Mainframe systems could play a significant role in cybersecurity and artificial intelligence advancements in years to come and IBM is investing in those areas to ensure system Z mainframes have a stake in those growing tech markets.
IBM mainframe sales grew some 69% during the second quarter of this year, achieving the highest year-over-year percentage increase of any other business unit. Some industry observers attribute the unexpected performance to the fact the z15, introduced a year ago, is still in its anticipated upcycle. Typically, mainframe sales level off and dip after 12 to 18 months until the release of a new system. But that might not be the case this time around.
Ross Mauri, general manager of IBM's Z and LinuxOne mainframe business, discussed some of the factors that could contribute to sustained growth of the venerable system, including IBM's acquisition of Red Hat, the rise of open source software and timely technical enhancements.
Mainframe revenues in the second quarter were the fastest-growing of any IBM business unit, something analysts didn't expect to see again. Is this just the typical upcycle for the latest system or something else at work?
Ross Mauri: A lot of it has to do with the Red Hat acquisition and the move toward hybrid clouds. Consequently, mainframes are picking up new workloads, which is why you are seeing a lot more MIPS being generated. We set a record for MIPS in last year's fourth quarter.
How much of it has to do with the increase in Linux-based mainframes and the growing popularity of open source software?
Mauri: Yes, there is that plus all the more strategic applications [OpenShift, Ansible] going to the cloud. What also helped was our Capacity On Demand program going live in the second quarter, providing users with four times the [processor] capacity they had a year ago.
Some industries are in slumps, but online sales are up and that means credit card and banking systems are more active than normal. They liked the idea of being able to turn on 'dark' processors remotely.
Some analysts think mainframes are facing the same barrier Intel-based machines are with Moore's Law. Are you running out of real estate on mainframe chips to improve performance?
Mauri: What we have done is made improvements in the instruction set. So, with things like Watson machine learning, users can work to a pretty high level of AI, taking greater advantage of the hardware. We've not run out of real estate on the chips, or out of performance, and I don't think we will. If you think that, we will prove you wrong.
But with the last couple of mainframe releases performance improvements were in the single digits, compared to 30% to 40% performance improvements of Power systems.
Mauri: In terms of Z [series mainframes], they are running as fast as Power. We know where [mainframes] are going to be running in the future. As we move to deep learning inference engines in the future, you'll see more AI running on the system to help with fraud analytics and real-time transactions. We haven't played out our whole hand yet. The AI market is still nascent; we are very much at the beginning of it. For instance, we're not anywhere near what we can do with the security of the system.
Ross MauriGeneral manager, IBM's Z and LinuxOne mainframes
We have started to put quantum encryption algorithms in the system already, to make sure security was sound given what's going on in the world of cybersecurity. You'll see us continue to invest more in the future when it comes to AI. We'll build on that machine learning base we have already.
Is IBM Research investigating other technologies that would sit between existing mainframes and quantum computers in terms of improving performance?
Mauri: Our [mainframe] systems group is working closely with the quantum team as well as with IBM Research. We are still in the research phase; no one's using them for production.
What we're exploring with IBM Research and clients is trying to determine what algorithms run well on a quantum computer for solving business problems and business processes that now run on mainframes. For instance, we're looking at big financial institutions where we can make use of quantum computers as closely coupled accelerators for the mainframe. We think it can greatly reduce costs and improve business processing speed. It's actually not that complex to do. We're doing active experiments with clients now.
What are you looking at to increase performance?
Mauri: We are looking at a whole range of options right now. We have something we do with clients called Enterprise Design Thinking where they are involved throughout an entire process to make sure we're not putting some technology in that's not going to work for them. We have been doing that since the z14 [mainframe].