39
Figure 1: TCO‐Reduction with growing size of datacentre (Microsoft)
Ten years ago, a thousand server datacentre would be considered a large
scale datacentre. At that time hardware was more expensive, bandwidth
limited and management less automated. However, technology changes
over time, as well as the price for technology. During the last decade we
have seen a dramatic decrease in hardware cost, a decrease in bandwidth
cost and capability as well as highly virtualised and more automated systems
management software being created. This has changed the equation on
what is considered to be the “best economical size” of a datacentre. In
recent years the IT industry has been able to effectively build datacentres
with hundreds of thousands of servers and to have a good understanding of
the economy of scale associated to these new large scale datacentres.
So, let’s reflect on what the diagram above actually means. As you can see,
the current combination of the price for hardware, bandwidth cost, labour
cost, energy cost etc., indicates that a large scale efficient Cloud computing
provider can reduce
its own
total cost of ownership (TCO) with up to 80%
going from 1 000 to 100 000 servers. It’s no wonder these mega datacentres
are being reported as having the potential to massively reduce the price for
computing resources for the world.
2.4.2
How does cheaper technology get adopted?
So, does cheaper technology necessarily mean adoption of that technology?
If we ignore technical, security or legal considerations for a while, then the
