Pronet Advertising

Learning from Google’s data centers

Neil Patel

Have you ever wondered how many data centers the search engine giant Google has, or why it is important to know each one? Currently Google has around 39 or more data centers. They can tell us a lot about how Google works; such as giving us an indication of when they are doing an algorithm update or when they are updating pagerank. If you are wondering how your website ranks among the datacenters, feel free to run searches on random data centers. If you know of any data centers that we might be missing, feel free to leave a comment.
Data Centers
1. [66.102.1.104](http://66.102.1.104)
2. [66.102.7.99](http://66.102.7.99), [66.102.7.104](http://66.102.7.104), [66.102.7.147](http://66.102.7.147)
3. [66.102.9.99](http://66.102.9.99), [66.102.9.104](http://66.102.9.104), [66.102.9.107](http://66.102.9.107), [66.102.9.147](http://66.102.9.147)
4. [66.102.11.99](http://66.102.11.99), [66.102.11.104](http://66.102.11.104), [66.102.11.107](http://66.102.11.107)
5. [216.239.37.99](http://216.239.37.99), [216.239.37.104](http://216.239.37.104), [216.239.37.107](http://216.239.37.107)
6. [216.239.39.99](http://216.239.39.99), [216.239.39.104](http://216.239.39.104), [216.239.39.107](http://216.239.39.107)
7. [216.239.51.99](http://216.239.51.99), [216.239.51.104](http://216.239.51.104), [216.239.51.107](http://216.239.51.107)
8. [216.239.53.99](http://216.239.53.99), [216.239.53.104](http://216.239.53.104), [216.239.53.107](http://216.239.53.107)
9. [216.239.57.99](http://216.239.57.99), [216.239.57.103](http://216.239.57.103), [216.239.57.104](http://216.239.57.104), [216.239.57.107](http://216.239.57.107), [216.239.57.147](http://216.239.57.147)
10. [216.239.59.99](http://216.239.59.99), [216.239.59.103](http://216.239.59.103), [216.239.59.104](http://216.239.59.104), [216.239.59.107](http://216.239.59.107), [216.239.59.147](http://216.239.59.147)
11. [216.239.63.99](http://216.239.63.99), [216.239.63.104](http://216.239.63.104)
12. [66.249.81.104](http://66.249.81.104)
13. [66.249.83.104](http://66.249.83.104)
14. [66.249.85.99](http://66.249.85.99), [66.249.85.104](http://66.249.85.104), [66.249.85.107](http://66.249.85.107)
15. [66.249.89.104](http://66.249.89.104)
16. [66.249.91.104](http://66.249.91.104)
17. [66.249.93.99](http://66.249.93.99), [66.249.93.104](http://66.249.93.104), [66.249.93.107](http://66.249.93.107)
18. [64.233.161.99](http://64.233.161.99), [64.233.161.104](http://64.233.161.104), [64.233.161.107](http://64.233.161.107), [64.233.161.147](http://64.233.161.147)
19. [64.233.163.104](http://64.233.163.104)
20. [64.233.167.99](http://64.233.167.99), [64.233.167.104](http://64.233.167.104), [64.233.167.147](http://64.233.167.147)
21. [64.233.169.104](http://64.233.169.104)
22. [64.233.171.99](http://64.233.171.99), [64.233.171.104](http://64.233.171.104), [64.233.171.107](http://64.233.171.107), [64.233.171.147](http://64.233.171.147)
23. [64.233.179.99](http://64.233.179.99), [64.233.179.104](http://64.233.179.104), [64.233.179.107](http://64.233.179.107)
24. [64.233.183.99](http://64.233.183.99), [64.233.183.104](http://64.233.183.104), [64.233.183.107](http://64.233.183.107)
25. [64.233.185.99](http://64.233.185.99), [64.233.185.104](http://64.233.185.104)
26. [64.233.187.99](http://64.233.187.99), [64.233.187.104](http://64.233.187.104), [64.233.187.107](http://64.233.187.107)
27. [64.233.189.104](http://64.233.189.104)
28. [72.14.203.99](http://72.14.203.99), [72.14.203.104](http://72.14.203.104), [72.14.203.107](http://72.14.203.107)
29. [72.14.205.99](http://72.14.205.99), [72.14.205.104](http://72.14.205.104), [72.14.205.107](http://72.14.205.107)
30. [72.14.207.99](http://72.14.207.99), [72.14.207.104](http://72.14.207.104), [72.14.207.107](http://72.14.207.107)
31. [72.14.209.99](http://72.14.209.99), [72.14.209.104](http://72.14.209.104), [72.14.209.107](http://72.14.209.107)
32. [72.14.211.104](http://72.14.211.104)
33. [72.14.215.104](http://72.14.215.104)
34. [71.14.217.104](http://71.14.217.104)
35. [72.14.219.104](http://72.14.219.104)
36. [72.14.221.104](http://72.14.221.104)
37. [71.14.223.104](http://71.14.223.104)
38. [72.14.235.104](http://72.14.235.104)
39. [71.14.253.104](http://71.14.253.104)

0 Responses to “Learning from Google’s data centers”

  1. great stuff … i was finding this to help me with my stats and other stuff

  2. Harris says:

    So how can you say that an IP address equates to a entire data center?

  3. Matt says:

    These are just different subnets. Large companies such as Amazon.com have several subnets within a single datacenter.
    There’s no reason for Google to have 39 datacenters.

  4. shmapty says:

    The inherent problem with your list is that you are using anecdotal proof that these are indeed separate data centers.
    You are better off by actually figuring out how these IP addresses are being routed on the Internet. If so, you’d find that your “Data Center #1″ is better described by perhaps 66.102.0.0/23 and #2 by 66.102.6.0/23, etc.
    If you did that, it would become clear that you are “missing” four /24 ranges in between, considering that Google owns the entire /20 block of 66.120.0.0, as well as several others.
    By tracking any full routing table on the Internet, you’d see that 66.102.0.0 takes a different path than 66.102.2.0, which takes a different path than 66.102.4.0, etc, etc.
    But, the larger flaw is assuming that these are different data centers at all and are not simply multiple paths via different carriers to the same dc.

  5. Tim says:

    Agree with Matt on this one.. what makes you think this is an entire data center?

  6. Neil Patel says:

    Thanks Kid, it looks like I missed a lot of IPs, but I think I got all of the C Blocks.

  7. Sam says:

    Each represented IP does not necessarly mean a server. This is a wrong statement to make, each server has an IP adress does this make it a physical site? NO….server can have numerous IP addresses and looking at the list I am able to say that google has numerous subnets…such as 1, 7, 9, 37….217, 219….etc. They have alot of machine under each subnets and they all are load balance and contribute to the overall seach!!!!

  8. Neil Patel says:

    Tim, my method was probably flawed and I should have used the method Shmapty mentioned.
    Thanks for the feedback.

  9. shmapty says:

    Neil. I didn’t check all the IP addrs you listed, but you are mistaken if you believe that you have “all of the C blocks.” All of the addrs that I tried from you list were in fact not “C blocks.”
    Also, another fatal flaw of your list is the concept of IP anycast. I don’t know if Google announces the same block from several dcs, but the point is that unless you peer with Google at every edge, you cannot know if this is not the case.

  10. Neil Patel says:

    Shmapty, I should have been clearer. I meant “I think I got all of the C Blocks listed on [Webmaster World](http://www.webmasterworld.com/forum30/34828.htm)”.
    Thanks for pointing out the flaws; they will help me from make the same mistakes in the future.

  11. shmapty says:

    Here is a simplified list, using your original 39:
    64.233.160.0/19 32 Cs
    66.102.0.0/20 16 Cs
    66.249.64.0/19 32 Cs
    72.14.192.0/18 64 Cs
    216.239.32.0/19 32 Cs
    possibly not Google
    71.14.217.0/24 1 C
    71.14.223.0/24 1 C
    71.14.253.0/24 1 C
    as an example of what’s being announced, here are the routes being advertised now for 64.233.160.0/19:
    64.233.160.0/23
    64.233.162.0/23
    64.233.164.0/23
    64.233.166.0/23
    64.233.168.0/23
    64.233.170.0/23
    64.233.172.0/24
    64.233.173.0/24
    64.233.176.0/23
    64.233.178.0/23
    64.233.180.0/23
    64.233.182.0/23
    64.233.184.0/23
    64.233.186.0/23
    64.233.188.0/23
    64.233.190.0/24
    64.233.191.0/24

  12. g1smd says:

    Matt Cutts has already confirmed that all IP addresses found on one C-block are typically located in one datacentre. It’s in Video #9.

    So, Google has UP TO 39 datacentres. They use 39 class-C IP blocks in total.

    Of course, several C-blocks might resolve to the same building; but I expect them to resolve to another bank of machines on another floor, or in another room.

    I would speculate that one IP address equates to something like 1000 machines, but that would be rampant guesswork.

Privacy Policy