From our side of the fence, cloud computing is a wonderful paradigm. We have all this instant-on power available to us, spinning up instances left right and centre without consideration for the actual underlying hardware.
Have you ever considered the actual hardware underneath to allow us to consume 1000 instances at a time without even getting a phone call from the cloud provider asking "what on earth are you up to?". Even the virtual world has to touch reality at some point.
For example, just how much disk space does Amazon have on tap for S3? How much does Nirvanix have? How many servers can GoGrid scale up to be they run out of kit? All fascinating questions, and all closely guarded secrets.
But I discovered this wonderful piece from Data Center Knowledge who listed, as far as they could tell, the number of physical servers each of the big providers where managing. Some interesting highlights:
- Amazon: $86M spent on servers in 2008; more than Microsoft
- RackSpace / Mosso: has over 50,000
- Google: over 450,000 servers
- Microsoft: has 218,000 servers with 300,000 servers in their new data center
- Sun: Spent $350M on a new data center at Las Vegas for their Cloud server
- WikiPedia: 300 servers
And what of Apple? Well they have just committed to $1B for a new data center in the US. Is Apple the next company to start looking to get into the cloud business? Or have they got more consumer orientated plans for their servers?
There will be a generation of architects coming behind us that won't even think about the hardware when building out new systems but at it is still the top question asked at the boot camp, just how many machines can I run up?
At the moment though, we still like to think in the real world, of physical boxes with flashing lights and whirling fans.