Bob Eve, EVP Composite Software and author of the new book, "Data Virtualization"
Why did you write the book?
There's certainly an opportunity in the marketplace for people to improve their business agility, for mergers and acquisitions, regulatory purposes, competitive things, all those are drivers for changing your business. To do that you need new information and our customers are using data virtualization to help make that happen. It's important to get that story out so the pragmatist users, not just the early adopters can understand what's possible. We saw at our data virtualization day a couple weeks ago in New York that our customer champions have the same issues I have. They're just selling the people, process and technology change that has to occur internally, and I thought, 'we're in the same boat.'
We constantly attach terminology to tech marketing, so how does data virtualization connect passionately to business versus all the things we've called it in the past? Why is it relevant now?
That's an important question. The first thing business knows is change and we all understand the importance of information alignment, whether it's changing a business process, using new information etc. We understand there are a lot of ways to get at new information. You can consolidate it in a warehouse, you can replicate it, you can virtualize it. That's a choice everybody has.
I think what's different about data virtualization is the time to solution and agility as well as the more effective use of resources, and there are multiple examples of those things in the book. The iPhone 4S just came out and you can imagine how long [our customer] Qualcomm has been working on those chips. They have had data virtualization implemented in their IT organization for two years, completed 24 projects, they saved $2.2 million on the first five projects. Today, business comes to IT with a request, they pop back a business view to them right away and now IT waits on the business as opposed to the other way around. That is a different, dynamic to do business. There are resource savings in that you don't have to stand up as much hardware and you have a more streamlined iterative process. The iterative process is what you need to move quickly where decisions need to be made quickly, and in onetime decisions like we often see in R&D.
Are you selling data virtualization as a one size fits all thing or is it just another tool in the box?
From an IT point of view or an integration competency view it should be viewed as another tool in the box. As each business problem comes along you go at it with the full toolbox. I think our customers often take a hybrid approach, they are using some consolidation, some replication, some virtualization. Or maybe they are virtualizing previously consolidated content or they're extending consolidated content with external content or they're in M&A and want to virtualize the entire organization and the whole research pipeline. So it's sort of, 'why would you not want this tool in the box?'
Yet when it comes up in a meeting, executives will say, 'didn't we already do this?'
Let's compare virtualization to a data warehouse consolidation. The great thing about a data warehouse was one place to go and a business view of the data. That was the fundamental business proposition, and who wouldn't want that? Unfortunately, one place to go has been proliferated into multiple silos because of big data and analytic appliances and we never did get quite everything into the warehouse and the business view is not just a dimensional model or fact table.
There is a widening out of information business needs. We've tried to simplify the complexity by saying it can be one place to go except it can be a virtual place and we'll figure out whether we go get the data from the warehouse of whether it's from an outside source etc. Then that unified business view turns into building business views, canonical models. One of our customers, an oil company [I write about] in the book has over 600 canonical models built into 30 domains. They've created standards for wellheads and what a pipe is and certain key business entities that get used in multiple reports in multiple uses. That standardized layer is what the business application, the BI application can see.
The big data theme here at the conference implies that we have to go everywhere and cook the ocean to figure things out. Federation implies that too, that we can go anywhere but can we determine if it's worth our time?
Making it real is important. Certainly big data silos with a tightly bound analytic engine are awesome for a certain use case, for unique things and understanding trends etc. But they tend to be isolated. They are purpose built. The kind of business case we run into, for example at Putnam Investments, they market their mutual funds through advisors to end buyers. They do a "Wall Street Journal" ad and want to know the impact of that campaign.
Sure, there is some clickstream data they are analyzing in a Hadoop store, but what was the investment advisor behavior they pick up in Salesforce.com? What if they combine that with an email campaign in Unica? They have some other information in older systems they haven't entirely migrated off of. They want the 360 view of the advisor, and also the impact of the marketing campaign and whether the ad was successful. It takes a whole set of joins to answer that question. The click stream on their website is just one part, the important thing is whether the ad was successful.
