A confirmed technologist, Composite Software Chairman and CEO Jim Green seemed destined for the Silicon Valley, where he’s lived since 1974. “When I was getting out of school I was extremely fortunate to meet a guy who confirmed, against my intuition, that there really was a place called Sunnyvale, California.” Ever since, he’s found it compelling to “invent, create, build, have fun” and bring products to market that have generated more than $1 billion in revenue. Before joining Composite in 2003, Green was CTO and EVP at webMethods, where he arrived as the former founder of Active Software. When not spending time with family and children, “the most important thing in my life,” or taking an ocean kayaking breather, Green is likely to be talking virtualization, as he recently did with DM Review Editorial Director Jim Ericson.
DMR: Composite Software’s marketing almost reads like a manifesto on the move from ETL [extract, transform and load] and batch processing to virtualization. What can you tell us about the rate of change?
Jim Green: First, why are we talking about batch going into 2008? Everybody is moving toward service-oriented architecture [SOA], everyone’s trying to leverage the network. It’s very clear that the speed of business has accelerated, and we need to leverage all the computing horsepower that has emerged through Moore’s Law, which now can allow us to do things in real time versus offline because of [old] limitations in processor or network speeds. Those limitations are not what they were 18 months ago and were wildly different 36 months ago. Another obvious thing is that the schemas of the database were designed for application developers; they’re obtuse, sometimes obscure and not in a form intended for human consumption. So even as you move to real time, it’s really not possible to access data directly and have it make business sense. Therefore, you need to virtualize information so there is a cleaner, more business-focused presentation that can be mapped into back-end databases with proprietary access mechanisms. Think of your car. You can open the hood and see a big mess, or you can hop inside where the dashboard, wheel and pedals all make sense. We’re doing something analogous with virtual presentations that allow people to access and leverage corporate information in a more productive way.
DMR: Your message of virtualization doesn’t sound limited to the scope of business intelligence.
JG: If you look at the old style of data integration where the predominant tool was ETL, the target was to drop data into a data warehouse, a sophisticated staging mechanism that supported various business intelligence applications. With virtualization and real-time data integration, we can target a wide variety of applications that want to access data directly. Many of the people who use our tool are not running business intelligence applications. They just need to access the data and don’t want to build sophisticated manipulation into applications. They want to push that down into our server so applications are simpler and the work is better partitioned. It’s similar to people who write applications that talk to the Web. They found there is a whole lot of infrastructure required, and they push that down into an application server so their applications are much simpler.
DMR: Can you describe this in an example?
JG: Let’s take the scenario where a data-intensive application might pull data from multiple places. It might have to merge relational and XML data and deal with systems with different security mechanisms. If you’re dealing with packaged applications, you’re not doing a SQL fetch, you’re doing an API call, and there might be some digging around to be sure the data is complete and authoritative. It really doesn’t have anything to do with the business logic in the application. It has to do with bringing the data forward in such a way that the business logic can deal with it cleanly and simply.
DMR: What about the reported drawbacks of virtualization in terms of scale and performance?
JG: When you start from the business objective, say a single view of the customer, you usually find that the data you need is scattered. Orders are in one place, shipments are in another and payments in a third. You have to do some really exotic work to achieve high performance, and my argument is, with virtualization and a server like Composite, it will happen far faster than if you try to do it in your application. That’s because we’ve deployed lots of exotic algorithms and spent years studying the problem. An application developer wouldn’t even consider the techniques we use to improve performance.
DMR: Are you saying federated data is as scalable as traditional data structures?
JG: It can be, yes. We acknowledge that if you retrieve data from more than one place, it won’t be as fast as doing it in one place, so there are performance considerations. But you don’t have to guess. You can determine the performance you’re going to get from a system before you build it. This is not a leap of faith; it’s just math and can be computed and rendered to actual numbers fairly quickly.
DMR: Some people argue that virtualization actually makes things more complex for IT, given all the data movement.
JG: Complexity always exists, so the real question is, who gets exposed to the complexity? By implementing virtualization, IT is taking responsibility for managing complexity, but above that layer it’s simpler. I don’t know that you’re adding or subtracting complexity, but you’re certainly putting complexity in an area where the experts will maintain it as opposed to the business users.
Be the first to comment on this post using the section below.