home   |   contact us   |   subscribe to rss

Grid computing’s value proposition

March 16th, 2005 by mhjones

The Economist’s latest Technology Quarterly picks up on the Grid Computing meme, and pours cold water on Sun Microsystems in the process.

"Who wants to buy a computon?" (sub req) implies that delivering computing power in a utility-like fashion has merit, but the technology’s evangelists – notably Jonathan Schwartz – are losing people amid all the geek-speak. We’re not hearing the right economic argument.

Since I don’t have an online sub, I’m forced to actually type a few grafs from the magazine to explain:

Offerings such as Sun Grid, while novel, do not solve the ultimate problem: the efficient allocation of networked resources. People do not think of their computing needed in terms of, say 50 processor-hours; instead they have specific tasks of varying importance and urgency, and want to get those tasks done economically, using whatever resources are available.

Since I’ve been following the grid story for years, my first reaction was to dismiss this notion as nonsense. But the story goes on to quote Bernardo Huberman, a researcher from HP whose developed grid software called Tycoon. Tycoon is a "stockmarket or clearing house" for grid computing users. You’ve got an urgent processing job? Then pay more for your computing cycles than the next guy and you can jump straight to the front of the processing queue. It’s the eBay of grids.

It’s an interesting idea. But to me what’s more important is how we understand/sell these ideas to the wider IT community. Huberman proposes calling the unit of value grid customers buy a "computon" (mix of computation and photon). A computon is meant to signify a performance & value benchmark that combines processing cycles, memory, storage, and network bandwidth.

Despite The Economist’s attempt at uncovering a brilliant new insight, it turns out computon is not a new idea (eg, this Grid blog story from 2003). So that then raises the question: why have Sun or HP apparently not marketed the ‘computon’ or an equivalent term.

Sure, Sun’s got it’s $1 per processing hour model. But what if Sun jacks up the price to $1.50? What’s the common benchmark across the industry that gives users a clear sense of comparitive value? As much as I hate jargon, it strikes me that IT professionals need a term like computon & associated standards that finally brings grid computing into the real world of enterprise computing.

Tagged: IT business
Comments: 4 Comments »

4 Responses to “Grid computing’s value proposition”

  1. Jonathan Schwartz Says:

    But… what’s the standard measure of a barrel of oil? Did you mean crude? Or light sweet? Or diesel? Or kerosene? Or check out the SLA on a “standard” bandwidth contract (it’s anything but simple).

    The point is, there will be an evolving set of definitions around true network utilities – but our bet (for now anyways) is that time is the most fundamental meter. Same for electricity (kilowatt-hours) bandwidth (megabits per second), cable service (per month), etc. Not the only measure, but the most fundamental.

    If you recall, IBM did try to market their way around this way back in the era of timesharing – it was called a MIP. I haven’t found a CIO that thought that or their timesharing contracts were all that transparent…

  2. Mark Jones Says:

    Hey Jonathan. Yep, I remember MIPs. And that’s my point – confusion still reigns.

  3. Jonathan Schwartz Says:

    Not sure I’d agree – there’s been tons of “On Demand” this and “Utility” that – driving the confusion. But I’d like to believe cpu-hrs are the meter the world will use. I’ll make a bet that a few years from now, that’ll be the case.

    And like light sweet vs. crude, a barrel isn’t a barrel – nor will all cpu-hrs be the same.

  4. Mark Jones Says:

    Ok, fair enough. If we have various scales of costs associated with time taken for processing jobs, let’s see the vendor collective agree on an approach to evaluating how to charge for the shades of grey.

Leave a Comment