Depressingly, however, throughout this year, my confidence in Netezza has been slipping. I've been a huge Netezza fan since I first learned about them almost three years ago. I spent quite a bit of time telling everyone I could find that they needed to get on the Netezza bandwagon or get run over by it. So my sense that the database world was running away and leaving Netezza behind was downright depressing.
Turns out I was right. I just shouldn't have been depressed about it.
I think my concern about Netezza's future was somewhat justified, as this year's assault on Netezza has been three-fold:
- The advent of column-oriented databases was the first real assault on Netezza's major advantage - query performance.
- Plug-in query accelerators like Dataupia and ParAccel have attacked their primary weakness - compatibility.
- The variety of MPP database vendors offering white-box solutions (GreenPlum, Vertica, etc.) has certainly accentuated Netezza's secondary weakness - proprietary hardware.
It occurred to me a few days ago, however, that the seemingly innocuous things I've heard from Netezza lately actually add up to some faith-restoring conclusions. The fact that they exhibited at SC07, for example, lends some credence to my theory that they aren't just a database vendor anymore. Now granted, I wasn't at SC07, so maybe I'm seeing what I want to see. Could be.
But what really soothed my soul was the discovery that SPUBoxes are available to Netezza Development Network members . SPUBoxes have long been a reality inside Netezza, and I never understood why they weren't a reality outside Netezza. I guess I still don't have an answer, but the important thing is that they are available outside Netezza.
Now, as a software guy, hardware doesn't usually do much for me, but I think this is key. One of Netezza's long-standing issues has been how to get the system in the hands of developers, partners, etc. so that they can push the platform forward. Loaning people multi-million dollar systems just isn't a scalable approach, after all. The SPUBox finally solves that problem. That should mean explosive growth in the number of people familiar with the system as well as the number of applications developed for it.
Or maybe I should say with it. SPUBoxes are specifically aimed at developers who want to use the new UDFs functionality to push work down to the SPUs. That seems like an odd goal for a database company, even one with proprietary hardware.
Unless, of course, Netezza doesn't really consider itself a database company anymore.
And so, we've come full circle. I think the database world is running away from Netezza, but not because Netezza is standing still. I think they're just running in a totally different direction.
I wonder how long it'll take the world to catch up with them this time...