Hands-off operation is not a new concept in data centers by any stretch. The term “Lights out operation” has been around for more than a decade, and most data centers are operated by just a handful of people whose main job is to fix broken hardware.
But with the advent of artificial intelligence (AI) and machine learning in recent years, server vendors have taken automation to a whole new level, using AI to do mundane or repetitive tasks and thus free up the humans to do more important work.
So is AI ready for prime time? Some think so. Anthony DeLima, head of digital transformation and U.S. operations for Neoris, a digital transformation accelerator, said AI is good for repetitive activities that are being automated. “Intelligent process automation has been around for years but has gotten to a level where you should let automated processes take over with repetitive tasks, rather than risk a higher error rate if you have a human do it,” he said.
“I’m more nervous about human error,” said Craig Wilensky, CEO of JASCI Software, a SaaS platform supporting ecommerce logistics and distribution. “We’ve had instances where staff are doing things and make a mistake. Sometimes those mistakes are big. Autonomy is designed not to make mistakes. I think these smart systems are just at the beginning and it’s just going to get better.”
[ Related article: Is RPA really an AI process or much less? ]
Jake Ring, CEO of data center provider GIGA Data Centers, said the hyperconverged infrastructure market has taken a lead in autonomy, which is why that market has seen such dramatic growth. HCI was supposed to be 35 percent of converged market by 2019, but by 2018 it was 46 percent.
“The ease of use and automation behind it means a generalist can do the job without requiring an expert,” he said. “You can take repetitive things people are doing and reallocate resources to something else. We have a human resource shortage in our industry, so the more we can free them up and apply them to customer and other initiatives, the better.”
[ Related article: How to future-proof your IT career in the age of AI ]
Oracle makes the push
No one is pushing this harder than Oracle. OpenWorld in 2018 was all about the Autonomous Database, with its self-tuning to optimize performance. In 2019, Oracle was pushing autonomous servers and Oracle Linux, which would be self-correcting and optimizing and self-patching. The whole Oracle stack, from hardware to apps, is now self-patching and self-optimizing.
“The tech is ready for prime time. We have thousands of customers looking at it, whether they go all in and let their hands off the wheel or maybe say I want to review a security patch before I apply it,” said Steve Daheb, senior vice president of Oracle Cloud.
[ Related: Is AI judging your personality? ]
JASCI Software's Wilensky is sold on Oracle automation, saying autonomy has resulted in 100-fold performance gains on the database alone.
“I can probably say we’re saving 20 percent of our energy internally by not having to focus on infrastructure,” he said. “Now we can look at our staff where we once had a portion of them dedicated to managing infrastructure and the database and updates. Now we’ve found Oracle has brought their expertise to the tables so we don’t have to worry about those things. I can allocate funds and attention to what we do best.”
Oracle is hardly alone in this initiative, just more aggressive in its marketing. HP Enterprise (HPE), Dell Technologies, Lenovo and Cisco have all touted AI operations in their data center equipment as well. Bob Moore, director of server software and product security at HPE, said autonomy is leaps ahead of the older technologies.
“AI has been around for some time in some form or fashion,” he said. “The difference now is we’ve got AI and deep machine learning that provides a lot more capabilities than we’ve seen in the past. These algorithms are really able to deliver on a journey we couldn’t make 10 or 15 years ago.”
HPE’s primary efforts are built around its InfoSight predictive analytics, which spots issues, security or quality, and makes performance recommendations. “AI can let you know a component is failing long before it does, make assessments to tune performance to much higher capability, and spot bad behavior. All of those capabilities are new and relatively new,” said Moore.
In the future, for server management, HPE is looking forward to the time when AI will fix problems before they even happen by automatically opening a support case and ordering spare parts and having them delivered to be installed long before hardware fails, Moore added.
[ Related article: Only AI can save us from a world of fakes (a world AI is also creating) ]
Can we all get along?
One issue hanging over autonomy is heterogeneity. It’s one thing for Oracle to automate its own hardware, OS, database and application layer. But what happens in virtually every environment where you have a mix of brand names?
“It does get harder with heterogeneous environments, that’s why you’ve got to be careful with the scripting you put in place,” said Ring. Autonomy “requires homogeneity. Yes it will be autonomous within our platform, but once you go into other plats, well it depends. It’s not just an API hook. It becomes a custom effort that takes more time and expense.”
Moore says HPE recognizes that the customer is likely in a hybrid IT environment and not a homogeneous environment. “So autonomy is probably easier [in a homogenous setting] but we do believe that AI can extend into a heterogeneous environment and that’s what we’re working to do,” he said.
[ Related article: Should you tell customers they're talking to AI? ]
Security is a Driver
While there are several drivers for automation, Deheb said there is a lot of interest around security breaches, because in as much as 85 percent of the cases where a breach occurred there was a patch available for up to a year before it happened.
“When you look at what’s going on from a cybercrime perspective, you have bad actors using advanced persistent threats,” said Daheb. “They are using automation to break things down. If all you are doing is throwing human resources at that, it can’t scale. You need machine to fight machine. So in these cases AI will be more ready for prime time than people are in related to security.”
But patching has a caveat: The patch better be solid. On more than one occasion this year Microsoft has been forced to roll back or remove Windows patches because they broke PCs. It’s not a fair comparison, of course. Microsoft has to support countless thousands of PC configurations. Oracle, HPE and Dell have to support only their own hardware.
So trust in the integrity of the patches is necessary, say the IT executives. “It’s got to be where you can show a history and track record of making the automation work seamlessly so we can truly take our hands of the wheel, or at least a Tesla situation where I have to keep my hands close to the wheel in case I have to course correct quickly,” said Ring.
“There will probably some horror stories and setbacks in 2020,” said DeLima. “I think it’s likely that those setbacks would not be caused by infrastructure issues, they will be because algorithms didn’t behave the way we predicted they would behave, or a hardware failure. But the biggest impact will be intelligent algorithms not performing as predicted.”
“At the end of the day, you want to leave core decision-making to the folks. Some things I would want [autonomous systems] to do for me, like security. I would look to it for recommendations, but ultimately make the decision myself,” said Daheb.