It seems like just the other day I was shaking my head at the determination of relational database vendors to keep fine-tuning their legacy software in order to optimize it for dedicated proprietary hardware. In essence, to deliver the last drips of performance by acquiring huge alloy wheels, an enormous spoiler, new carbs, a noisy exhaust pipe, and painting flames down the side of their database. To the outside world it looks like it should go fast, but inside there’s still the same rickety old steam engine under the hood. 


This is very similar to what I recall from my first days in the database market. IBM and the big box vendors, ICL, Unisys, etc. had a stranglehold on the market with file technology that was two decades old. Remembers IMS, IDMS, TOTAL? They all held COBOL as the centre of their Universe. COBOL did deliver secure, scalable, and reliable environments. It was also expensive, slow, and delivered sub-par applications compared to what could be delivered with Unix, Client/Server, and Relational Databases.

How did you fix things back in the day, especially if the problem involved more users, more data, more functionality? The answer was more disk, more memory, more CPU, and lots more money. (On a personal note, thank goodness for that!)

I started my software career in this market in 1990 and with a better, faster, cheaper solution we had a lot of budget to work within and deliver ROI.

But then those products started to mature, to become entrenched, and to stop innovating on future-thinking software in order to preserve past revenue streams.

Rinse, wash, repeat.

It’s twenty years later and how do you get Oracle, Teradata, Netezza, and IBM’s data to go faster? More disk, more memory, more CPU, and lots more money. Why else do you pay nearly 10 times revenues?

If the definition of insanity is doing the same things over and over again and expecting different results, then continuously throwing hardware at what’s clearly a software problem is insane.

Once again, it’s time for a new approach.

SAND believes only a massively parallel, column-oriented database can properly handle the requirements of the modern database market. More specifically, we believe only SAND CDBMS has the core technology to handle the petabytes of data, tens of thousands of users, and truly complex, deep-dive analytics increasingly critical to the enterprise.

So when the team asked me today what I thought about IBM’s acquisition of Netezza, I said it was what’s to be expected, and there’s likely more to come.

Now that Mark Hurd’s gone will we finally see HP ditch NeoView (or is that NoView?) and buy Teradata? With Kickfire going to Teradata, DATAllegro to Microsoft, Sun to Oracle to make their own appliance, and now IBM getting Netezza, that leaves the independent software vendors with the innovative software uniquely poised to deliver the real solutions.

At least if we’re not confused about who the real competition is, unlike some who will remain named (ParAccel). Our competition is SAP’s Sybase IQ (hmm two acquisitions and a funeral?), Vertica, Infobright, and yes, ParAccel too. Then again, I remember ParAccel clearly talking about Oracle and Netezza and somehow forgetting Vertica. Perhaps they don’t really have a columnar database? They did start out as an appliance…

Either way, it feels like 1990 all over again.

Now, time to pull out my double-breasted suit, Nirvana CD’s and my favourite Keds.