For marketers
who love technology
Home » , , , » How many datacenters does Google use?

How many datacenters does Google use?

Google lists some datacenters on its website.


This list may be incomplete. For instance, given the attention Google pays to performance, it's strange to see such a big blank hole in Eurasia (Russia / China / India) and Africa.

So, how can we complement the map? Well it's easy: I have shown you how to identify Google's IP addresses thanks to BGP looking glasses. We just have to geolocate them with maxmind for instance. I've written a quick and dirty Python program to do that, and I have filtered the results located in Mountain view, as it's not a big surprise that Google hosts some servers there...

Here are the results
{   'AS15169': {   'Atlanta (US)',
                   'Berlin (DE)',
                   'Brussels (BE)',
                   'Buenos Aires (AR)',
                   'Charleston (US)',
                   'Council Bluffs (US)',
                   'Country: BELGIUM (BE)\nCity: Brussels\nIP: 74.125.122.120\n',
                   'Country: UNITED STATES (US)\nCity: Buffalo, NY\nIP: 1.1.1.120\n',
                   'Country: UNITED STATES (US)\nCity: Fairfax, VA\nIP: 1.2.3.120\n',
                   'Country: UNITED STATES (US)\nCity: Mountain View, CA\nIP: 74.125.63.120\n',
                   'Dublin (IE)',
                   'Frankfurt am Main (DE)',
                   'Groningen (NL)',
                   'Kuala Lumpur (MY)',
                   'Lappeenranta (FI)',
                   'Morganton (US)',
                   'Mountain View (US)',
                   'New York (US)',
                   'Schaumburg (US)',
                   'Taipei (TW)',
                   'Taoyuan City (TW)',
                   'The Dalles (US)',
                   'Tulsa (US)',
                   'Washington (US)',
                   'Zaventem (BE)'},
    'AS16591': {'Kansas City (US)', 'Stanford (US)', 'Mountain View (US)'},
    'AS19448': {'Kansas City (US)', 'Stanford (US)', 'Mountain View (US)'},
    'AS22577': {'Chesterfield (US)', 'Mountain View (US)'},
    'AS22859': set(),
    'AS24424': {'Beijing (CN)'},
    'AS36039': set(),
    'AS36040': {   'Country: UNITED STATES (US)\nCity: Aurora, CO\nIP: 63.117.14.120\n',
                   'Country: UNITED STATES (US)\nCity: Des Moines, IA\nIP: 12.216.80.120\n',
                   'Country: UNITED STATES (US)\nCity: Houston, TX\nIP: 63.117.68.120\n',
                   'Country: UNITED STATES (US)\nCity: Katy, TX\nIP: 63.88.73.120\n',
                   'Country: UNITED STATES (US)\nCity: Miami, FL\nIP: 65.196.188.120\n',
                   'Country: UNITED STATES (US)\nCity: New York, NY\nIP: 63.96.4.120\n',
                   'Country: UNITED STATES (US)\nCity: Winchester, VA\nIP: 63.110.67.120\n',
                   'Mountain View (US)',
                   'Palo Alto (US)',
                   'Warzachewka Polska (PL)'},
    'AS36384': set(),
    'AS36385': set(),
    'AS36492': {'Palo Alto (US)', 'Mountain View (US)'},
    'AS36561': set(),
    'AS36987': set(),
    'AS41264': {   'Country: UNITED STATES (US)\nCity: Macomb, MI\nIP: 209.119.80.120\n'},
    'AS43515': {'Mountain View (US)'},
    'AS45566': set()}

There is no guarantee that all these IPs are used for datacenters. But this list gives us a clearer idea of the location of AS15169 IP addresses, as long as the geolocation database can be trusted (they often contain mistakes). In particular, we find known datacenters (Council bluffs, the dalles, etc.) but also several other cities.

Note that for AS15169, my script was not able to geolocate many prefixes! Also do not forget that Google is likely to use third party datacenters in locations where the business case for building in house datacenters is unclear. Finally, be aware of the Google Global Cache (GGC) program through which Google manages to put its servers inside ISP networks. These servers bear an address from the ISP's ranges are are thus more difficult to track.

Enhanced by Zemanta
SHARE

About Gilles

0 comments :

Post a Comment