So today, I saw this:
BlackBox:~ jlixfeld$ host google.ca 8.8.8.8
Using domain server:
Name: 8.8.8.8
Address: 8.8.8.8#53
Aliases:
google.ca has address 206.126.112.166
google.ca has address 206.126.112.177
google.ca has address 206.126.112.172
google.ca has address 206.126.112.187
google.ca has address 206.126.112.151
google.ca has address 206.126.112.158
google.ca has address 206.126.112.157
google.ca has address 206.126.112.173
google.ca has address 206.126.112.181
google.ca has address 206.126.112.155
google.ca has address 206.126.112.147
google.ca has address 206.126.112.185
google.ca has address 206.126.112.143
google.ca has address 206.126.112.170
google.ca has address 206.126.112.162
google.ca has IPv6 address 2607:f8b0:4006:808::100f
google.ca mail is handled by 50 alt4.aspmx.l.google.com.
google.ca mail is handled by 30 alt2.aspmx.l.google.com.
google.ca mail is handled by 20 alt1.aspmx.l.google.com.
google.ca mail is handled by 10 aspmx.l.google.com.
google.ca mail is handled by 40 alt3.aspmx.l.google.com.
BlackBox:~ jlixfeld$
That is not Google IPv4 address space, and those IPv4 IPs are not being announced by 15169.
Am I dumb in thinking that this is weird or is this sort of thing commonplace?
Local Google caches at QIX?
-- Stephen
^ my thought, they're on the QIX public block
Those IPs appear to be used by to Google cache servers at the QIX. It's
common for CDNs to utilize provider space and not maintain their own
layer-3. E.g. cache servers connected to switch, connected to provider,
without the requirement of a router.
/Steve
Seems like an odd waste of resources; what if Google, Akamai, Netflix, and
anyone else who wanted caches wanted IPs in that block? The IX would be out
of address space pretty quickly, forcing a majority of users to re-number
because of a small number of other users.
-Dave
I'm surprised they don't set aside a small piece of their IP space that an
ISP can anycast routes locally to the cache (Maybe a /24?)
one of the legacy uses of an IX was to place “content” near the eyeballs. For the adventurous, this meant placing NTP chimers, DNS & route servers,
and even content directly on the switch mesh. Akamai might have been the first to pull back from that and move its services behind an Akamai router
that was on the switch mesh.
These days, most folks use BGP “condoms” to protect themselves.
manning
bmanning@karoshi.com
PO Box 12317
Marina del Rey, CA 90295
310.322.8102
I'm surprised they don't set aside a small piece of their IP space that an
'they' and 'their' here are confusing, which 'they' and 'their' did you mean?
ISP can anycast routes locally to the cache (Maybe a /24?)
that sounds like a recipe for disaster with respect to latency and
jitter and control of destination of the user request, eh? Have you
tried using 6to4 gateways?
In this case they being Google and their being Google's IP space (but you
can replace Google w/ any provider using caching servers).
When I was using Comcast 6to4 gateway, the latency wasn't too bad (I'm a
Comcast customer). If the anycast is being broadcast from a regional
datacenter and your ISP has good connectivity, the latency shouldn't be bad
(in the QIX case above, the latency shouldn't be any worse than QIX's IPs
unless someone started advertising the same anycast subnet w/ a lower cost).
> I'm surprised they don't set aside a small piece of their IP space that
an
'they' and 'their' here are confusing, which 'they' and 'their' did you
mean?
In this case they being Google and their being Google's IP space (but you
can replace Google w/ any provider using caching servers).
this has the same uncertainty problems with other folk I imagine.
When I was using Comcast 6to4 gateway, the latency wasn't too bad (I'm a
Comcast customer). If the anycast is being broadcast from a regional
datacenter and your ISP has good connectivity, the latency shouldn't be bad
(in the QIX case above, the latency shouldn't be any worse than QIX's IPs
unless someone started advertising the same anycast subnet w/ a lower cost).
there are a lot of ifs there... and failover is where the problems arise 
Hi Jason,
This is not commonplace. That prefix is from a specially designated IXP
micro allocation block. See http://bit.ly/1OEcHde for detail. The use of
these specially designated blocks is for IXPs only.
We (Akamai) don't have equipment numbered into this type of address space
nor do we have any evidence that we have in the past. We certainly won't in
the future. If someone knows of anything that we missed, contact me
directly and we'll arrange to renumber.
Hope that helps.
Best,
-M<