The end of corporate computing
sloanreview.mit.edu
sloanreview.mit.edu
The initial costs associated with building a data center seem enormous. The maintenance costs also must be staggering. Does anyone know what percent of DC construction/maintenance is spent on redundancies? (e.g. backup generators, UPS, multiple carrier services, etc).
Abstracting away the physical infrastructure for a 'logical' infrastructure (if you will), seems very attractive to me. Particularly if I can have an on-line balance between compute cost and benefit.
One major concern with this approach is security, both physical and otherwise...
You pay your money, get your own private space in a DC, they provide the power, cooling and physical security and you supply the kit and buy connectivity from one of the ISPs/telcos that infest these places.
Yes, firms can co-locate their equipment in other firms DCs but its not a silver bullet either. You still run up heavy costs associated with renting the physical space, and hardware maintenance. If we're talking about a company of significant size that is going to be part of a paradigmatic shift -- the cost/benefit of co-locating doesn't scale well (and that's why you'll see many sizable non pure tech companies that own and operate their own DCs)
[simplified] Example: $1 million to outsource to a provider = $1 million that can be written off in that year.
Build a DC, install servers etc. at capital cost of $2 million and yearly operating costs of $250K - you can only write off perhaps $200K of depreciation plus the $250K in operating costs = $450K per year that can be written off.
Imagine there exist two companies who are competitors, A and B. Now company A outsources its data with a third party service. When company A managed its own hardware infrastructure it had tight control over physical access to every machine. Now that its data is hosted with a 3rd party, there is a greater possibility that the data becomes physically compromised.
Also, your A/B argument works about as well against commoditized electric power as it does against commoditized data storage. Why is it a no-brainer for company A to give up its generating infrastructure but not its server infrastructure?
Yes, there will be a few exceptions like hospitals, which keep generators on site and have very strict confidentiality rules to observe. But in the main, if a factory can risk a temporary loss of power to all its machines and still come out ahead by dealing with a utility, why can't an office supply company also come out ahead by risking temporary loss of access to its CRM data?
I do think that for startps and smaller companies, receiving and paying for storage/compute power like electricity is the way to go.
But back for a moment to the A/B example. My point here is really with companies that are dealing with sensitive information. Let A = JP Morgan & B = Morgan Stanley.
JP Morgan, I would guess (somewhat wildly) probably owns and operates in the neighborhood of 18,000+ servers. At least half of those (roughly), probably contain mission critical programs and data. The other half, that don't contain mission critical data per se, probably contain either 1) backups of mission critical applications for emergencies, 2) email, documents etc of every single employee (sensitive data).
Now as long as JP gets the electricity that it needs who care where it comes from? It just needs to run its business. But the moment it lets important data out the door, all bets are off. Who accessed what? Is the network secure? Is the building physically secure from intrusion and corporate espionage? Does Morgan Stanley now know the JP automated trading strategies? and so on... Which is why I would say its not a no brainer to simply give up control of server infrastructure.
The data providers aren't idiots. They know what issues corporations are going to have with outsourcing. And they have provided solutions to those issues.
So yeah, getting closer all the time.