Submit Your Article Forum Rules

Results 1 to 5 of 5

Thread: Google Datacenters Explained

  1. #1
    Senior Member Garrett's Avatar
    Join Date
    Jun 2003
    Posts
    316

    Google Datacenters Explained

    I heard that Google changed its datacenters recently from this post in WebProWorld that described a datacenter update. Of course, this led me to ask what exactly a datacenter was and what datacenters have to do with optimization.

    I wrote to several search engine experts to get their explanations of the datacenters, and some discussion regarding their relevance to you.

    What Is A Google Datacenter? Here is what a Google datacenter looks like: http://216.239.39.98/ (there are many of these), and here's a great definition from Daniel Brandt: "As I understand it, the data centers are different copies of the Google database located in various parts of the U.S. and Europe, and perhaps India.

    "They are used for load balancing and perhaps now for geo-located SERPs. Other things being equal, you probably get connected to the data center closest to you."

    Datacenters and Updates. Andy Beal describes the Google update and how it relates to the datacenters:

    "Before Google updates its main index, www.google.com, it makes changes at its various datacenters first. This allows Google to test the new results across smaller user samples and tweak them before making an update available to everyone. This process usually happens over 2-4 days."

    Why Watch The Datacenters? Since there are so many people who watch datacenters closely I asked why. Here's what Daniel said:

    "They are watched closely because in the old days, which was a year ago, the Google data centers were more predictable. The "Google dance" would begin with new backlinks on www2.google.com and www3.google.com, and then propagate to the other data centers over the next few days. The new backlinks were an indication that the rankings had been recalculated for the entire index, which used to happen almost like clockwork, once a month."

    "Once the new data appeared on www2 or www3, it would be stable. Therefore, if you watched these data centers, you had advance notice of the new index, and an indication of how your site would do over the next month until the next update."

    So people watched the datacenters to have advance notice on how Google had changed their algorithm. Or, they used to. As Daniel said, "in the last few months, Google hasn't been predictable at all."

    Dan Thies was not so impressed with the Datacenter analysts:

    "When changes happen at one of the data centers, people with nothing better to do will post hundreds of messages at Webmaster World about it.

    "Then they'll all switch the IP address in their hosts file to see if their PageRank went up or down. Then they'll all query the specific data center to see how they're doing on backlinks. Then they all panic and say it's the end of the world."

    Jim Hedger found a bit more value in watching the datacenters, though perhaps not too much more than Dan: "I don't really spend much time checking the various datacenters as the SERPs are all the client really cares about. The only time they are really interesting is times like this when we are trying to figure out what the heck Google is up to. My trolling has produced no solid leads though."

    Thanks to:
    Daniel Brandt
    Google-Watch.org
    Andy Beal
    WebSourced.com
    Dan Thies
    SEOResearchLabs.com
    Jim Hedger
    StepForth
    Garrett French
    Editor, WebProNews.com
    Hidden Content

  2. #2
    Senior Member Garrett's Avatar
    Join Date
    Jun 2003
    Posts
    316

    Google Datacenters

    From Andy Beal:
    Before Google updates its main index, www.google.com, it makes changes at its various datacenters first. This allows Google to test the new results across smaller user samples and tweak them before making an update available to everyone. This process usually happens over 2-4 days.

    For example, when Google "dances", it is merely changing the indexed results at the datacenters one at a time. Because a visitor may get their Google.com results from various datacenters they may one minute see results from "www-cw" then the next from "www-ex". As Google is updating all of its datacenters, you can see results from the "new" and "old" index until all datacenters have been updated and the "dance" is complete.

    Monitoring the datacenters is useful for detecting a dance/update, but has little use for actually optimizing your web site. You are merely catching glimpses of a Google update.

    I hope that helps.

    Regards,
    Andy Beal
    Vice President Search Marketing
    WebSourced, Inc
    Garrett French
    Editor, WebProNews.com
    Hidden Content

  3. #3
    Senior Member Garrett's Avatar
    Join Date
    Jun 2003
    Posts
    316

    Daniel Brandt on Datacenters

    From Daniel Brandt:
    As I understand it, the data centers are different copies of the Google database located in various parts of the U.S. and Europe, and perhaps India. They are used for load balancing, and perhaps now for geo-located SERPs. Other things being equal, you probably get connected to the data center closest to you. I know that from the Chicago area, www.google.com resolves to nine different IP numbers, apparently depending on load conditions. Once you talk about data centers, you have to add that the mapping of any data center to IP number is quite dynamic, which makes it difficult to even discuss data centers. I've noticed that more data center watchers are now using the IP number rather the subdomain when they're discussing data centers at all.

    They are watched closely because in the old days, which was a year ago, the Google data centers were more predictable. The "Google dance" would begin with new backlinks on www2.google.com and www3.google.com, and then propagate to the other data centers over the next few days. The new backlinks were an indication that the rankings had been recalculated for the entire index, which used to happen almost like clockwork, once a month. (For a long time, it appeared to be synchronized with the appearance of the full moon, so that now I still cannot look at a full moon without getting reminded of Google!)

    Once the new data appeared on www2 or www3, it would be stable. Therefore, if you watched these data centers, you had advance notice of the new index, and an indication of how your site would do over the next month until the next update.

    In the last few months, Google hasn't been predictable at all:

    1) Google keeps adding new data centers. There might be as few as 10, or as many as two dozen by now.

    2) Even if you think there are a dozen or more data centers, if you resolve them to IP numbers, you might discover that half of them resolve to just a handful of different IP numbers. The resolution of domain to IP is very dynamic these days.

    3) The new index, when it does appear, frequently appears on some obscure data center now rather than www2 or www3. It is rarely the same obscure center that it was the last time.

    4) Lately, a good number of the data centers haven't resolved at all, meaning that the DNS changed or they are off-line.

    5) The freshbot has, in the last six months in particular, caused the so-called "update" to be less definable. New pages will show up in the index and then drop out again, all before the so-called "update." For sites that are very dynamic, such as blogs and forums, it's almost meaningless to speak of an "update."

    6) The rankings of old and new ecommerce pages also get changed these days by a filter that appears to be separate from the main algorithm ranking. (That's how the Scroogle Scraper is able to work at www.google-watch.org/scraper.html ) This means that whatever Google is doing with the filter is at least as interesting as the update itself. The main updates are no longer one month apart, but closer to two months apart. The evidence of new backlinks and PageRank appear separately now (perhaps a few weeks later) than the evidence of an update itself.

    I think the main reason why the data centers are the object of so much attention is that there are a number of tools out there that compare the rankings across different data centers. One, for example, is at http://www.seochat.com/googledance They work by querying the selected data centers in their own inline frame, so that you can see them side by side. They keep the myth of the "Google dance" alive. Sure, Google still dances, but it's more like these days, Google is drunk all the time and dancing around the clock. I'm not sure these tools are perfoming much of a service anymore.

    I think a more interesting tool along these lines would be one that allows you to compare different Google results according to geolocation -- such as www.google.co.uk or www.google.de -- because there is quite a bit of evidence that results are skewed according to the target location. However, this gets difficult because it's clear that Google uses the IP number to determine your location to a large extent, and the question becomes one of whether your IP number would override the particular Google that you are searching.

    Also, I'm not sure how relevant the backlink count and PageRank are these days. If your backlinks are suddenly cut in half (as mine were cut in half by Google's count, a few weeks ago for www.namebase.org), and your PageRank drops from 7 to 6 (as mine did a few weeks ago), does this mean anything? My www.namebase.org referrals from Google, which are normally quite stable and based on large numbers (about 17,000 per day), have been inching upward in the last two weeks despite my drop in backlinks and PageRank. In my case, it obviously doesn't mean anything. It was Google's counting that changed for www.namebase.org -- nothing changed on the site itself, and no new backlinks were kicking in.

    If I dropped in PageRank from 7 to 5, or from 7 to 4, it might be worth worrying about. Also, the backlink count is very flaky. Google shows you the backlinks from the higher-PageRank sites, but is using more than these to calculate PageRank.

    Even the total number of pages that Google shows for your site is very flaky, if you have more than a few hunderd pages on the site. I cannot trust it within plus or minus 50 percent, and that's based on knowing exactly how many files are in each of 160 directories. Google used to be fairly accurate, but since last September I don't even look at the total counts because they're so screwy. They are meaningless as an indicator of how deep and thorough Google is when indexing my site.

    Daniel Brandt
    Google-Watch.org
    Garrett French
    Editor, WebProNews.com
    Hidden Content

  4. #4
    Senior Member Garrett's Avatar
    Join Date
    Jun 2003
    Posts
    316

    Dan Thies On Datacenters

    From Dan Thies:
    A data center is a place with lots of computers and lots of Internet connections. Google has multiple data centers in different locations to handle all of the search requests they get.

    People watch the data centers, because when Google updates their database, the results have to propagate out to the various data centers. A sudden change in results from one of them can signal an update.

    Real hard-core junkies also fiddle with their "hosts" file on their system to show a different IP address of the server that the Google toolbar queries to get a PageRank score to display. This allows them to control which data center their toolbar's PageRank display comes from.

    When changes happen at one of the data centers, people with nothing better to do will post hundreds of messages at Webmaster World about it.

    Then they'll all switch the IP address in their hosts file to see if their PageRank went up or down. Then they'll all query the specific data center to see how they're doing on backlinks. Then they all panic and say it's the end of the world.

    As for what more data centers might mean, they probably need more data centers to improve their users' experience. New data centers coming on line might mean that old data centers could be slated for removal. It doesn't matter anyway.

    Couple links for you:
    http://www.adaptability.co.uk/googlefacts.htm
    http://www.webmasterworld.com/forum3/7298.htm

    Dan Thies,
    SEO Research Labs
    http://www.seoresearchlabs.com
    Garrett French
    Editor, WebProNews.com
    Hidden Content

  5. #5
    Senior Member Garrett's Avatar
    Join Date
    Jun 2003
    Posts
    316

    Jim Hedger on Datacenters

    From Jim Hedger:
    Trolling Google's datacenters gives us several snippets of information when making predictions about client-site placements and about how Google is operating in general. An important point IMHO is that this information does not really make a huge difference in optimization, just in monitoring.

    First of all, checking the various datacenters can tell you when a Google-Dance is in progress. By looking at the number of "related sites" in the top right hand corner (Results 1 - 10 of about 1,260,000) and seeing if the numbers differ between the various data centers, we can tell when Google is updating itself.

    Next, the various datacenters can give us an indication (though very unscientific) of what will be happening in coming weeks. An older and (no longer valid) rule of thumb states that www.google.com is the current listings. www2.google.com and www3.google.com show what is coming up in the near-future. Accessing www2 and www3 used to make us look like psychics or sages (or both).

    Now, things get a bit strange when one considers that there is a co-relation between datacenter updates and the Hilltop, Florida, and (current) Austin updates. When the Hilltop update occurred, a new datacenter, (www-gv: 216.239.59.100) went up. (It has since gone down). When the Florida update occurred, www-sj.google.com went down and is still out of commission. At the same time, three new datacenters have been introduced since August, (around the time of Hilltop algo update), one of which are still online. That's the one I am going to be paying a lot of attention to in the coming days. (www-lm: http://66.102.9.100)

    All that said, I don't really spend much time checking the various datacenters as the SERPs are all the client really cares about. The only time they are really interesting is times like this when we are trying to figure out what the heck Google is up to. My trolling has produced no solid leads though.

    Jim Hedger
    StepForth
    Garrett French
    Editor, WebProNews.com
    Hidden Content

Similar Threads

  1. Difference in rankings when searching Google datacenters
    By wige in forum Google Discussion Forum
    Replies: 37
    Last Post: 01-14-2008, 05:45 PM
  2. Google Directory Datacenters
    By clasione in forum Google Discussion Forum
    Replies: 1
    Last Post: 06-16-2005, 12:15 PM
  3. Google datacenters
    By scisoft in forum Google Discussion Forum
    Replies: 0
    Last Post: 03-04-2004, 05:33 AM
  4. Google PR simply explained
    By CraigAllen in forum The Castle Breakroom (General: Any Topic)
    Replies: 4
    Last Post: 01-28-2004, 09:58 PM
  5. Datacenters...
    By Runawayrentals in forum Google Discussion Forum
    Replies: 2
    Last Post: 01-22-2004, 04:49 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •