|New Google Datacentre - Sept 2006|
Is this the start of The Dalles operation?
| 7:57 pm on Sep 26, 2006 (gmt 0)|
Three brand new class-C IP blocks belonging to Google have recently come online serving search results.
Is this the start of operations at The Dalles perhaps?
|This thread follows on from the 2006 June thread at |
[webmasterworld.com...] which lists
the many other IP addresses already found.
[edited by: tedster at 7:54 pm (utc) on June 2, 2008]
| 8:00 pm on Sep 26, 2006 (gmt 0)|
I don't know the GFE datacentre names for any of these IPs yet.
These are all in Google's previously unused 18.104.22.168 - 22.214.171.124 block.
| 11:37 pm on Sep 26, 2006 (gmt 0)|
Oooops. How did I miss these before?
| 11:38 pm on Sep 26, 2006 (gmt 0)|
Here they are again, matched with their GFE names, where known:
gfe-ag 72.14.247.x x104 x99 x147
gfe-td 72.14.255.x x104 x99 x147 x103
gfe-fk 209.85.129.x x104 x99 x147
gfe-mu 209.85.135.x x104 x99 x147 x103
gfe-? 209.85.143.x x104 x99 x147
and again in a longer format:
126.96.36.199 - gfe-ag
188.8.131.52 - gfe-ag2
184.108.40.206 - gfe-ag3
220.127.116.11 - gfe-td
18.104.22.168 - gfe-td2
22.214.171.124 - gfe-td3
126.96.36.199 - gfe-td4
188.8.131.52 - gfe-fk
184.108.40.206 - gfe-fk2
220.127.116.11 - gfe-fk3
18.104.22.168 - gfe-mu
22.214.171.124 - gfe-mu2
126.96.36.199 - gfe-mu3
188.8.131.52 - gfe-mu4
184.108.40.206 - gfe-?
220.127.116.11 - gfe-?2
18.104.22.168 - gfe-?3
Phew! Any more?
| 10:38 am on Oct 7, 2006 (gmt 0)|
SearchMash.com currently resolves to all of these simultaneously:
| 12:22 pm on Oct 7, 2006 (gmt 0)|
g1smd, what tools do you use to monitor the search results on several data centers at once (besides accessing IPs via browser - I've also found an online tool), and what steps do you take to monitor keywords once you discovered they generate traffic? Some hints would be greatly appreciated, as I notice you always tend to be well documented.
| 1:17 pm on Oct 7, 2006 (gmt 0)|
I do a manual search at several gfe-xx.google.com datacentres, using searches stored in the browser bookmarks... with the search preferences set to &num=100 and &filter=0 most times too.
| 1:52 pm on Oct 7, 2006 (gmt 0)|
Are these DC's or IP's? I think it's important to note that DC's and IP's are completely different. Matt Cutts has said something to this effect in one of his videos. He also said it's "possible" a IP "may" pull results from different DC's... even on a refresh. Part of their load balancing I would imagine.
| 2:06 pm on Oct 7, 2006 (gmt 0)|
Google currently uses 44 different Class-C blocks of IP addresses.
On each Class-C block, there are several dozen IP addresses that are active. These usually end in 17, 18, 19, 44, 80, 81, 83, 84, 91, 93, 95, 98*, 99, 100, 101, 102, 103, 104, 105*, 106*, 107, 115, 133, 147, 184, 189, 210* and 214 (* Note: .98, .105, .106, & .210 seem to be no longer accessible). These show results from the "Google English" search screen.
Each Class-C block also has a two-letter GFE alias. These are always mapped in a specific way. For these, gfe-xx always corresponds to x.x.x.104, most blocks also have gfe-xx2 which is x.x.x.99 and a few also have gfe-xx3 which is x.x.x.147. Finally, and for a very few, there is gfe-xx4 which is always x.x.x.103. The search screen is simply announced as "Google" without the "English" logo.
See the previous thread for a list of GFE names mapped to Class-C IP addresses. For the moment, these are the currently known GFE names: gfe-ag, gfe-ar, gfe-au, gfe-bf, gfe-bp, gfe-bu, gfe-bx, gfe-cw, gfe-dc, gfe-ed, gfe-eh, gfe-ff, gfe-fg, gfe-fk, gfe-gv, gfe-he, gfe-hk, gfe-hs*, gfe-hu, gfe-ik, gfe-in, gfe-jc, gfe-jp, gfe-kc, gfe-kr, gfe-lm, gfe-lo*, gfe-mc, gfe-mu, gfe-nf, gfe-nz, gfe-od, gfe-po, gfe-py, gfe-qb, gfe-rn, gfe-ro, gfe-td, gfe-tw, gfe-ug, gfe-ui*, gfe-va, gfe-wr, gfe-wx, gfe-yo with an unknown name for 22.214.171.124 as yet.
- gfe-ui resolves to the same IP as gfe-hs.
- gfe-lo is not currently active.
Matt Cutts confirms that all the data at all of the IP addresses within one Class-C block should usually be the same, as typically they are all the same datacentre.
So, Google probably consists of 44 large banks, or groups, of PCs, but there may be multiple unconnected banks within one building.
I find it enough to check just one IP from a Class-C block, and to check a few different Class-C blocks. There is no need to check all 700+ active IP addresses.
| 10:40 pm on Oct 19, 2006 (gmt 0)|
Anyone find any more?
| 10:35 pm on Oct 22, 2006 (gmt 0)|
OK. Here is the next batch of Google IPs coming on stream:
| 10:36 pm on Oct 22, 2006 (gmt 0)|
I don't know any of the GFE names for these yet:
gfe-? - 209.85.133.x - x104 x99 x147
gfe-? - 209.85.139.x - x104 x99 x147
... and here is the same list in a different format:
126.96.36.199 - gfe-?
188.8.131.52 - gfe-?2
184.108.40.206 - gfe-?3
220.127.116.11 - gfe-?
18.104.22.168 - gfe-?2
22.214.171.124 - gfe-?3
These have come online for the first time in only the last few weeks.
| 3:11 pm on Oct 23, 2006 (gmt 0)|
Thanks for the "heads up" and the info! I added the new datacenters this morning.
I tried to reply to your stickymail BUT your mailbox is full so it wouldn't send...
[edited by: tedster at 3:28 pm (utc) on Oct. 23, 2006]
| 12:42 am on Nov 3, 2006 (gmt 0)|
And there is more:
These have come online within the last two weeks.
| 12:43 am on Nov 3, 2006 (gmt 0)|
The GFE name is as yet unknown:
gfe-? - 209.85.155.x - x104 x99 x147 x103
and here is that data in another format:
126.96.36.199 - gfe-?
188.8.131.52 - gfe-?2
184.108.40.206 - gfe-?3
220.127.116.11 - gfe-?4
| 12:55 am on Nov 3, 2006 (gmt 0)|
hmm could this mean they now have more space, so they dont have to list all as supplemental and when searching getting the omitted results in the face after 3 pages.
| 2:52 am on Nov 3, 2006 (gmt 0)|
These certainly are not (all) in Oregon.
I just ran traceroutes on a couple of random ones. One was in Germany, another in Japan.
You know, you can get a lot of information out of traceroutes and ARIN IP block lookups. If you are interested in where these are, I suggest you use traceroute. I don't have either the interest or the patience to run the whole list. :)
I don't think there is any particular consistency to this list. Some are apparently in the U.S., some are in various parts of the world.
The 72... addresses are rather interesting, as they are accessed through quite a number of router hops on the same 72... network, none of them returning reserve DNS. The 72... addresses all belong to Google. May be part of the network Google is reputed to be building.
Take this one: 18.104.22.168. Tracing from home, from a Cox Communications connection, the reverse-DNS stops as soon as it leaves San Diego. There are three hops that are obviously Cox, but unnamed - which is unusual for Cox. But an ARIN lookup on the IP block reveals that Cox hand-off directly to Google in Atlanta. (Cox is pretty good about identifying their geography in their IP block data.)
Same address, traced from Dallas: Level3 to Seattle, peered with Google probably in Seattle, then another 8 mSec and 4 hops on the 72... network. So, this one could be in Oregon.
22.214.171.124 - hands-off to Savvis in Palo Alto. Savvis to SF, LA, Dallas, Atlanta, Washington, Paris, Frankfort. Then goes a few hops of unnamed 72... Google-owned IPs. That's an ugly route! From Dallas, I get Savvis with an even uglier route, going back to L.A., Washington, etc.
A few hints if you want to pursue this - carriers often use airport codes as part of domain names to identify the locations of their routers. Some carriers will identify locations as part of ARIN block names. Cox is good at this, for example - each class C is named "city-abbreviation-block" or some-such. Google seems to do neigher, though.
OK, one more: 126.96.36.199. Cox to L.A., hand-off to Equinix. A hop across the Pacific to Asia Netcom. Not enough info in the DNS or ARIN lookup to tell where, but according to their coverage map, the only connection from L.A. is to Osaka. They obviously hand-off to Google in Osaka, because it is the same time (127 mSec) to the first Google hop. Not even enough time to make it to Tokyo.
This one is pretty intesting from Dallas - SprintLink from Dallas to Ft. Worth, Kansas City, Chicago. Hands-off to Google in Chicago. Takes an all-Google route the rest of the way to Japan. Can't tell exactly how, but it's fairly direct. First hop is 3mSec from Chicago, then next hop is 100mSec. Another 30-msec hop, then finally 8mSec. So, they have a direct link to somewhere in Asia outside of Japan from a point 3mSec from Chicago, then a link from there to Tokyo, then the final 8mSec to Osaka. Just a guess. :)