| 9:45 pm on Mar 8, 2008 (gmt 0)|
People use several methods, but most of them are subject to error and might only be correct 95% of the time.
The best way I know of is to use Firefox with the ShowIP add-on. That will display, in real time, the IP address that served the page you are currently viewing.
However, you will find if you take that Google IP address and query it directly, the results may still vary from what you just saw. Between the data center and the final results on a Google domain, there are some extra steps in the process that are not all obvious. Some of those changes seems to handle geo-targetting but that's not the whole story. All we know for sure is that there is still some "secret sauce" being applied.
| 10:56 pm on Mar 8, 2008 (gmt 0)|
|How do i know which data centers are being used at each time? |
Go to google.
Type in your keyword.
Hover you mouse where it says "cached" and it will give you the DC that you are looking at.
| 11:09 pm on Mar 8, 2008 (gmt 0)|
That's not always the case. In fact, I'd go so far as to call it a commonly held myth. What you see when you hover is the IP address for the server that will deliver the cached page if you click. But that's not necessarily the IP address that just supplied the results.
I just did a search where the links for the cached pages are on 184.108.40.206, but the page itself was supplied by a combination of four different IP addresses: 220.127.116.11, 18.104.22.168, 22.214.171.124 and 126.96.36.199. Multiple IP addresses on the same C-block is almost always the case.
Another common approach is to ping google.com just before or after you make the query. But with Google's complex load-balancing, this also is not a sure-fire approach. Every request can possibly be routed differently.
[edited by: tedster at 11:22 pm (utc) on Mar. 8, 2008]
| 11:13 pm on Mar 8, 2008 (gmt 0)|
The ShowIP extension is about the only reliable way to discern this information. It tells you all about the current screen of information.
Most of the other methods make some other request, even moments later, and that will often get a different result. Cache URLs are especially dangerous.
There's also differences seen between the various results returned, from querying the bare IP address, the 'equivalent' GFE name, and the 'equivalent' xx-in-fnn name.
| 11:29 pm on Mar 8, 2008 (gmt 0)|
I now remember what I had been told before and that is what Lame Wolf quoted.
Now I have some more info, I do use firefox so I will use the add on.
| 2:08 am on Mar 9, 2008 (gmt 0)|
I just installed it, and it's showing the IP address as "188.8.131.52 +3"
What is the +3?
When I copy to clipboard, I get
| 2:42 am on Mar 9, 2008 (gmt 0)|
Different parts of the page are being supplied from different servers. There's the organic rankings and the Adwords for two. Not sure about the thrid, but I'm beginning to suspect that the description snippets are supplied independently - that is just a guess right now.
| 3:00 am on Mar 9, 2008 (gmt 0)|
it is my igoogle home page, with my webmasterworld google search news widget, (which is currently showing as unavailable), a finacial widget and my rss feed to my site.
| 8:42 am on Mar 9, 2008 (gmt 0)|
As far as I remember, you're getting results from one of them (notice that they are all on the same class C block) and the other three are there as fail-over in case the chosen one is non-responsive.
| 8:20 pm on Mar 9, 2008 (gmt 0)|
g1smd, thanks for pointing out the same class C block. Believe it or not, I used to be able to subnet networks with a given mask, but alas, with Cisco, if you don't use it, you loose it. And unfortunately I forgot more then I remember.
silly me for not noticing that earlier. :P
| 2:41 am on Mar 10, 2008 (gmt 0)|
|.. I'm beginning to suspect that the description snippets are supplied independently - |
that is just a guess right now.
I'm seeing strangely different snippets across multiple datacenters for some SERPs;
I'm thinking this is a pretty accurate guess, Tedster
| 6:34 pm on Mar 10, 2008 (gmt 0)|
Not meaning to hijack the thread, if i am say so, but my site ranked on the 3rd to 4 th page on google for a very competitive term for a couple of months, making steady progress to page 1. Suddenly, last night it jumped to page one, position 4 :). The next time I check its at position number 32 :(. The position varies from #32 to first page. Is their preferred data center which serves the UK? (my website is UK based)
I have checked at least 30 data centers all of which put the website in the first page :). Should I be happy or not?
| 6:53 pm on Mar 10, 2008 (gmt 0)|
| 6:59 pm on Mar 10, 2008 (gmt 0)|
| 6:59 pm on Mar 10, 2008 (gmt 0)|
Welcome to the forums, gingerman. Google's infrastructure does not seem to work simply by assigning a data center to a country-specific Google. In fact, you can use Firefox to see the exact data center that just responded on a google domain name, then query that data center directly by IP address, and still get different results.
See this thread for more information:
| 9:41 pm on Mar 10, 2008 (gmt 0)|
Going on about six months now we have seen a good deal of real movement every couple of days, primarily on the first few of pages. This initially led us to what the OP is asking about, variation between data centers, but for us things syncs up quite well when looking across data centers. So observing placement by data center hasn’t really shown much differential, or insight into these changes, because data center identification does seem so elusive, and if you wait a bit, it changes on a particular DC you think you see.
Why these rapid, but small changes are occurring on such a regular basis is very curious. Never thought I would enter this camp, but am beginning to think they are in fact tracking click through rates, bounce factors and general overall user experience and in order to do that effectively different sites get “placed” in different positions. ("placed" is an outrageous word I know, but even done algorithmically there has to be some sort of approach on getting sites in different positions in an orderly testable fashion?)
Any one else seeing this steady churning?
| 9:47 pm on Mar 10, 2008 (gmt 0)|
There are some observations about new sites in the SERPs that line up with your idea about churn - actually"rotation" in this case. That's being discussed in the thread March 2008 Google SERP Changes [webmasterworld.com].
| 10:28 am on Mar 11, 2008 (gmt 0)|
There certainly has been a churn/cull. One of the terms which I am optimizing for gave a return of 100,000,000 pages last week. Today it only returns 29 million pages, a spring clean from Google.
[edited by: tedster at 4:59 pm (utc) on Mar. 11, 2008]