The future of computing demand
February 09, 2007
Sun Microsystems has posted videos of its Analyst Day presentations from earlier this week. Following a tip from CEO Jonathan Schwartz, I watched CTO Greg Papadopoulos's presentation on the future of computing demand (pdf of slides). It's well worth a look.
Papadopoulos shows that the traditional big driver of computing demand - basic business computing - has lost its force. The growth in computing power through Moore's Law now far outstrips the growth in demand from traditional business computing. That means, in essence, that businesses will need far fewer computers in the future to fulfill their demand - a fact already manifesting itself in IT departments' emphasis on server consolidation and virtualization. The big opportunity here is simply to improve your utilization of existing capacity.
But, argues Papadopoulos, there are three new drivers of computing demand that far outstrip the expansion in supply guaranteed by Moore's Law:
1. Rich multimedia content delivered through the broadband Internet (think YouTube and VOIP).
2. High-performance supercomputing (think weather modeling and drug development).
3. Software as a service (think Salesforce.com, Webex and Office Live).
In combination, these three sources will produce an exponential leap in demand for computing - Papadopoulos calls it the "Redshift" - that will far outstrip the increase in supply produced by Moore's Law alone. In other words, you're going to need a lot more computers (or at least a lot more computing cycles). Whereas the focus of those organizations running traditional business-computing operations is cost cutting through consolidation, the focus of those operating the new mega computing operations (like Google, say) is achieving efficiency at massive computing scale (through, for instance, reducing electricity consumption). There's a fundamental split opening up in the market, in other words: two very different sets of customers (one with stagnant demand and one with burgeoning demand) with very different needs.
So how will the new sources of demand be fulfilled? Papadopoulos offers two scenarios, which might be labeled "Sun's nightmare" and "Sun's dream." Under Sun's nightmare, the new mega-scale computing operations follow the Google model and "disintermediate" traditional computer companies. They build their own systems from scratch. The old enterprise computer business turns into a "commodity parts market," as Papadopoulos puts it.
Under Sun's dream, the mega-scale computing giants continue to look to outside suppliers to provide the sophisticated computing systems they need. Google, in this scenario, is an early pioneer of the engineering of mega-scale computing but it is not a model for the future supply of such computing. (A historical analogue for Sun's dream is mega-scale electricity production a hundred years ago - the big utilities didn't all build their own generating systems; they bought them from GE and Westinghouse.)
In Sun's dream, computing (like electric current) becomes a commodity but computers (like electric generators) don't. Those who dominate the computer business of the future, under this scenario, would be the engineering giants, not the assemblers. The winners would not be those who pump out generic cheap boxes but those who are able to build highly efficient mega-scale machines. In some ways, that would mean the future of the computer industry would look more like the mainframe era than the PC era.
a telling slide to me was 18 where he talks about brutal efficiency - and nowhere does he mention "cost". Every large vendor already has massive scale but dilutes it in their sales pitches by offering the other C - customization. Sun and other industry leaders are their own worst enemies in the field. They incent their field to "value sell" - all kinds of customized services and distributed functionality. No matter what Jonathan says they want to sell fancier generators, not utility computing yet.
Posted by: vinnie mirchandani at February 9, 2007 01:03 PM
You might want to check, but I beleive the current electricity situation in the US has to do with regulations: the anti-trust doctrine would tolerate the GE monopoly, but not having a continental giant for operation. One of the way you can understand that is by noticing how, abroad, many of those are public, or simply local monopolies. If it is the case -- and as Internet regulations is more subtle then the "anyone having more then that should go to jail" approach of the Sherman Act --- then the nightmare is more likely. That would be if we face a relevant computing power becoming a commodity market; my opinion is more that this commodity will only serve a more important market for managing personal information, driven by innovation; in such a market, CPU time might be as relevant as water is to the food industry: essential, but not decisive.
Post a comment
Thanks for signing in, . Now you can comment. (sign out)(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)
"Riveting" -San Francisco Chronicle
"Rewarding" -Financial Times
"Ominously prescient" -Kirkus Reviews
"Riveting stuff" -New York Post