someone post some more details of their complaint here around weasel words. Unless it is very serious why not just discuss it on the talk page rather then marking up the article itself? Augustz 16:57, 15 April 2007 (UTC)

Average = Traffic
The Sentence "Critics of the 95th percentile billing method usually advocate the use of a flat rate system, billing per byte of data transferred, or using the average throughput rather than the 95th percentile." suggests, that "billing per byte of data transferred" and "using the average throughput rather than the 95th percentile" are different things. They are not, right? --84.61.140.24 03:29, 2 December 2007 (UTC)

No, they are in fact different things. Most hosting plans targeted at small users charge per byte of data transfered (usually per gigabyte of data transfered) amazon ec2 does this, as does my own company, and as do most web hosting providers. But in this case, if on monday, you transfer 20 gigabytes of backups to your server in the first hour of the first day of the month, (this would require a speedy connection)  then transfer no other data for the month, you are charged for 20 gigabytes of data. (in the case of a company charging $0.10 per gigabyte transfered, that would be $2.) if you billed on average transfer, that would be, uh, I"m too tired to do the math right now, so I should shut up, but 20 gigabytes worth of bits divided by the number of seconds in a month, which would turn out to be a rather small number of bits per second.

Note that unlike hard drives, bandwidth is still generally sold in MiB rather than MB. LukeCrawford (talk) 11:14, 8 October 2009 (UTC)

They are the same thing, although the capacity would be expressed according to the billing period rather than per second (e.g. as 3TB/month rather than as 10Mbps)

In the example above, this would be 64kbps vs 20GB/month

87.194.211.97 (talk) 19:21, 6 February 2010 (UTC)

dur, uh, you are correct, and I am wrong. LukeCrawford (talk) 10:42, 26 June 2010 (UTC)

Problem in understanding
So the example graph has a max Inbound of 6,28 Mbit/s. 95% of that would be 5,966 Mbit/s, but the 95e line is at 2,62 Mbit/s? -- Thomas —Preceding unsigned comment added by 195.37.120.44 (talk) 10:54, 5 August 2008 (UTC)

Problem in understanding explained
In response to the above poster. It is NOT 95 percent of your max usage, it is a 95th percentile of all the samples collected for the timeframe in questions. So, lets say you want the 95th percentile for a 100 minute timeframe (1 hour and 40 minutes). With a 5 minute sample interval you would collect 20 samples during this time. Here the fictitious values you got for those 20 samples:

The above is just an example. A standard month has 8640 samples if taken every 5 minutes for 30 days. Here the results with the different approaches:

The 95% percent of the max usage wouldn't be fair for a normal business customers who has a large spike. The 95% of the average usage isn't fair for a ISP when the customer is a normal business customer with a bell shape usage during the day and next to nothing during the night. The 95th percentile is a middle ground, removing spikes in the usage but gets closer to what the customer is actually using on a regular basis. —Preceding unsigned comment added by 205.196.176.95 (talk) 21:19, 27 April 2009 (UTC)

Error in "Criticism for end user billing"
The "Criticism for end user billing" section includes: If a website used 2Mbps for 5% of the time and not a single byte for the remainder of the month, the site would be billed as if they had sent 2Mbps for the entire month. It would then be actually cheaper to continually send data in hopes of reducing the 95th percentile (where 95% of the peaks fall).

Which is simply incorrect.

It would be true to say that the site would be billed the same as if they were to use 2Mbps for the entire month.

87.194.211.97 (talk) 19:15, 6 February 2010 (UTC)

Article has no longer has any valid (inline) references
There was only one reference, last retrieved in 2006, to an ISP industry magazine (I'm not sure, as it doesn't exist online anymore). Anyway, that is now a victim of link rot.

This article needs some attention, particularly without any references. It is being used for citation purposes in research, which is how I found it. That makes me uneasy, in the sense that it either needs to have at least one valid reference, or be merged, or deleted.

It isn't Wikipedia's fault that people are submitting research papers to the ACM using Wikipedia articles for references, I am not saying that at all! They should have checked the references in the article just like I did, as it was published in 2012, and the single valid reference was retrieved back in 2006. But I mention this to make the point that we need to be correct, although that would be true regardless of whether or not ACM papers relied on this article. (I am referring to this research paper: TailGate: handling long-tail content with a little help from friends, see reference #4, in WWW '12 Proceedings of the 21st international conference on World Wide Web, ACM, April 2012. ISBN: 978-1-4503-1229-5). --FeralOink (talk) 18:58, 19 April 2012 (UTC)

remove references to software packages?
I am not sure that providing links to (an incomplete) list of software packages that provide 95th percentile counting is useful. RichardLetts (talk) 15:57, 20 June 2013 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified 1 one external link on Burstable billing. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20080720163115/http://www.isp-planet.com/news/2006/cogent_price_drop.html to http://www.isp-planet.com/news/2006/cogent_price_drop.html

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at ).

Cheers.— InternetArchiveBot  (Report bug) 03:59, 11 November 2016 (UTC)