| Having been in this industry for as long as I have, | | | | searching from San |
| I often forget some of the basics. Well, it's not | | | | Francisco will likely be served their search results |
| that Iforget, it's just that I assume that everyone | | | | from a data center near them, such as an |
| else in the industry has the same knowledge level | | | | Oregon or |
| as I do. | | | | California data center. |
| So when I get a "newbie" asking a question like | | | | It is because of the differences in these data |
| "Why are my Google results different than my | | | | centers that someone searching in New York will |
| clients" I have totake a step back and realize that | | | | see differentresults than someone in Los Angeles. |
| we aren't all equal. | | | | It is reasonable to think that each data center |
| Therefore in this article I'm going to answer this | | | | acts somewhat independently of the others. That |
| one seemingly simple question. While it may be | | | | means that theirupdate schedules are different as |
| simple tothose of us "in the know" this isn't | | | | are their crawling schedules. One can even |
| always the case with others. | | | | assume that the algorithm changeswhich affect |
| One of the scariest things you can do as a new | | | | the index happen at different times as well. |
| SEO is talk to a client. That's because you are | | | | This accounts for why there are differences in |
| always worriedthat they are going to ask you a | | | | search results. Because of Google's perpetually |
| question that's over your head. And while you can | | | | updating index,the results you see near you may |
| fake your way through somequestions when you | | | | be similar to results in other data centers but |
| are new, there are some that just stump you. | | | | ultimately different. |
| Of course even the greenest of most SEO's | | | | This could be due to how the crawlers retrieve |
| usually know more then their clients so you don't | | | | sites - a crawler may find a site closer to it more |
| often have to fakean answer. Of course I would | | | | easily andtherefore add it to the index sooner |
| never advocate faking it. When I don't know, I've | | | | than a geographically far away crawler would find |
| found the best thing you cansay is "I don't know, | | | | the same site. |
| let me find out and I'll get back to you." | | | | For example, a site in Vancouver, Canada may |
| And that's just what happened the other day. A | | | | appear in the California data center days before |
| new person came to me and said "I had a client | | | | the eastern datacenters. Because this site would |
| who saw different | | | | be added to the west data center sooner, it will |
| Google results than I did, and I didn't know what | | | | have an impact on the searchresults returned |
| to tell them." | | | | sooner. |
| So let me give you the answer I gave him. | | | | Consider it like the ripples you see in the water |
| Google, like all the other engines, is comprised of | | | | when you drop a rock into a pond. If you drop |
| literally tens of thousands of servers. Each server | | | | just one rock,you see the ripples move out from |
| is partof a larger cluster of computers. Each | | | | where you dropped it. However, if you drop 2 |
| cluster forms part of a datacenter. Each | | | | rocks close to each other and atslightly different |
| datacenter then acts as anindependent branch of | | | | times you see how the ripples interact with each |
| Google. | | | | other when they meet. |
| These datacenters are found all over the US. For | | | | The index changes reflect this type of interaction. |
| the longest time Google only had about 13 data | | | | One site can have a subtle but noticeable effect |
| centers thatserved all the results to the world. | | | | on theindex. Yet the effects aren't noticed across |
| Now the number is estimated at around 80. | | | | all data centers at the same time. We can also |
| While some of these data centers are used for | | | | see the changes inthe index grow over time, so |
| pre-testing results (for example, testing a new | | | | that one Vancouver site's effect grows over |
| algorithm outbefore moving it to the main data | | | | time, but the effect is differentacross the data |
| centers) most are used just to deal with the load | | | | centers because changes happening with other |
| that Google receives everyday. | | | | sites also has an impact. |
| These data centers are dispersed throughout the | | | | As you can see, this is why you will see different |
| US in geographically specific areas. They have | | | | results across the data centers. It's not |
| done this sothat queries are served to the data | | | | necessarilybecause of one single event. Like SEO |
| center nearest to the user. | | | | itself, it's a culmination of smaller events which |
| For example, while there are a large number of | | | | causes the noticeabledifferences. |
| data centers on the eastern coast, a person | | | | |