Will computing flow like electricity?

Will computing flow like electricity?

Summary: News.com's Martin Lamonica surveys several industry executives on whether they believe that computing will eventually flow like electricity.

TOPICS: Hardware

powergrid.jpgNews.com's Martin Lamonica surveys several industry executives on whether they believe that computing will eventually flow like electricity. The consensus is that "utility" computing will work for some applications and companies, but won't become a pervasive driver of IT transformation overnight. Charles Giancarlo, Cisco CTO, said, "We think (utility computing) makes sense for some small and medium-size businesses. But for large businesses, the decision to host applications outside or inside of the network depends on many different factors, including cost and network efficiency," said Giancarlo. "Some of the largest companies can run their own applications much cheaper and more efficiently than any utility computing provider."  

Martin is playing off of Nick Carr's article, "The End of Corporate Computing," in which the author argues that the shift from IT as a fragmented capital asset to a centralized utility service will create far more upheaval than the introduction of the PC and the Internet did in past decades. I blogged about Carr's article back in April, mostly agreeing with him that computing will eventually flow like electricity but not until cultural and economic barriers (accepting the fact that many IT functions can be better delivered as services and the cost savings are obvious. Whether utility computing is viewed historically as more significant than the PC or the Internet should be left to the historians. On the other hand, those who fail to gear up culturally and technically for utility computing in the next few years will end up at a competitive disadvantage...

Topic: Hardware

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Is there a single example where grid is used?

    A recent study of major pharmaceuticals has shown almost no interest in grid. Everyone keeps blaming culture, but ignore the technical difficulties of grid. On an economic front, does it make sense to save a million dollars in hardware but have to spend 100 million dollars in additional bandwidth? Even if gigabit WANs were free, there are still serious limitations to grid. There has yet to be an example of a localized grid that can be shown to benchmark well in database applications.

    The only good examples of grid computing are SETI and key cracking and those are highly specialized applications that require very little bandwidth to operate. Very few enterprise applications fit this description.
    • Confused

      Are you confusing Grid Computing [lots of small computers connected to make big computers, and rented by MIPS and Time slots] with the next generation of Application Services [My business needs a payroll application - I lease all the core processing activity from a service provider, and have no internal IT department]?

      I appreciate that, as the PC's power to weight and power to cost ratios continue to favor power growth, the overlap between these two views of computing becomes bigger. Even so, it is not a given that people will continue to buy even PCs.

      In addition you really must look at this prejudice you have of there being costly communications - bit rate to dollar ratios have outstripped PC power to dollar ratios in the last decade many times over. High bandwidth at low cost is already a given in many parts of the World. This trend can only continue.

      If you forget that the arguement for utility computing is essentially economic, predicated on continuing hardware and software development trends, then you lose sight of the ultimate business models that will emerge.

      Or, to put that another way, the trends we see in hardware and software development today seem likely to continue. If we assume that they will, and look into the future, how might we do things differently? IT departments are expensive both as individual entities and as a pan-economic function and, when viewed as service providers, they often fall far short of an ideal business support function. What might replace them? We could reorganise IT departments into more centralised service providers. From the quality of service and cost perspectives this means balancing special needs against common needs.

      For example: A school offers lunch to all it's pupils every weekday. Although it offers options for common fads and fashions - such as vegetarian options - it's customer base is too mixed to offer less common requirements (Kosher, Nut Allergy, or Halal, perhaps). For these pupils it still offers some common services; Water, Trash Disposal, Summer Benches, Winter Lunch Hall, etc. etc..

      In ICT even such rare problems as Nut Allergy (whatever the ICT equivalent might be) can be served because distance is no longer an obstruction (except for very high bandwidth server-to-server or client-server needs - though these are extremely rare, and even here watch the cost of comms continue to fall).

      "So how does it work?" I hear you ask. Well, there is one thing we have to remember about ICT. ICT is business, and business is about people. The most expensive things in ICT are people - the experts and the users. Reduce the numbers of either group, make it easier to train either group, and costs quickly drop.

      Centralizing ICT functions also has the counter-intuitive effect of substantially raising quaility of service (as any good application service provider will happily demonstrate for you today).
      Stephen Wheeler
      • Multiple definitions of grid

        The definition of grid depends on who you ask. You're talking about the outsourcing aspect of grid where "IT doesn't matter any more" which I personally think is nonsense.

        My post pertains to grid super clusters. You can read more about my position of grid here http://blogs.zdnet.com/Ou/index.php?p=36.

        As for cheap high-speed bandwidth, please show me where I can get it and I?ll jump a mile high. VDSL is not my idea of high bandwidth. When I say high-speed, I mean minimum of 150 mbps OC3c type connection speeds. Even then, it?s not fast enough for most grid applications. You speak in theoretical terms, I speak in realistic (what can I get now or within the next year) terms.

        You seem to feel that PCs are obsolete. The market doesn?t seem to agree with you. We?re not going to go back to the glory days of dumb terminals.
  • grid hype

    I am an HPC expert - what I do might be called grid, but I certainly do not call it that. why? because grid is misleading hype.

    grid is based on the assumption that computing can or should become completely commodified. not merely widely available or easy to get, but completely generic, like tap water. the analogy with the electrical grid is slightly more difficult, since electricity is not completely generic (AC or DC, what voltage, frequency, how many phases?). for a grid to truely exist, there would need to be a single instruction set used everywhere. even if the whole universe were to suddenly embrace the Java bytecode format (the Java language itself is irrelvant), there are still complexities in the "compute grid" that don't appear in the electrical grid (security, access to libraries, storage/persistency, etc).

    this is also very separate from more PHB-ish concepts of grid (on-demand, utility-computing, etc) which mainly revolve around partitioning and virtualization, not real generification.