Found a curious article from Wired discussing about an idea to have cheaper infrastructure with physical servers than from IaaS-providers. Approach of the article is curious as it is against current trend where companies are trying to virtualize their infrastructure. Scope of discussion makes it even more curious: “Why Some Startups Say the Cloud Is a Waste of Money”.
I would tend to think you need to have large volumes to reach economies of scale without sharing the infrastructure. Usually startups don’t have this merit.
The comparison: Iaas $324,000 vs. physical servers seem focusing comparison of pears against apples. Physical infrastructure needs data connections, cooling, fire safe location, physical duplication ( in some cases even distributed duplication), disaster recovery and maintenance personnel. Comparison does not look fair, if you only take into account the price of physical box. At same time you have on other side of equation all the work and other parts of the infrastructure counted into equation.
Naturally a startup, which has only few servers doesn’t need too extensive organization maintain servers, but here article is talking about “few hundred” servers. Operation of this size should take into consideration few other parameters on top of servers too.
I think trick here is that if you don’t look the big picture, you lose some important cost components which are hidden into personnel costs and surrounding over heat related to physical servers. Time consumed among boxes does not usually contribute to innovation and progress of a company (if boxes are not their trade).
Despite all of this thinking, I still agree with the end conclusion of the article: “In some cases, the cloud works. In others, it doesn’t.” You just need to understand the full business case of your own infrastructure and the magnitude of your operation. Then you are ready to make your choice.