Engineering…

I have seen the future of data storage and I weep for it.

A few random rants at 2am after a data migration didn’t go because I’m not willing to kill a backup process that has a hold of my mount-point.

—-

Engineering seems to be a thing of the past.  We’ve graduated to the world of “that’s good enough for launch, we’ll fix the bugs in future releases” in spades.

I’m so tired of hearing people explain to me that a solution is “good enough” or compromising technically due to cost, when we all know that a badly engineered product will suck the life and profitability out of any company.

Most vendors, and EMC is included in this generalization, have taken to solving hardware engineering problems with software.  Enter the “appliance”.  Consumer grade hardware thrown in with crappy software designed in some third-world country without any thought to the long term failure rates on such combination.

Poor engineers make the mistake of assuming that an overly complex solution must be better than a simple one.  IBM GPFS, Sun Shared QFS, EMC Celerra MPFS are all examples of a psychotically overcomplicated solution to a very simple problem that can be solved with NFSv4 and a decent network back-end.

The more moving parts you throw at a problem, the more chances something stupid, or someone stupid, is going to foul up the mix.

And the more external vendors you buy your parts from the more your chances of having to deal with a problem that is of SOMEONE ELSE’s making, since you’ve given up control of your product and Quality is truly a thing of the past.  (You’ve also made your product *WAY* more expensive than it needs to be, because you’re not supporting the margin of everyone else between you and the supplier)

And if you buy the cheapest product to do a job, remember that it’s cheap for a reason.  IBM has been known to give up to 90% discounts on the XiV platform in order to compete with EMC.

90% – wow.  Reminds me of the usual assumption about guys and big trucks.

If you’ve got to give that much of a discount to convince people to buy your product, then it’s not much of a product now is it.

—-

4 comments

Skip to comment form

    • Ryan on September 21, 2010 at 9:42 am
    • Reply

    I suspect that “it’s good enough to ship” is going to be a big nail in the coffin of IT. We’re not going to die of course, but at some point, there are going to be some HUGE and HIGHLY VISIBLE outages causing corporate embarrassment and RGE (resume generating events) and it’s going to be ‘good enough’.

      • Jesse on September 21, 2010 at 5:35 pm
        Author
      • Reply

      The good news is, the more they cut corners on product, the more they’re going to spend on support, and even with all of the illiterate overseas support reps third-world countries have to offer, it *still* means more money will be spent supporting the products.

    • Ed on March 13, 2011 at 4:16 pm
    • Reply

    you make one inaccurate assumption about NFSv4 vs MPFS. I agree that NFSv4 is way simpler than MPFS… but you need to understand that MPFS was a predecessor to v4. In fact, EMC helped develop the v4 standard based on a lot of the features in MPFS.

    1. But NFS is still a simple and elegant solution. Whereas I’ve gone crosseyed at the implementation of MPFS a few times…

      But, I stand corrected.

Leave a Reply

Your email address will not be published.