Past And Future Collide At PowerUp 18
May 21, 2018 Alex Woodie
COMMON kicked off its first inaugural PowerUp conference Sunday in San Antonio, Texas, where an estimated 1,000 attendees are gathered to learn about the Power Systems platform and share their knowledge with other users. The IBM i server’s 30 years of reliable service was a big topic of discussion, but IBM executives Steve Sibley and Stefanie Chiras were determined to talk about Power’s AI future.
Any platform that lives to be 30 years old in this day and age is doing something right. While today’s IBM i server is not the same beast as the first AS/400 that came out of Rochester, Minnesota, back in June 1988, there’s a common digital DNA that links the machines across space and time. (In fact, the IBM i heritage goes beyond that, into the System/38 developed in the late 1970s and the System/36 that came out in the early 1980s. You can trace it all the way back to the System/3 in 1969 if you want to.)
That remarkable continuity is practically unheard of today, and is simultaneously the source of great pride and some frustration for IBM. During a panel discussion on the Future of the IBM i Sunday morning, a PowerUp attendee brought up the “L” word. Namely, why does it have to be so hard to convince technology decision-makers that it’s not a washed up legacy platform, especially when 5250 green screens are what users see?
That got Steve Will, the chief architect for IBM i, going a bit. “You think this operating system isn’t modern?” he said during the panel. “Give me a modern characteristic that IBM i isn’t, because I’m going to show you the characteristics and we’ve got it.” (Will is actually presenting a session on this topic Wednesday.)
“The difference between us and them,” he continued, “is that you can still, and many people still do, use us the way you used us in 1990. Nobody uses a PC with DOS anymore. Nobody would do that. But our community is willing to use our system as they’ve used it and we’ve allowed them to do that. And I don’t ever intend to change that. I don’t ever intend to take it away.”
That central contradiction – the juxtaposition of a system that is simultaneously backward compatible with applications originally developed on punch cards on the one hand, and on the other capable of running the industry’s’ most powerful deep learning algorithms in support of cutting-edge artificial intelligence applications – is part of what makes this platform so unique and so fascinating.
Obviously, Chiras and Sibley did not want to talk punch cards. Instead, the presentations they delivered during the Opening Session Sunday morning focused on the bright future they see for IBM i and the Power business as a whole.
“I hope you take great pride that you have worked to shape what IBM i is today,” said Chiras, whose title is vice president of IBM Power Systems offering management for systems of engagement. “But even more importantly, it’s about what IBM i will be tomorrow.”
The rapid proliferation of machine learning technologies and techniques is on the cusp of having a profound impact on established views about what business computing should be, Chiras said. “We use to run businesses with what we call process,” she said. You take a set of business rules for what you do today, you coded that in, you ran it through your computer, and you hoped that you could do that better and faster next time. And that the next generation it would run faster.”
But businesses are now looking beyond that type of process technology to find other ways to gain a competitive edge. “Now it’s about what insights can you get from the data. What can I learn from that data that’s buried in relationships that I don’t know in my business rules and process? How can those business rules and processes get smarter yesterday than they were yesterday?”
Companies like Uber have capitalized on this form of computing — what was originally called “big data” and which now many refer to as machine learning or artificial intelligence — and in addition to being very powerful for those who master it, it’s also changed consumers’ expectations for how they interact with businesses.
This is both good and bad news for IBM i shops. While IBM didn’t say it, it’s clear that IBM i shops that continue to look at technology through a green-tinted 1990s lens will likely struggle against their more technologically nimble peers that are leveraging things like predictive analytics and machine learning. But on the bright side, IBM i shops are in possession of something that gives them an edge in all this: a very powerful server that can do much more than run an RPG applications at close to the speed of light.
It’s about having a platform, Chiras said. “When people ask me what the persona of Power Systems is in the industry, I call it the enterprise platform for innovation,” she said. “It delivers what your business needs today and will continue to need tomorrow and gives you the opportunity to do cognitive and AI like no other platform in the industry.”
Chiras, who prides herself on being a chip person, held up a Power9 chip with a fair degree of wonder. “It’s an amazing piece of engineering,” she said. “It’s 8 billion transistors on this piece of real estate. Over 15 miles of wiring on this processor chip. It’s incredible what goes into this. What’s even more incredible is what it can deliver on the other side to the business.”
IBM has amassed a hardware advantage over X86 technology with its Power9 processor, which offers twice the performance as X86 on a core-by-core basis, according to IBM, and a 1.5X advantage over Power8. When you factor in hardware accelerators like GPUs from Nvidia and speedy interconnects from Mellanox Technologies, then you have a setup that’s extremely powerful. It’s no coincidence that the Department of Defense is currently installing just such a setup at Oakridge and Lawrence Livermore National Laboratories, which will use the supercomputer to ensure the efficacy of the nation’s nuclear weapons arsenal.
“Commodity is not good enough,” Chiras declared. “This era of commodity types of hardware is over. IBM i has never been this way. We took the hardware, we took the software, we pooled them together and it was better together . . . . And this era of AI is re-igniting that space in the industry.”
With Power9, IBM is well-positioned on the hardware front to benefit from a surge in interest in deep learning. The Power9 systems have a sizable performance advantage over Intel X86-powered systems to do the processing-intensive model training. However, a new wave of specialized AI-based hardware, such as Google’s own Tensor Processing Unit (TPU), is on the cusp of delivering orders of magnitude speedups in training.
What IBM would like to do is to get its IBM i shops interested in the software that unlocks the power of machine learning and AI. It has developed a collection of products, including PowerAI and the Data Science Experience, that allow developers to develop predictive models using open source deep learning libraries such as Tensorflow, Caffe 2, Torch, PyTorch, and Theanno. Considering how much transactional data that IBM i servers (and mainframes) hold, the idea that useful correlations could be just waiting to be discovered in one’s data, is a potentially interesting one.
The idea is to make the AI tools easy enough to use so “it’s as simple for your developers to develop new AI applications as it was for them to develop RPG or RPG Free Form applications,” said Sibley, whose title is vice president of Power Systems offering management.
“We are on the verge of this new era of computing that is coming,” he continued “And we liken it very much to what many of you or all of you carry around in your pocket. A mobile phone. Ten years ago Apple introduced the iPhone and it has changed the lives of most of us over that last decade. We think that when we’re meeting at COMMON ten years from now and you look at the application and the capabilities you have running in your business, you’re going to see the same kind of transformation in how you’re able to deploy application in new ways to change [your business] as well.”