With prices dropping even as performance increases, the data warehousing market is as competitive now as it has ever been. Vendors are scrambling to stay one step ahead of the competition, and data warehouse appliance vendor Netezza is no exception.
While its claim that it "invented" the data warehouse appliance is debatable -- just ask Teradata -- Netezza can at least take partial credit for popularizing the appliance model -- preconfigured bundles of hardware and software -- for analytic databases.
Now Netezza is embracing two recent technological developments -- in-database analytics and the use of commodity hardware -- to further bolster data warehousing compute power while keeping costs below the $20,000 per terabyte threshold.
SearchDataManagement.com recently sat down with Tim Young, Netezza's vice president of corporate marketing, to discuss the vendor's recent partnership with SAS Institute, its move to IBM blades over its own proprietary hardware, and its plans for the New Year. Below is an edited version of that conversation.
SearchDataManagement.com: You recently announced a partnership with SAS to embed the vendor's analytic capabilities into Netezza's data warehouse appliance. Why did you make this move, and how will it benefit your customers?
Tim Young: Certainly, about half of our customers use Netezza together with SAS, so I think the more we can strengthen the technical relationship between the two companies, or at least the more we can tie the technology together, the more our customers are going to benefit.And I think SAS recognized that in order to give their customers a high-performance, advanced analytics solution, it makes sense to move the analytic processing power and perform functions into the database. Traditionally, in a historical kind of architecture, you'd move data from a database into their SAS client and then analyze it there. So by performing the analytics within the database, you're avoiding moving data from a database into the client.
SearchDataManagement.com: With TwinFin, the latest version of your data warehouse appliance, you turned to IBM blades to power the software over Netezza's own proprietary hardware. How is commodity hardware affecting not just data warehouse appliance prices but performance as well? And why are you not giving customers the choice of which commodity hardware they can use, as some other data warehouse appliance vendors are?
Young: We're still delivering this as a black-box solution, and as far as our customers are concerned, they really shouldn't worry about what's in the cabinet, but what the cabinet is delivering. The benefits we get from using commodity parts is that we can spend less of our money on building blades [than] we were doing in the past. By using commodity blades, we can now divert those resources toward making TwinFin faster and more functionally rich.Whenever people use the term commodity, that does somewhat suggest that the blades are interchangeable, that we could take our software and we could run it on top of a Sun blade or on top of a Hewlett-Packard blade. The reality is that we probably could do that and the effort would probably be minimal, but I think the bottom line is that the value we are delivering to the customers is that preconfigured solution, that black-box solution that's ready to run.
SearchDataManagement.com: In a related point, Forrester Research analyst Jim Kobielus characterizes the data warehouse market as extremely competitive, with vendors needing to keep their prices below the $20,000 per terabyte threshold. Do you agree with that assessment?
Young: A few years ago, this was a market that was dominated by and really stitched up by Oracle, with a bit of Teradata and obviously IBM. But I think data warehouses were very much largely regarded as commodities. This was definitely not a market that people were willing to invest in.The market today is completely different. This has become a very dynamic, very busy, very competitive market. I suspect that lots of the competition has occurred as a result of the VCs funding startup companies. I would agree with the $20,000 per terabyte mark. I think Netezza was the one that actually put that bar in the ground having announced that price point first. But I would see that really as just a natural evolution of technology. I think the capability of hardware and software evolves over time, and the price point falls over time. The reality is that whilst we may talk about the cost per terabyte falling, the amount of data that people are managing is growing exponentially, so you could argue that the money that actually gets spent at the end of the day is probably about the same, if not more.
SearchDataManagement.com: Much has been made lately about data warehousing in the cloud, both private and public. What role do you see cloud computing – both internal, private clouds and public clouds like EC2 – playing in this market?
Young: Whenever you talk to somebody about cloud computing, you have to put in the caveat that it means 100 different things to 100 different people. I think that this is something that's going to become very attractive to a potential buyer, but I think it really is in its infancy at the moment.The only hesitation I have is that it's an incredibly complex environment, particularly if you want to deliver anything that resembles decent performance. So if you look at the way data analytics or advanced analytics is heading at the moment, people want to capture information about customers and analyze and take action on it while they still have that customer as a captive audience – either still in the store or on the website. Now, if you're using an external cloud service, just getting that data across the network into that cloud environment is tough, and then with the low levels of performance that the cloud environment has, it's not really a feasible solution. There's no doubt that the technology is definitely going to evolve -- that things like Hadoop and MapReduce have a very significant role to play -- but this is basically a technology in its infancy, and I think the talk of cloud computing is somewhat ahead of what can actually be delivered and indeed what people actually want to buy.
SearchDataManagement.com: As the New Year approaches, what can we expect from Netezza in 2010?
Young: Netezza's really going through a huge transformation at the moment. If you look at where we've come from, we've been basically a one-product company. We're using a hammer to crack a nut. As we move forward, we'll deliver a whole portfolio of appliance products that have different capabilities.We've already started to announce prepackaged appliance solutions. We announced the Kona solution with Kalido a few months ago. That's kind of like a pharmaceutical application in a box, ready to go, and we'll be making a lot more of those applications. And very significantly for us, we're going to put in place technology that allows systems to be federated. Today, the bottom line is that a TwinFin machine can't talk to any other TwinFin machine, or at least not without going through an ETL process; and, of course, that's not what customers want to do. So our larger customers have turned to us and said [that they need us] to fix some of these issues and make our technology deployable on an enterprise scale. So we'll be bringing out federation capabilities and software systems that allow us to better integrate our machines with things like Oracle. This time next year, we will have the portfolio of products for different types of applications. Those products will all be able to talk to one another with data replicated across the machines as appropriate, with data moving freely between the machines, with queries crossing those machines.