Interview: CA set to move on distributed computing

The distributed computing discussion has largely focused on vendors such as IBM, Sun, and HP. Now Computer Associates wants some of the action. In a conversation with InfoWorld Executive News Editor Mark Jones, Editor at Large Ed Scannell, and Staff Writer Brian Fonseca, CEO Sanjay Kumar and CTO Yogesh Gupta reveal their ambitions to see CA become a player in the computing-on-demand space.

InfoWorld: What is CA doing to address the emergence of distributed computing models?

Kumar: If you go back a year ago, I talked about this thing called Computing Everywhere and [was] telling people there’s this on-demand utility-pervasive grid, there’s many permutations and combinations of this. It’s like your phone system. No company, other than the telephone company, has a big telco switch in their basement because only Verizon can afford that and that’s what they do for a living. That’s where this industry is heading. Big companies will still insist on having their own [switch], but they will still buy services from somebody else. No one vendor will be able to conquer and own this marketplace. IBM has its strategy, Sun has its own, HP has its own, others are working on their own. The premise of computing on demand, autonomic computing implies a very heterogeneous platform to start with.

We’re all doing little things. We have in the next generation of UniCenter, for example, some autonomic stuff that worries about network. Let’s say you have file transfers in progress and a network connection dies. We’d pick up all that stuff and dynamically we move it and we re-establish it. So in our scheduling product, we’re working [on] autonomic things where if the job scheduling system is supposed to run a job here and move data and we tell it a path to move it from that’s not available, it’s going to go find another way to move it. [It’s] early stages, still early days, simple stuff. Every one of us has got a little thing, [and] as more of us do these things, over time things will get better. I don’t think any one person will have a nirvana strategy.

InfoWorld: How far down the road are you toward being able to offer some of the automated computing features companies such as IBM, Sun, and HP are touting?

Gupta: We have a wonderful road map and technologies coming out throughout this year. We have demonstrable technology today on resource allocation on demand. Today we do single-click resource allocation — it does require a human being to intervene [but] we believe that we’ll be able to do completely automated, policy-based resource allocation by end of this calendar year. We believe we are at least three years ahead of where the market is in terms of competition.

InfoWorld: What’s your take on companies that offer software on a hosted basis?

Kumar: We will not host, but I will tell you for sure we will allow more and more of our software to be offered by hosting companies. People completely underestimate the depth of the CA business model. Whether I sell you a three-year license and I book it one month at a time on that three-year license, or I truly get a monthly fee from a hoster, there’s no difference to CA. There’s a lot more depth there than people get.

InfoWorld: How do Web services play into this?

Kumar: In Web services we have two issues. One is developing technologies to support Web services directory and security in front of the directory in a Web service environment. Today customers are building Web services [and] the logic for deciding who is calling the service and what they can see is in the application. That’s horrible, that’s like building an application 20 years ago when the security was hard-coded in it. “Is it Bob? OK, allow him these transactions.” That’s nasty, but that’s what they’re doing. We’re building technology that allows a directory to decide what components of the service you [want] to expose. Let’s say you’re Fidelity and you have a stock quote service that’s published. You want one application for John Doe to come around to get a delayed 20-minute quote on the Web. If you’re a Fidelity customer and you logged in and security knows who you are, it’s a real-time quote. If you’re a Fidelity broker, maybe you will see the dealer book and bid, but same application.

No. 2, we have to re-host many of our technologies on a Web services platform. Because if we get to the model that you’re talking about, which is hosted technology, one of the things I would like to do is offer the hosting in a Web service environment. So let’s say security. We’re not there yet — and I’m not saying we’re close, we have work to do — but imagine a third-party service provider hosting an access control offering as a Web service, which is Web service-enabled, Web service-aware, and a Web service itself, that we get paid by the click.

InfoWorld: You’re implying that the great utility computing model is a lot closer than we think?

Kumar: I think a lot closer, but not in the technical sense of where people think it is. Most people are focused on [utility] computing and you’ve got big banks of computers. I think there are many more things that will happen way before then, and Web services can potentially unlock it. Linux will drive prices down, will drive more commoditization. A customer was telling me they replaced a specific brand of Unix servers and were going to Linux. They were able to increase their power of the server four times, they were able to cut the cost to one-fifth. Not every one of those boxes [moved] — for example, they’re running SAP and they can’t move [that] yet — but a bunch of things moved and essentially he said [he] cut the expense down and added way more capacity than he needed and [is] creating now [his] own kind of utility computing bank.           

InfoWorld: We’ve talked to people recently who are using Linux in self-healing networks to rebuild on-the-fly the kernel of the application and restore it. Are you doing that?

Gupta: We have done that. Self-healing is a very interesting thing. What plenty of people are talking about when they say self-healing is really [replacement], it’s not really healing. I find it disappointing because really if you want self-healing, you want things to find out a problem before it happens and fix it. We’ve offered some of the technologies historically in the network and systems management space. Automatic failover is really what this is. The ability to, if you have a whole bunch of Linux environment and servers [and] one of the Linux server fails, you [re-gen] another Linux server, bring it up online, apply it to the load that is being offered. And then while that is going on, you balance the load among existing servers, and then once this new thing is available, you re-balance it to the new server. We do that today. What’s interesting is we don’t limit it to Linux, we’re also doing it for Sun Solaris, we’re also doing it for Windows NT. I don’t think that this has to be only Linux.

InfoWorld: What’s your strategy when it comes to pervasive computing devices?

Kumar: I think [there are] two fronts. You have to look at the technology and you’ve got to adapt products for technologies. But No. 2, potentially you have to cannibalize your own business model. But I think it’s a little further out in that adoption sense. When customers think about this stuff and ask, “What are my priorities and what am I thinking about things, what’s exciting to me?” I think self-healing autonomic stuff is closer. Utility computing is a little bit further out and pervasive is a bit further out.

Gupta: To me, there are three major things that make pervasive computing happen: The devices themselves, the connectivity through wireless, and the integration [through] Web services. Those are the base three building blocks of technology that then basically lead to a computing model where there are completely dynamic transactions happening. When you do something, you’re not really saying, “I am going to connect to something” [or] “I’m going to use that particular application.” You’re doing something on your PDA, or on whatever device it is, and it is basically figuring out things in the back. It’s dynamically looking up what service is available that’s most appropriate for what you’re trying to do. In that environment, there are lots and lots of issues. There’s this whole issue of identity, this whole issue of security and privacy, this whole issue of management and service levels, and there’s this whole issue of accountability — if I try to use a service and it doesn’t really do what I told it to do, who do I hold responsible and who do I shoot?

InfoWorld: How have these technologies changed the way enterprises are making IT decisions?

Kumar: In large companies, I tend to see CEOs or CIOs independently. In a small to midsized company, we tend to see them together, generally.

InfoWorld: Do you see more involvement with CTOs?

Kumar: I think CTOs are much more focused on the architecture. They are much more aligned with business strategy than ever before.

InfoWorld: How influential do you think the chief technologist is becoming in the buying decision?

Kumar: I think in many of the point solutions and to fill out the gaps in the overall architecture, they’re becoming very influential. A company might have a strategy that says “Linux, Windows NT, and the mainframe are our platforms of choice.” The technology guy decides where to host the application. The CTO and the CIO decide “These are the standard platforms for the big businesses.”

InfoWorld: What drove CA to break up into six brand units?

Kumar: You’ve got to be able to compete on individual products with best of breed. Customers still prefer suites, but they want to know that every little piece in there is the best one, which is a change from three or five years ago. Today there’s a blend. The guy buying suites is saying, “I want to make sure everything in the suite is good.” And the person who is buying point solutions is saying, “For the first time I’m willing to look and consider some number of collective products from a vendor.” So I wanted to be in a position to offer both.