This is a close to actual quote of John Landry, former CTO of Lotus (and a lot of other stuff) at last week’s MassTLC unConference on the Future of Software and the Internet. This comment came in one of the breakouts on #CloudComputing led by Judith Hurwitz. John was objecting to the use of the term “Private Cloud” to describe cloud-like operations inside the corporate firewall.
In many respects, Landry is right. It is the data center. But it does change how the data center is managed and accessed by the developers and business. It’s easiest by looking at a comparison of old vs. new.
Provisioning: getting servers for your application for dev, test and production
- Old model: make a request to IT, wait from a day to several months and maybe you get your servers (or maybe not)
- Cloud model: use a console or API and have the servers you need in minutes
Scaling: growing or shrinking your capacity based on actual demand
- Old model: for scaling up, see provisioning above. For scaling down – don’t bother. Tell them you expect the demand to increase and “hoard” the servers you have for as long as possible. Eventually IT may reclaim them and put them somewhere else. If you then need to scale up, see provisioning above. Note that this model applies in both dedicated and virtualized environments in most enterprises.
- New model: for scaling up – assuming the private cloud has the capacity – it should happen automatically (auto-scaling). For scaling down – same answer. If Monday’s are always busy, but the rest of the week is dead, you can set your capacity to scale up Sunday night and down Tuesday morning. Scale is “elastic.”
There are more aspects to explore here, and hybrid clouds which utilize both private and public infrastructures are another interesting issue. The bottom line is that Landry is right, but he is also wrong. It’s the data center, but it’s not just the data center as it’s operated today. It’s a new operations model for the data center – and it does change things.