Powers Unfiltered

An entrepreneur’s journey into grid computing and partnering with Microsoft, by John Powers

Powers Unfiltered header image 2

Windows is optimal OS for grid computing, says Digipede

August 3rd, 2006 · 7 Comments

I adapted the title of this post from an article that appeared this week in PC Welt, entitled “Linux is optimal OS for grid computing, says Oracle.” (I found it through the diligent grid coverage provided by Greg Nawrocki in Grid Meter — thanks Greg!)

The article is primarily coverage of a presentation by Guy Cross, director of Business Development for Oracle’s Asia Pacific Linux Business Unit; the article is a little sketchy, and one hopes Mr. Cross did a better job of linking the platitudes with the conclusions. For example:

To leverage on Linux and the grid, Cross advised enterprises to take the following steps:

1) Standardize. Take inventory to find out what you are running, and ask if the vendors will be around in 10 years, he said. “Do research and find out what the vendors are rallying behind.” The answer, he said, lies in the “O3 zone’ ” open source, open standards and open systems.

2) Consolidation.  Have a 360 view of your business and start to migrate to do more with less, so that there is less cost to manage. Start at the hardware layer and then move to the database and then applications, to higher levels of abstraction.

3) Automate. Take advantage of grid computing by deploying groups of small, cheap servers, or leverage on Oracle on demand to have software delivered as a service, so that the enterprise can focus on its core business.

What?

1. Standardize. OK, and you do that by trying to guess who will be around in ten years?? Maybe industry giants like Compaq, or Digital Equipment, or Silicon Graphics, or Lotus… Get serious. Any tech purchase made today based on a 10-year analysis is doomed; any grid project that doesn’t pay for itself in one or two years isn’t worth funding — something far better will be along in less than two years. This is typical big-company FUD, nothing more.   By the way — what’s the over/under on Oracle’s continued independent existence? If it’s ten years, please excuse me, because I have to call my bookie.

And as for the “rallying” of vendors — vendors are rallying behind two separate banners. First (yes, first — go check), there’s Microsoft and .NET. Second, there’s the “O3” banner that Mr. Cross identifies, with dozens (or thousands) of sub-banners that permit vendors to claim to be rallying while they’re actually feuding.  And who advocates rallying behind a big, established company working (a bit) with open source? Why, it’s the head of bizdev for Oracle’s Linux group in Asia.

2. Consolidation. Rubbish. Absolute rubbish. The tech industry was born fragmented, and has been fragmenting ever since. Over and over I hear people say “oh, big enterprise IT departments are consolidating — they want to do more business with fewer vendors, because it’s easier to manage.  They want to have fewer platforms, fewer “special projects,” more focus, simpler systems, more control.  Time to standardize and consolidate — you’ll sleep better.”  Guess who I hear it from?  Not from CIOs, who understand where innovation comes from.  Not from department managers or application owners who are in pain and need a solution instead of platitudes.  No — I hear it from big incumbent vendors (and occassionally the analysts they employ).  Every few years we hear about a new giant wave of consolidation, and when the dust clears there are more platforms, not fewer; more vendors in more categories, not fewer; more new technologies to master and monetize, not fewer.  Note to Mr. Cross — that’s a tech trend called “innovation.”  It’s a tech trend frequently resisted by big incumbents too accustomed to comfy margins. 

3. Automate. Automate? What does building a grid using “small cheap servers” have to do with automation? And where does Linux come into it? More small cheap Windows servers are sold every day than small cheap Linux servers. You lost me. Again. And as soon as we’re lost, who advocates for just handing everything over to Oracle so we can focus on, um, something else? Why, it’s the head of bizdev for Oracle’s Linux group in Asia.

My turn:

To leverage on economics and the grid, Powers advises enterprises to take the following steps:

1. Standardize. Take inventory of your computers — what OS do most of them run? Take inventory of your applications — on what platform do most of them run? Take inventory of the skills in your architecture and development group — on what platform are they most productive? Then choose a grid technology that leverages those resources and skills.

2. Embrace fragmentation. Empower departments and application owners with control over resources and development direction. Support them — don’t ignore them.  Start with the applications, not the hardware. Hardware is cheap and getting cheaper! Deploy applications on your grid in rank order of greatest benefit/cost ratio.

3. Simplify. Stay away from grandiose visions that require ten years. Acheive value now by reducing complexity in grid design and implementation, and by keeping the solution focused on where deploying applications to the grid provides the highest benefit/cost ratio.

For each of these steps, Windows dominates Linux in most enterprises as the right OS for grid computing. With a few exceptions, enterprises have more Windows hardware, more Windows software, and more Windows development expertise than anything else. Grid computing is important — but not important enough to overturn platform decisions already in place. “Tear everything out so you can do grid computing?” I don’t think so.

In the end, a grid is for applications, and applications are written in .NET for the Windows operating system.  (You may want to check out the latest figures if you don’t believe me — kudos to Darryl Taft of eWeek.)  Are there exceptions?  Sure.  Enough to make Linux the “optimal OS for grid computing?”  Check back with me in ten years and let’s see.

Tags: Grid applications

7 responses so far ↓

  • 1 Phil // Aug 4, 2006 at 6:25 pm

    You bring up some good points/best practices. Although I think that grid computing is independent of the operating systems of the machines that participate in the grid (at least a grid should support this), I don’t think that Linux is better/worse than Windows (or Macs for that matter).

    I am looking at implementing a grid for logistics analysis and I wonder what the benefits of the Digipede product is over other open source .NET products?

  • 2 john // Aug 4, 2006 at 8:54 pm

    Thanks for your comments, Phil —
    In terms of your question about how the Digipede Network stacks up against open-source .NET products, you might want to check out this earlier entry; I talked in that post about Alchemi, and the same goes for any more recent efforts. In open source, critical mass is, well, critical, and none of the .NET grid projects have come close to that.
    Even with better-supported open-source projects, you need to examine your own level of expertise, patience, budget, specific requirements, and so on. In my (biased) opinion, Digipede’s combination of great funtionality, documentation and support (not to mention low entry price!) make it more useful in most situations than anything produced through open-source .NET projects. Email me (john [at] digipede [dot] net), and I’ll send you some more detailed info. I’m interested in learning more about your logistics analysis application; I hope we can help.
    I’ll take up the advantages and disadvantages of Windows as a grid OS again shortly, because I think there are some cases where one OS is indeed better or worse than another, but that’s for another day.

  • 3 Diego // Aug 10, 2006 at 3:55 pm

    Standardize? Ian Foster started describing this new paradigm including the adjective “heterogeneous”, so, this is not a good approach.

  • 4 john // Aug 12, 2006 at 4:38 am

    Diego —
    Ian Foster is a smart man. I have read his work thoroughly, attended his presentations, and had several worthwhile discussions with him. He’s not wrong; he has different objectives than we do. I would not disagree with the word “heterogeneous;” but one need not have multiple operating systems to have a valuable grid. In my opinion, in most enterprises, moving from the current state (no grid) to some hypothetical end state (multi-OS grid) in one jump is counterproductive. The first step is moving from no grid to a grid that handles the most common, highest-priority applications without needless complexity. For those who want to attain the most value from grid computing now, I’ll stand by my recommendation.

  • 5 Phillip Fayers // Aug 14, 2006 at 9:22 am

    I think I agree with you on the first two steps, but the problem comes in maintaining the balance between them – if you lean too heavily towards Standardization you take away opportunities for Fragmentation.

    I agree whole heartedly with you on step 3. You simply can’t second guess what the IT industry is going to look like in 10 years. The problem is that a lot of companies have 5 and 10 year plans and they expect their IT people to fit in with those plans. I think it was the authors of Peopleware who said that if your company has a five year plan it sets the tone for everthing in the company, so don’t expect change to happen any faster than that 5 year schedule.

  • 6 john // Aug 14, 2006 at 4:23 pm

    Thanks Phillip. I think you’re right — many do companies have five-year plans, and aligning IT with those plans is important. My issue with Guy Cross’s presentation (or the way it was described in the article that started this whole discussion) is his contention that a five (or ten) year plan has anything to do with vendor selection. No buyer I know, however large and plan-constrained, limits its IT vendors to those who “will be around it 10 years.” Buyers select IT vendors based on who gives them a competitive advantage now and in the future. Certainly, part of that consideration is stability — of the vendor, and of the vendor’s technology. So we all need to do a good job of showing our multi-year roadmaps to buyers, to explaining how we’re aligned with their current and planned needs. But I have no reason to believe that a slow-growth IT company with lots of staff turnover does that any better than a smaller company with a great management team and a growing staff.

  • 7 Powers Unfiltered » Blog Archive » Grid Computing for Windows, part deux // Aug 27, 2006 at 10:44 pm

    […] As my loyal readers know, I recently spotted an article called “Linux is optimal OS for grid computing, says Oracle,” and I took a swipe at it because, well, I disagree. Today I’ll take a little more time to explain my comments about the Microsoft platform, and why we’ve bet our company on grid computing for Windows. […]

Leave a Comment