[ih] Total Internet traffic [was: Ingrid Burrington on North Virginia]

[Patrick W. Gilmore]

On Jan 11, 2016, at 2:40 PM, Jack Haverty <jack at 3kitty.org> wrote:
> 
>> the text includes:
>> 	Today, up to 70 percent of Internet traffic worldwide travels through this region
> 
> I don't understand how anyone can make any quantitative statement about
> worldwide characteristics of the Internet.
> 
> I'm curious how people actually measure "Internet traffic worldwide" in
> order to be able to draw such conclusions.
> 
> If you accept Vint Cerf's definition of the Internet - paraphrasing as
> "communications between devices using TCP/IP" - how does someone measure
> that traffic?   For example, I have lots of devices on my own LANs which
> send terabytes of information around the house over TCP. I bet you do
> too.  And the company you work for.  Who's measuring all that
> traffic...?   And how are they doing it?
> 
> Same question about other "worldwide" statistics, like number of
> attached computers, number of users, etc.
> 
> Yes, it is sad that marketing-generated "factoids" like "up to xxxx" is
> so easily interpreted as hard facts.
> 
> Back in the 80s when the Internet was young, we didn't have the
> capability to take such measurements with any confidence of accuracy or
> completeness.  When did that change.....?

Not much. Just got harder. But what hasn’t changed is the fact some people are willing to make shit up. And many others can’t tell when someone is making up stuff.

That said, there are measurements which can be made, and which are actually useful.

For instance, if you sample a large enough set of “broadband” ISPs in certain countries, you can find that the vast majority of traffic going down DSL/Cable modems come from 3 companies (Google, Netflix, Akamai). Is that “all” traffic? Of course not, but it is a valid, useful statistic.

What tweaks my brain is whether Google sending traffic from a Google-owned machine in DC1 to a Google-owned machine in DC2 over Google-owned fiber counts as “Internet traffic”? Does it matter if those machines have globally unique, publicly accessible addresses? Suppose the machines have 1918 addresses, but the link between the two DCs is on the “public” Internet? I could go on, but you get the point.

Lather, rinse, repeat for all the less obvious but similar companies / situations.

Which is why I prefer to look at things like total traffic to eyeball / broadband / access networks. (Or maybe just ’cause that’s what I used to do for a living. :)

-- 
TTFN,
patrick