How big is the Internet?

Wed Aug 14 20:27:00 UTC 2013

On Aug 14, 2013, at 15:00 , Sean Donelan <sean at donelan.com> wrote:

> I should have remembered, NANOG prefers to correct things.  So here are
> several estimates about how much IP/Internet traffic is downloaded
> in a month.  Does anyone have better numbers, or better souces of
> numbers that can be shared?

I think you are not defining things precisely enough to be corrected. What does "downloaded" mean? For instance:

1) If a Google server pulls traffic from another Google server in another datacenter over the Google backbone, is that "downloaded"?
2) How about if an an Akamai server pulls traffic from another Akamai server in the same building but two different networks?
3) How about if the two servers are on the same switch?
4) What if I am playing X-Box with another user on Comcast on the same head end?
5) Two different head ends in the same city?
6) Different cities?

Etc.

It is actually even harder than the above illustrates. Most people define "Mbps on the Internet" as inter-AS bits. But then what about Akamai AANP nodes, Google GGC nodes, Netflix Open Connect nodes, etc.? They are all inside the AS. Given that Akamai claims to be 20% of all broadband traffic, Google is on the same order, and NF claims to be 30% of US peak-evening traffic, it seems like it would be foolish to ignore this traffic.

I could go on, but you get the point. Definitions are a bitch.

Once you define what you mean by "how bit is the Internet", I'll be happy to spout off about how big it is. :)

All that said: My back-of-the-envelope math says the Internet is order of 1 exabyte/day, as defined by my own rules on what counts as "the Internet"[*].  I could easily be wrong, but you asked.

-- 
TTFN,
patrick

[*] I count Company-to-Company traffic. This is _mostly_ inter-AS traffic, but on-net nodes (e.g. Akamai, Google, NF) -> Provider _do_ count. Things like Google -> Google over Google backbone do not count. Things like as701 -> as702 would count, but not as701 -> as701, even if the traffic is between two single-homed customers. It is a weird definition, but that's how I define it. (Although I may be biased, since counting only inter-AS traffic leaves off $SOME_PERCENTAGE of the traffic from my company.)

> Arbor/Merit/Michigan Internet Observatory: 9,000 PB/month (2009)
> Minnesota Internet Traffic Studies: 7,500-12,000 PB/month (2009)
> 
> Cisco Visual Network Index:
> 	Total IP: 55,553 PB/month (2013)
> 		Fixed IP: 39,295 PB/month (2013)
> 		Managed IP: 14,679 PB/month (2013)
> 		Mobile Data: 1,578 PB/month (2013)
> Telegeography via ITU report: 44,000 PB/month (2012)
> National Security Agency: 55,680 PB/month (2013)
> 
> 
> Individual providers/countries
> Australian Bureau of Statistics (AU only) : 184 PB/month (2012)
> AT&T Big Petabyte report (AT&T only): 990 PB/month (2013)
> CTIA mobile traffic (US only): 69 PB/month (2011)
> London School of Economics (Europe only): 3,600 PB/month (2012)
> TATA Communications: 1,600 PB/month (2013)
> 
> Historical:
> NSFNET: 0.015 PB/month (1994)
> 
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 495 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20130814/aebc57e1/attachment.sig>