DCW Keynote: Apollo Expert Frank O’Brien - A Realist for 2020
by Danny Bradbury
“I’m coming off like a crotchety old man,” says an apologetic Frank O’Brien halfway through our interview.
O’Brien, who will keynote at Data Center World--A Virtual Event on Aug. 27 at 11:30 a.m. (EST), is trying to come up with some predictions for the data center world in 2020, and he’s having trouble. It’s not through a lack of insight; It’s because he’s seen more history than most of the people working in the industry today, and he’s more of a realist than most.
When other kids were reading Boy’s Life, the 13-year-old O’Brien was reading a set of Apollo Guidance Computer manuals that his father gave him for Christmas. He earned a computer science degree and worked as a corporate programmer at Sperry-Univac, under contract to Bell Labs where he wrote a mainframe version of Unix. He spent nearly 40 years at Colgate-Palmolive, first as an IBM MVS system programmer, and latterly as an administrator working on databases and SAP systems.
He never abandoned his initial fascination with spaceflight, though. After writing the definitive book on the Apollo guidance computer, he became an editor of the Apollo Journals on the NASA website. He has also volunteered for NASA’s Jet Propulsion Laboratory for the last 25 years.
So, when he hears breathy industry predictions about how AI will take over the data center or everything will run on a blockchain, his eyes roll. “I’ve been set up and disappointed for so many years on so many different technologies,” he says, adding that the outcome is often the same: “Yeah, boy, that sounds like a great prediction. Let’s see how it goes in a couple of years. And of course, you know, a couple years later, you never hear of it.”
He singles out AI-operated data centers for particular skepticism. It’s limited to managing specific, highly constrained tasks, he points out, which makes it difficult to get any real traction in the data center.
“Right now, we have fantastic tools, we really do,” he says. “Honestly, they’re pretty good at the management, to monitor performance, and to take care of routine troubles. Like blockchains, AI is a solution for looking for a problem."
O’Brien may not buy into the more outrageous predictions for the future, but that doesn’t mean everything will stay the same, he points. out. Technologies will change the world because development is speeding up. The problem is that we don’t know which ones will prevail or how people will use them. He knows one thing for sure, though: Data center managers must prepare for an uncertain world.
In a world where data center managers don’t know where the next innovation is coming from or what it will look like, they must prepare for anything, he says. That means building data centers that ready for any development that comes along.
“They have to be agnostic about what is being connected into their network. Much of this stuff is not going to be designed or managed by the data center,” he warns. “It is all going to be man-aged by people further out in the user world.”
That means building IT architectures that extend beyond the enterprise data center to meet those challenges, O’Brien says, adding that the linchpin will be data communications infrastructure. Ninety percent of a data center’s infrastructure challenges in 2020 will be in the network. “That turns out to be the real sticky wicket,” he says, thanks to one overarching prediction that he makes: The continuing advance of edge computing, driven by technologies like IoT.
Companies will move away from centralized data centers to more distributed operations around the world as network technology and IoT systems evolve to support this model, he says. The requirement for low latency isn’t new — it’s something he understood when he worked at Colgate, a company with a presence in 200 countries — but it’s mounting.
“When you’re talking of a plant in Asia integrating with the local Google data center in Washington, that traffic mounts up in terms of response times,” he says, adding that you’ll see these problems crop up in IoT deployments, where sensors and control units might interact directly with production equipment and products rolling off the production plant. “You will have to abandon having a centralized cloud data center and start putting the processing much closer to the user."
Analyst reports bear this out. According to Gartner, around 10% of enterprise-generated data was already created and processed outside a traditional centralized data center or cloud environment in late 2018. That will jump to 75% by 2025, the company says.
O’Brien predicts that most of this will happen through the cloud providers. Large enterprise companies are unlikely to build out their own mini-data centers in far-flung locations, especially where their revenues don’t exceed a certain threshold. Instead, they will take advantage of those built by dedicated cloud service providers for whom extended geographical coverage will be-come a core competency. Expect to see large cloud players stepping up with more data centers as they realize that they need to provide lower-latency services in multiple regions, he tells us.
This doesn’t mean that centralized infrastructure in head office is going away, he says. Rather, it means that most infrastructure investment, especially in networking, will grow outside the central enterprise data center.
To help manage these more complex global networks, the really smart money will be in SD-WAN technology, O’Brien says, adding that it will make a significant difference for companies grappling with these issues in the next few years.
“A bazillion different vendors are offering their own idea of a product. That will mature, that’ll settle down, I don’t have any doubts on that,” he says. “But SD-WAN is the only way you’re going to be able to manage this. You won’t be able to deliver the flexibility to handle all these unique cases coming down the line without some form of SD-WAN.”
IDC’s research bears him out. In July, the organization published its SD-WAN Infrastructure Forecast. It predicted that the market for this technology will grow at a 30.8% compound annual growth rate (CAGR) from 2018 to 2023, reaching a heady $5.25 billion.
The move to cloud-based infrastructure at the network edge will have several side effects, warns O’Brien. One of them is the drive for greater cloud interoperability. As large enterprises rely increasingly on the cloud not just for centralized computing but for edge-based use cases, they will drift further toward multi-cloud environments, he predicts. The drivers will include avoiding vendor lock-in, improving reliability, and forcing better performance through competitive stresses. “If I was in a C-suite position, I wouldn’t want all my eggs in one basket,” he says.
However, this means that cloud providers must solve interoperability problems for their clients. “I shouldn’t have to hire a separate staff to manage an Amazon cloud versus Microsoft, because all of a sudden the value of putting all that stuff outside my data center goes away.”
The increasing popularity of open source software will help drive that interoperability, he predicts. Open source continues to move further up the cloud stack with developments like Open-Stack (whose developers are increasingly exploring edge-based use cases), Docker, and Kubernetes.
Cloud management tools like these, which hide the nuts and bolts of system hardware management from administrators, may have their own adverse future effects, he warns; they will change the job landscape for sysadmins.
“When I first started, for a medium-sized data center, we had half a dozen people running our operations on shift work. That was for two mainframes,” he says. Forty years later, he sees four people running around 1,500 virtualized servers. The consolidation of administration staff will continue and speed up, driven by increasingly easy-to-use tools that can govern large numbers of virtual machines at once.
“In the next couple of years, you will see — I don’t want to say the death of the system admin or operations shift person — but it’s getting kind of close to that,” he warns. Those people will have to move into different roles, particularly around automation. Instead of grappling to keep a single server running and back up its data, they’ll switch to managing large collections of virtual systems and configuring automated tools that can monitor and maintain the health of the herd. They must do this in close concert with the developers writing the software to run on those servers. They’re moving from pure operations roles to DevOps.
Those operations workers will also administer more infrastructure than ever before, spread be-tween many regions. The increasing use of cloud service providers spread around the world, combined with their own enterprise data centers, will boost the adoption of hybrid cloud technologies. “It will keep on growing,” he says.
New technologies are coming, O’Brien says, along with new ways of using them we haven’t even thought of yet. The smart data center manager in 2020 will focus on adaptability, so they can meet the new challenges that businesses throw at them.
What’s the first step that managers should take to prepare themselves? Start with the network, he concludes, adding one final afterthought: “Your bits always have to get home.”
Danny Bradbury is a technology journalist with over 20 years' experience writing about security, software development, and networking.