dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
948
share rss forum feed


espaeth
Digital Plumber
Premium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2

Methodology problem

This report is only using data off public exchanges (mostly international) and university networks. It doesn't take into account private peering, which is by far the most prevalent method of interconnection for the Internet.

Run a few traceroutes to... well.. anywhere. How many times do you see a public exchange in between carrier hops? The interconnection method is to use carrier neutral meet-me facilities in strategic locations throughout the world; the capacity and link performance at these locations remains largely private, and hence cannot be factored into this report.

The other problem with weighting the data heavily with university network exchanges is that the overwhelming majority of university networks are governed by "fair use" policies that rate limit or otherwise restrict high-bandwidth applications like P2P, Skype (particularly when supernode election in an issue), and excessive FTP traffic. Of course traffic growth isn't going to be huge from these networks -- it's already throttled down more aggressively than any public ISP complained about in these forums!

Lazlow

join:2006-08-07
Saint Louis, MO
And why is it that they only use public data? Because the ISPs will not let their data out. Just look how long it took to get bell to release their data. Look the the BS story comcast came up with for the FCC request (oh it would take too many hours, funny how everybody else got it done). One of the rules the new FCC should initiate is transparency. Force the ISPs to show what their data load is. There is absolutely no security risk in doing this. The only risk is that people will be able to see how much BS has been spread and how much greed is actually going on.


espaeth
Digital Plumber
Premium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
said by Lazlow:

And why is it that they only use public data? Because the ISPs will not let their data out.
.. but the public data is a gross misrepresentation of the environment they are attempting to model. This is one of those cases where "any data is better than no data" certainly isn't true.

said by Lazlow:

Force the ISPs to show what their data load is. There is absolutely no security risk in doing this.
Publishing detailed statistics that showed usage in relation to capacity for specific providers would be extremely helpful to the purveyors of denial of service attacks. Watch all the smaller ISPs get taken out like the fat kid in dodge ball...

The only chance to gather that data would be to put together a study like the Internet Traffic Study in Japan where statistics from all the major players are collected and sufficiently scrubbed of identifiable information so the statistics could be made public. I'm sure if a University research body presented their case to the major ISPs and had strict non-disclosure contracts they could get this done in the US. This would be limited to only reporting on how the industry as a whole is doing on capacity though.

Lazlow

join:2006-08-07
Saint Louis, MO

2 edits
Its funny, in one line you are saying that it cannot be done(safely) and then a couple of lines down you admit that the Japanese are already doing it. To publish this data you do not need to provide attack addresses, I guess I thought that was obvious.

As far as the public data does; there is no real way to tell how valid it is UNTIL somebody (trustworthy) gets the ISP's data.


espaeth
Digital Plumber
Premium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
said by Lazlow:

Its funny, in one line you are saying that it cannot be done(safely) and then a couple of lines down you admit that the Japanese are already doing it. To publish this data you do not need to provide attack addresses, I guess I thought that was obvious.
You seemed to be implying the data should be supplied on an ISP by ISP basis. The data could be presented, but it would need be industry-wide and not company specific. (ie, grouping every cable operator in the country together and only reporting the summary results of the group)

Lazlow

join:2006-08-07
Saint Louis, MO
I agree that it should be industry wide. But each ISP will have to supply that data. There is just no way around that. I am not convinced of the necessity of reporting it as a group. I can see labeling is as ISP A, ISP B, etc. But if you pour all the data in a pot and mix it up, you are not going to be able to draw nearly as much information from the data.