Apr 1 ’05
Mainframe Perspectives: An Interview With Computer Associates’ Guy Harrison
Guy Harrison, senior vice president and principal architect at Computer Associates (CA), is an authority on mainframe computing. Before joining CA last year, he spent 29 years at IBM, where he was involved in the ongoing development of MVS and oversaw the transition to CMOS technology. He ran the MVS program for much of his tenure at IBM. Harrison is an outspoken advocate of mainframe computing. As principal architect at CA, he’s working on software tools to simplify the management and operation of mainframe computers. Early this year, z/Journal interviewed Mr. Harrison.
z/Journal: The experts regularly declare the mainframe is dead. Why doesn’t the mainframe die?
Guy Harrison: Frankly, sometimes we’re way too quick to throw the term “expert” around. When I ran MVS for IBM in the late ’80s and early ’90s, all the experts declared the mainframe a dinosaur headed for extinction. I said it was incorrect at that time and it’s still incorrect.
The reason the mainframe doesn’t die is Reliability, Availability, and Serviceability (RAS). In the ’80s, cost was a fair concern, but since the mainframe moved to CMOS technology, the mainframe has been able to chase technology price/performance gains just like other systems. This made a substantial difference in the cost equation on the hardware side.
Today, the mainframe is positioned to take advantage of technology gains in storage, memory, and CPU. Costs are dropping and continue to do so. Plus, you can run a mainframe at 90 percent CPU utilization. Try that with other systems and you would have a disaster pending.
Finally, all major applications today run on all platforms. You can pick the application you want and almost always find it for the mainframe.
So, what are the reasons people choose a platform: cost of hardware, applications, and cost of the people. If you look at the characteristics of the different platforms and consider cost and capabilities, the mainframe has become more cost-effective and has the application capabilities and operational characteristics that companies still want. That’s why the mainframe isn’t dead and has a bright future.
z/J: Are new customers coming to the mainframe?
GH: Yes. New customers come to the mainframe for the reasons I mentioned previously: competitive cost, RAS, high CPU utilization, etc. But at the heart of the mainframe business remains the same customers, or at least the same kinds of customers, such as financial services firms, although the names of the customers may have changed due to acquisitions, mergers, and such.
A better way to look at this question is to ask where the new application workloads are going and whether the mainframe is getting its share. The answer is yes; the mainframe is getting its share and more. This is because applications are being written to run on a variety of platforms. Once written, you can run the applications where you get the RAS and utilization characteristics you want. For any decent size company, the mainframe will be the best place to run these new workloads based on its operational characteristics. However, this may not be the case for smaller companies due to the mainframe’s complexity.
z/J: Mainframe complexity is another complaint cited by the “experts.” Is it possible to simplify mainframe operations?
GH: This issue presents a real opportunity for CA and other vendors. We want to reduce the number of people needed to run mainframes, especially the highly specialized experts needed to operate the mainframe. We have these people now. Some are really talented; they know network management, operations, everything. They know everything about the tools and the implications of what the tools find. But these people are getting older and we haven’t been attracting younger people. They don’t think mainframes are a good career choice. Our job at CA is to take this complex environment and integrate the knowledge that these experienced people have and bring it to any person, so they can see, enterprisewide, what’s happening and be able to run the mainframe environment. Basically, we need to create software that will let companies reduce the number of (highly specialized) people required to run a mainframe.
Autonomic computing is also an answer to this problem. This is where the systems can, to a great extent, manage themselves. However, autonomic computing won’t happen tomorrow. It’s going to take time. Meanwhile, the combination of software—including that CA and others are working on—and education can help.
z/J: In terms of education, colleges aren’t focusing on the mainframe. How will mainframe shops replace their aging mainframe staff?
GH: The universities never did much in terms of the mainframe and nothing really has changed. Universities don’t like to teach specific platforms, especially one that doesn’t give you the source code or let you change it. Unix was ideal for universities. But in the ’60s and ’70s, all the important work was on mainframes. Everybody wanted to be a mainframe guy. Ten to 15 years ago, that changed. The mainframe has suffered by being viewed as not sexy. For kids coming out of school, the mainframe wasn’t the game they wanted to play. This problem has been growing.
The skills issue is important. We need to find people to run these systems. CA, IBM, and others are focusing on this issue. IBM is reaching out to work with universities, so they’ll attack it through the university system. They’re even giving some universities mainframes, so students have one there. Also, companies such as CA are providing software and services to help customers deal with continuity. Finally, customers will pay differently to attract mainframe skills.
z/J: Open systems has become a hot buzzword. Do companies need to open up their mainframes? Can they?
GH: What you’re asking is whether mainframe customers should try to take advantage of open systems. Yes, they should consider it. However, there aren’t a lot of significant Linux applications on the mainframe yet, but that will change. Part of the issue with Linux is RAS. The zSeries is 40 years old. For 40 years, IBM wanted to make a large-scale commercial RAS platform and CA makes it more manageable. Linux is just getting into the RAS game. My advice: don’t attempt to put your family jewels on Linux at this time, but if you can find a set of functionality you can split out on zSeries to run on Linux, go ahead and try it.
I’ll say the same thing about Java and the zAAP engine. We’re testing our software on zAAP. This is a question of new workloads. Java is popular with the new workloads, but Java is interpretive, so it’s CPU-intensive. With zAAP, IBM puts Java on a separate processor. So, if I’m a mainframe customer and I’m looking at new technology for cost/benefit advantages, then Linux and Java make sense and zAAP is a good way to do it. But again, don’t touch the family jewels.
z/J: What role will the mainframe play in today’s real-time, on-demand enterprise?
GH: A mainframe can respond fast to an on-demand workload; the mainframe is the natural system to choose for that. It’s much more dynamic than many people think. Underneath the box of a zSeries are a ton of engines and components that let you plug in lots of things. This is what lets you do on-demand kinds of operations.
If you had to build all this in a distributed system or open systems, you’d need to build all the redundancy and RAS stuff from scratch. Then you’d have to spend years testing it to prove its reliability. With the mainframe, all that is already there and ready to use now.
z/J: How do you see the mainframe going forward in terms of hardware and software?
GH: The mainframe will become more dominant and more central to computing environments. If a customer has the skills in-house, the mainframe will be the most cost-effective computing method. The mainframe also will increasingly become the application platform of choice.