Artificial Intelligence in the Data Center: Seven Things You Need to Know

Artificial intelligence and machine learning are taking over both routine and advanced tasks. Will managers and employees be pushed out?

Most IT leaders think they have a complete grasp of data center management, operations, and planning. In reality, they don’t.

[[253489]]

No single IT leader or team of IT experts can exert precise (or even finer) control over the critical tasks of a data center—down to the second. Humans (even highly educated and trained ones) are often blinded by personal preferences, biases, and misunderstandings, and are unable to form a clear view of future planning and other important responsibilities.

Artificial intelligence (AI) has no such drawbacks. That’s why, even as data center operators grapple with hybrid environments, IoT and other challenges, they need to consider the impact AI will have on a host of critical data center operations and services.

Here are seven things all IT leaders need to know about how AI can turn data centers into more powerful, more efficient facilities.

1. Many different types of data centers can benefit from AI

Joe Merces, CEO of Cloud Daddy, a provider of enterprise backup and disaster recovery technology, and former CIO of the New York City Law Department, said all types of data centers can benefit from AI, but the ones that stand to benefit the most tend to be large facilities, such as large enterprise data centers, public cloud data centers, colocation hosting and outsourced data centers.

Tom Coughlin, a fellow at the Institute of Electrical and Electronics Engineers (IEEE) and president of data storage analytics firm Coughlin Associates, believes that all data centers can use AI methods such as machine learning to better manage internal resources and predict upcoming hardware and data needs. He noted: "AI is becoming one of the most important (data center) applications."

Paul Mercina, head of innovation at Park Place Technologies, a data center maintenance services provider, explains that machine learning is moving from basic pattern recognition and traditional algorithms to the more complex field of deep learning. “A key contribution of machine learning is its ability to discover structure in data using an iterative approach without requiring humans to start with any theory or hypothesis to test,” he says. Deep learning uses multiple layers of artificial neural networks to achieve extremely high accuracy in tasks such as object detection and classification, speech recognition, and language translation.

2. Artificial intelligence helps data centers improve energy efficiency

Over the past few years, AI tools have played an increasingly important role in reducing data center energy consumption and waste. “These applications help reduce power consumption, report on cooling inefficiencies and analyze the health of mission-critical systems to improve efficiency and save energy,” Mercina noted.

“The environment in data centers is constantly changing,” said Stijn Grove, managing director of the Netherlands Data Center Association. He said artificial intelligence can monitor current indoor and outdoor temperatures and predict future weather, allowing data centers to optimize cooling resources and save energy.

Servers are the biggest consumers of electricity in any data center. "When you want to automatically scale up or down cloud servers when needed, you can save even more energy by fully utilizing the potential of each server and shutting down unused capacity," Grove said.

AI can also significantly reduce storage energy consumption. By using AI-powered monitoring and analysis to predict the activities of various types of users, data centers can quickly move less frequently used data to storage sources with lower energy consumption and frequently used data to storage sources with better performance. "In addition, it is possible to use AI to minimize the amount of data that is moved back and forth during processing," Coughlin said. "Data in use can be intelligently laid out, which can bring data closer to where it is being processed, reducing the energy consumed by excessive data movement," he explained.

3. AI can improve data center security

The security needs of data centers are changing rapidly. Until recently, the biggest threats to data centers came from internal employees or relatively primitive brute force attacks from outside. "Today, hackers are creating AI-based algorithms that try to find weaknesses in data centers," said Param Vir Singh, associate professor of business technology at Carnegie Mellon University's Tepper School of Business. He noted that AI is the best technology to meet this challenge.

“AI applications enable data centers to adapt faster to changing security requirements while providing a safer environment for users without enforcing rigid rules,” Mercina said. “AI solutions also help detect malware and spam, analyze normal and abnormal activity patterns, identify weaknesses and strengthen protection against potential threats.”

AI can also isolate malicious intrusions in a 'honeypot,' Coughlin said, "where they can be closely monitored and even tracked down."

4. AI can optimize data center performance

AI can enable companies to run data centers at peak efficiency by constantly monitoring and adjusting resources, including where data is processed, networks and memory. “AI can be used to monitor load distribution, make infrastructure more scalable, and optimize efficiency in cooling and power consumption,” Merces noted. AI can also be used to optimize server configuration and utilization. “For example, AI can detect infrastructure problems and self-heal by moving loads and trying to fix them by restarting, cycling and re-imaging,” he said.

Coughlin believes that AI is particularly effective in optimizing server usage. “This could include directing the right processing power to processors that are specific to an application, such as GPUs and TPUs,” he said. AI can also optimize the performance of data center software. “For example, limiting polling of the same data in a database or limiting repetitive processes,” he added.

5. AI will improve infrastructure management

A Ponemon Institute study said the average cost of data center downtime across industries was about $8,850 per minute in 2016. "If we can predict maintenance issues, we can take preventative measures," Singh said.

Using ever-improving infrastructure management technology and smart sensors, neural networks can be trained to analyze the demand and capacity of existing infrastructure in order to meet that demand with the most appropriate equipment. “Because AI can process more information than a person or group of people, and do so almost instantly, AI-driven systems are more efficient and reliable,” said Tuck Northman, a partner at law firm Tucker Ellis who specializes in business and corporate law. Sensors can also help data center managers predict or mitigate catastrophic failures, he noted.

Mercina noted that today, most data centers are managed, monitored and maintained by humans who are trained to perform routine tasks, such as walking through rows of data centers searching for lights that indicate hardware failures. “AI and machine learning have the potential to revolutionize this outdated paradigm by removing the guesswork and empowering the entire ecosystem,” he said.

AI is expected to have a significant impact on the way data centers schedule routine maintenance tasks. AI will soon be able to predict when specific facilities will require service, upgrades, and replacements, simply by carefully reviewing all relevant data center resources. As a result, scheduled maintenance schedules will gradually be replaced by AI-generated recommendations, Grove predicts. “This will improve uptime and reduce costs,” he reports.

6. AI is becoming a powerful data center planning tool

One of the most exciting applications of AI is in data center planning. Northman says that by extracting a large amount of information from sensors in the data center and using the ability to learn from past situations, AI can provide sophisticated predictions and, more importantly, model differences in modified assumptions. "The longer the system is in place and the more information is available, the better the predictions that AI can make," he says.

“This is happening today,” Merces reports, with AI being used to plan and allocate power resources and predict cooling needs, for example. “AI is also being used to plan and manage network and bandwidth utilization and optimization,” he notes.

7. AI will manage more and more data center tasks with little or no human involvement

Grove said AI has a good chance of completely usurping data center tasks currently handled by humans. “The digital ecosystem requires more instant control and action that can only be achieved with AI and machine learning,” he said. “Also, with the advent of edge computing, you need AI to do things like this correctly to be able to manage large numbers of unmanned data centers.”

A fully automated data center that can monitor, diagnose and self-heal is the stuff of dreams. “That requires artificial intelligence, robotics and even augmented reality — machines helping each other,” said Roger Brooks, chief scientist at Guavus, a big data analytics company.

On the bright side, AI still won’t be able to reliably perform high-level reasoning and decision-making tasks, at least from a human perspective, with even a low degree of reliability. “AI will do jobs big and small, it will be divided into specific functions, and while those functions may be extremely efficient, in the end, they will not be intelligent,” Merces predicts.

Northman agreed, saying, “While managers will increasingly rely on AI to operate and manage data centers, I don’t foresee humans being completely removed from the process. While managers will no longer have as much responsibility for certain data centers… humans will continue to play a role in preventing failures.”