The future of datacenter build & co-lo
BY JAN WIERSMA
Last year my friend Tim Crawford wrote an excellent article on why CIOs should get out of the datacenter business. Tim focused on how current big cooperates are moving away from building, owning or renting datacenter facilities in favour of consuming IT at higher levels of the stack.
As he focused on the migration of leading big companies, it leaves the question; what about the future Fortune 500 companies?
How business would evolve in the past
In the past a small business would start with a single PC or multiple PC’s connected to a local network. As the company grew, the need for central storage and applications would arise. One server would lead to 5 servers and we would move the noisy systems in to a server closet. The closet grew in to a small server room and that would be our journey to building our own datacenters or moving systems to a co-lo facility.
How do businesses evolve currently?
Currently a small business starts with a laptop or tablet and an internet connection. Apps focused on personal productivity migrate in to collaboration apps as the company grows. Documents are created online using Office 365 or Google Apps. Storage is consumed on Box.com or Dropbox. Business specific apps are build using PAAS platforms or hosted on IAAS platforms. As all these platforms are massively scalable, they support the company from 5 employees to 5000 employees growth. No server is bought. No datacenter is build.
But; the datacenter will not go away. It will not die. It’s still very needed. Current IAAS companies will need datacenters as they experience massive growth. New IAAS companies will need datacenters to start their business. Companies that benefit from hardware optimisation (special servers) like Bitcoin miners will need datacenter space. As that benefit needs to be substantial, it means running ‘single purpose IT’ at extremely large scale to gain a real benefit. Those are the ones looking for datacenter space and can benefit for datacenter optimisations. (other examples: Facebook and eBay.) This trend means the market for datacenter builds and co-lo is changing.
I know some business developers will start to play the ‘security card’ (Cloud is not secure) or even the ‘privacy card’ (data needs to remain in X country) but those issues are being actively addressed by all the major cloud providers and governmental organisations around the world. Pandora’s box on IT-as-a-utility has been opened and can’t be closed again.
So where does it leave the datacenter business?
The mega datacenter will focus on innovation for reduction of operational cost. They will be highly engineered for specific hardware and workloads. It’s the realisation of ‘the datacenter as a computer’ that Urs Hölzle (Google) wrote about. All this innovation is great as it leads to more efficient and greener datacenters. Co-lo remains an other big part of the construction market.
Modular datacenters will continue to have limited success, as it was focused on eliminating pains from CIO’s. As that pain has evaporated with cloud consumption, the need for standard – low cost – fast deployment datacenters has slowed down. Maybe the co-lo market will benefit from this innovation but specifically real estate investment trust companies seem focused on traditional datacenter builds.
The co-lo business needs to focus on IAAS, some PAAS, PC-gaming and other very specific needs like Bitcoin miners, that are not served by mainstream IAAS. Short term some CIO’s will push out datacenter space from their own datacenters in to co-lo space, but more and more that step is skipped in favour of moving directly to IAAS.