Google causes 40% drop in traffic?

Given how much traffic these days is CDN and streaming, is that number
really supportable?

http://www.marketplace.org/topics/tech/down-goes-google-down-goes-internet

Cheers,
-- jra

http://www.seattleix.net/agg.htm
http://www.torix.ca/stats.php

Obviously not definitive, but I don't see a massive drop in traffic during the outage, at least on the two public exchanges listed above.

It would be interesting to see if there was a concurrent increase in traffic to other sites (Facebook, Yahoo, etc.) while people couldn't check their email.

Mike

In the interview they are saying that if Google is down, lots of people
don't have DNS anymore. So that accounts for an even larger drop than
just "no Youtube". Hmm - why would people use those resolvers, besides
being lazy in configuring a proper resolver-address.

Of course if Google is down we have no Google search (well, might be a
problem in some cases), no Gmail etc. (fine with me) and no Youtube
(hmm, but we'll survive without it). Come on ...

If the average user is *so* dependent on Google, we have an even larger
problem. Maybe like IPv6-day etc. lets try a "Google outage day" once a
year as a training :slight_smile:

Regards,
Stefan

It's just been pointed out to me that, even though Marketplace just posted
the link to that piece, the dateline is from August. My apologies for
not noticing.

Cheers,
-- jra

it's a bit perjorative to say 'lazy' isn't it? what if your ISP does
monkey business with nxdomain or other requests? would you like it
better if people used opendns? or 4.2.2.2? (a non-sla service from a
fourth party...)

I'm a fan of my own resolver, but not everyone has a dns resolver in
they back pocket, right?

From: "Stefan Neufeind" <nanog@stefan-neufeind.de>

> Given how much traffic these days is CDN and streaming, is that
> number really supportable?
>
> http://www.marketplace.org/topics/tech/down-goes-google-down-goes-internet

In the interview they are saying that if Google is down, lots of people
don't have DNS anymore. So that accounts for an even larger drop than
just "no Youtube". Hmm - why would people use those resolvers, besides
being lazy in configuring a proper resolver-address.

The amount of fun I had before I discovered "ipconfig /flushdns" suggests
to me that that's not really a supportable argument.

Of course if Google is down we have no Google search (well, might be a
problem in some cases), no Gmail etc. (fine with me) and no Youtube
(hmm, but we'll survive without it). Come on ...

If the average user is *so* dependent on Google, we have an even larger
problem. Maybe like IPv6-day etc. lets try a "Google outage day" once
a year as a training :slight_smile:

We just did; weren't you paying attention? :slight_smile:

Cheers,
-- jra

You sure? I think that there is an actual outage, unrelated to the
article you'd posted.

http://techcrunch.com/2014/01/24/gmail-glitch-is-causing-thousands-of-emails-to-be-sent-to-one-mans-hotmail-account/

Of course, I could be wrong. I've been wrong before.

A lot of people make value judgements on the relative likelyhood of finding
evil in DNS packets coming from 8.8.8.8 versus DNS packets coming from the
IP address handed to you in the DHCP reply....

Actual source

http://www.google.com/appsstatus#hl=en&v=status

-Gabe

There was an outage today, and that was why Marketplace linked their
piece on Facebook, which is where I saw, it, but the piece was not about
today's outage.

I was mislead, because there were 2 or 3 other pieces in the tech press
already in other places that *were* about today's outage.

Cheers,
-- jra

If it's just "some" DNS your provider hands out, I agree it's not much
better as well. (But you might possibly assume your provider has less
interst to spy on all your emails, your dns-queries and the like.)
What imho you'll want is a reliable resolver which is as close to you as
possible (and have it do DNSSEC-validation etc.).

Regards,
Stefan

From: "Stefan Neufeind" <nanog@stefan-neufeind.de>

If it's just "some" DNS your provider hands out, I agree it's not much
better as well. (But you might possibly assume your provider has less
interst to spy on all your emails, your dns-queries and the like.)

You might assume that, I wouldn't. If your access provider is a commercial
eyeball network like, say, Road Runner or Comcast, then there is, I believe,
evidence that they do DPI and possibly even ad injection, in addition to
playing NXDOMAIN games.

What imho you'll want is a reliable resolver which is as close to you
as possible (and have it do DNSSEC-validation etc.).

Sure; everyone should have their recursing resolver at the edge of their
network. But most consumers don't.

Cheers,
-- jra

There was a lot of discussion about this figure back in August when the
relevant outage occurred.

From memory, a large percentage of the traffic drop was from other sites

breaking as a result of Google not being available. ie, a site completely
unrelated to Google, potentially being served by a CDN, that was using
Google Analytics on every page could fail to load and/or load/render slower
as a result of the specific outage that Google had at the time. This
resulted in a traffic drop for far more traffic than just that sourced from
Google.

A non-trivial percentage of the Internet is in some way or other dependent
on things like Google Analytics/maps/etc, Facebook likes, Twitter recent
tweets, etc, such that if any of those services are not available the site
fails to load, either correctly or sometimes at all. The same is true in
many causes for javascript/etc libraries being loaded from 3rd party sites
like Google.

  Scott

interst to spy on all your emails, your dns-queries and the like.)

FUD much? have you read the public-dns ToS and privacy statements?

if you haven't you might want to:
<https://developers.google.com/speed/public-dns/privacy>

What imho you'll want is a reliable resolver which is as close to you as
possible (and have it do DNSSEC-validation etc.).

you might also want to read:
<http://googleonlinesecurity.blogspot.com/2013/03/google-public-dns-now-supports-dnssec.html>

and for some (quite a few) users of 8.8.8.8 it's faster and closer
than their ISP dns... which says something about the ISP provided DNS
server(s) I suppose :frowning: It's certainly not the answer for everyone, or
everything, but shrugging it off for FUD reasons is just silly and
makes little sense.

-chris

I wonder what percentage of large website operators whose site designs
have such external dependencies have had it occur to them to include those
external services in their monitoring systems?

Cheers,
-- jra

This presupposes that they actually have monitoring systems which actually perform useful checks. All too often, this isn't the case.

;>