As a self-professed fan of Harvard professor and author Clayton Christensen, Stuart Frost is drawn to words like innovative and disruptive. A long-time software engineer and entrepreneur, Frost founded SELECT Software Tools in 1988, leading it to a successful initial public offering in 1996. It was a good experience and case tools were interesting, but at the end of the day, it was a few hundred million-dollar market with 200 players, Frost recalls. Looking for a bigger pie with a few incumbents (another favored word) ripe for disruption, Frost would eventually found and lead data warehouse appliance upstart DATAllegro as chief executive officer. Since 2003, when the company came to market, the soft-spoken and youthful-looking executive has taken DATAllegro on a wild growth ride, as he recently explained to DM Review Editorial Director Jim Ericson.
DMR: The data warehouse appliance market seems to have arisen very quickly, almost out of thin air.
Stuart Frost: Several years ago, my background led me toward looking at the problems associated with large-scale database modeling. I was looking for something that would be of great importance to CIOs and disruptive to the market. I found data warehousing in particular was quite an interesting market that had been dominated by Teradata, IBM, Oracle and Microsoft. There had been very little innovation, partly because venture capitalists saw databases in general and data warehousing in particular as a no-go area. They thought it was all over, that Oracle had won the database market and that was it. That started to change when open source databases came out around 2000-2001, when MySQL and a few other things got funding.
DMR: Your competitor Netezza had already come to market with a similar idea.
SF: It was around that time when Netezza was getting their first customers, so I started looking at what they were doing. Id pretty much come up with the initial ideas completely separate from Netezza, but what theyd done was interesting. Theyd made good use of the appliance business model to address that gap between the software and hardware that DBAs [database administrators] find so difficult to tune. By having a vendor driving a tightly defined platform, optimizing and tuning it, you can get much better performance in a much more consistent way. To a certain extent, it was realizing what Teradata had realized 30 years before. The other thing was the shift from random I/O [input/output] to sequential I/O. Even though a lot of database engines try to do sequential I/O because theyre doing large scans, the structure means they end up doing lots of random I/O against a lot of disk drives. What you see in Teradata, Oracle or SQL Server is huge disk arrays typically shared across a number of servers doing lots of disk head movements. Netezza broke away from that. The downside of the way they approached it, though, was they ended up with proprietary hardware.
DMR: Your own strategy revolves around open source and key partnerships.
SF: The thinking was that we could take an open source database as a starting point, and rather than build a whole relational engine, build a massively parallel system around it and effectively create a kind of second-generation Teradata pretty quickly and cheaply. We wanted to achieve the same kind of breakthroughs through software intelligence rather than the hardware piece, and do it in a way that was portable across different hardware servers and storage. I realized there was a big benefit of being able to co-sell with the likes of EMC and Dell. That was the business - take the appliance business model, leverage an open source database, achieve sequential I/O under a heavy workload and do it on a complete set of hardware. We use Ingres as a standard relational engine, with some sophisticated software around it to parallelize a set of Ingres instances to turn it into a massively parallel database to do sequential I/O, and it works.
DMR: The appliance market was initially seen as a kind of stopgap to address specific data challenges. How have the economics changed, and are appliances now standard enterprise data warehouse infrastructure?
SF: To the last part of the question, Id say the answer is increasingly yes. Its well beyond stopgap; established vendors like ourselves now have the ability to build an enterprise data warehouse on this technology. Youve got a $10 billion market for software; if you include hardware, its $25 to $30 billion. A lot of the major players are trying to leverage whats going on in that disruptive scenario because it is such an important market. Youre seeing all this activity, HP with Neoview, Sun tying up with Greenplum, us partnering with EMC. Initially, companies like ours took advantage of the tactical requirements of companies building data marts or doing batch offloads. Weve been able to show, though, that we can build real enterprise-wide systems. When you can do that at one-tenth the price with 20 to 50 times the performance, its a very compelling value proposition.
DMR: Like Netezza, DATAllegro is strongly targeting Teradata customers right now.
SF: I like to think were staying focused and then exploiting our beachhead. Weve put out some press about Teradata utilities for coexistence and have a very aggressive node-for-node trade-in for Teradata customers. Were targeting Teradata customers today because we think theyre being grossly overcharged for the capabilities theyre getting. Weve proven that we can move workloads over very smoothly at dramatic savings but also with greatly increased capabilities, which is a pretty compelling value proposition for that customer base. Were not playing the game of trying to be like Netezza. We do produce a DW appliance, but there are quite a lot of differentiators around our grid technology and the way the platform works and scales and the partners were bringing to bear that I believe makes us uniquely attractive.
DMR: Lou Agosta, who works for IBM, wrote in the March issue of DM Review that early appliance innovators are already being leapfrogged by major power players. How would you respond?
SF: (Laughs.) The larger vendors have tried to respond in the classic incumbent way of trying so hard not to compete with their current mainstream products that they end up competing with no one. Clearly, HP has come out with Neoview, theyve paid lip service with NonStop SQL - its been a real struggle for them. I think thats true to say for IBM as well, despite Lous protestations. At the end of the day, the IBM BCU [Balanced Configuration Unit] is DB2 on a bunch of IBM servers. Its not an appliance, does not have the benefits and does not have the performance. So far, its clear that DATAllegro and Netezza are the leading disruptors in this market and have broken clear of the pack. But clearly they have access to markets that we dont.
DMR: Lou also predicted that, going forward, enterprises will need only one kind of database for both transactional and BI processes.
SF: That is so far removed from reality as to be ridiculous. The workload is just so different that no one engine can truly do that. Were actually seeing a fairly rapid divergence in those two things because of the way you drive I/O. OLTP [online transactional processing] is going much more toward being in-memory, solid state disk drives and so on. Were a long way away from being able to do that with a data warehouse of 100TB. Its simply too expensive to get anywhere close to that. If thats where Lou or IBM is going, I think theyre going to be on their own with that idea.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access