On October 27, 2008, Windows Azure was unveiled publicly by Microsoft Chief Architect Ray Ozzie at Microsoft’s Professional Developer’s Conference.

Less than a year later, on November 17, 2009, Windows Azure was unleashed on the world – anyone could go create an account.

Only a couple of months later, on January 1, 2010, Windows Azure turned on its Service Level Agreement (SLA) – you could now get production support.

And finally, on Feb 1, 2010, Windows Azure became self-aware – billing was turned on, completing the last step in them being fully open as a business.

That was 4 years ago today. Happy Anniversary Azure! I am not calling this a “birthday” since it isn’t – it was born years earlier as the Red Dog project – but this is the fourth anniversary of it being a fully-operational, pay-as-you-go, public cloud platform.

At the time, there were 6 Windows Azure data centers available – 2 each in Asia, Europe, and North America: East Asia, SE Asia, North Europe, West Europe, North Central US, South Central US. (Ignoring the Content Delivery Network (CDN) nodes which I plan to cover another time.)

What about today? With the addition in 2012 of East US and West US data centers, today there are 8 total production data centers, but more on the way.

Here’s a map of the Windows Azure data center landscape. (Source data is in a JSON file in GitHub; pull requests with additions/corrections welcome. CDN data is TBD.)

The lines between data center regions represent failover relationships drawn from published geo-replication sites for Windows Azure Storage. Mostly they are bi-directional, except for Brazil which is one-directional; the metadata on each pushpin specifies its failover region explicitly.

NOTE: this is a work-in-progress that will be updated as “official” names are published for geos and regions.

Also, be sure to click on the map pushpins to see which data center regions are in production and where are coming attractions. Not all of these pushpins represent data centers you can access right now.

There are three insets in order – first a GeoJSON rendering, second a TopoJSON rendering (which should look identical to the GeoJSON one, but included for demonstration purposes, as it is lighter weight), and the third is the raw JSON data from which I am generating the GeoJSON and TopoJSON files. [All the code is here: https://github.com/codingoutloud/azuremap. I plan to blog in the future on how it works.]

The map data is derived from public (news releases and blog posts for coming data centers and Windows Azure documentation for existing production regions).

The city information for data centers is not always published, so what I’m using is a mix of data directly published, and information derived from published data. For example, it is well known there is a data center in Dublin, Ireland, but where city should I used for US West region that’s in California? For the latter, I used IP address geocoding of the published data center IP address ranges. This is absolutely not definitive, but just makes for a slightly nicer map. It was from this data that I made assumptions around Tokyo and Osaka locations in Japan and San Francisco in California for US West.

Finally, this map is at the region level which equates roughly to a city (see the project readme for terminology I am using). A region is not necessarily a single location, since there may well be multiple data centers per region and though they will be “near” each other, this is not necessarily in the same city – they could be 1 kilometer apart with a city border between them.