Skip to main content

Prepare Your Data Center for AI

Jim Kandrac of UCG Technologies explains how AIOps will change data center operations.

An illustration of a brain hovers over a blue microchip.

If you’re not familiar with artificial intelligence operations (AIOps), it’s a good idea to start reading up now because you’ll be hearing a lot more about it in the future.

AIOps is the application of the tools of big data and automation to IT operations. The artificial intelligence (AI) component is the machine learning algorithms that can process data from multiple sources at a massive scale to identify problem areas, and more importantly, predict where problems may develop.

IT organizations need to develop the skills and infrastructure necessary to support AIOps if they are to realize the promise of digital transformation.

Data at the Center of Digital Transformation

Digital transformation is about recasting operations and customer relationships around data and digital connections. Although the concept has become a bit of a buzzword, it has captured the imagination of organizations, 70 percent of which say they either have a transformation strategy in place or are working on one (zd.net/2vqojwJ).

Organizations that put data at the center of their business will need to increase the speed and scalability of their IT operations by orders of magnitude. This will require huge changes in both technology and culture. Traditional on-premises operations infrastructure won’t be robust enough. The speed, scalability and availability demands of digital business will require organizations to use multiple clouds—both public and private. This will create added complexity, the management of which will be beyond the scope of human operators.

Insert AI for Added Efficiency

That’s where AI comes in. The same tools that IT organizations are bringing to bear on the efficiency of their businesses can also benefit their operations. Gartner, which coined the term in 2017 (gtnr.it/2RPZkjQ), describes AIOps as the application of “big data, modern machine learning and other advanced analytics technologies to directly and indirectly enhance IT operations (monitoring, automation and service desk) functions with proactive, personal and dynamic insight.”

Gartner predicts that 40 percent of enterprises will have adopted AIOps tools by 2022 (gtnr.it/2T4HDKJ), up from 5 percent today, and notes that vendors are rapidly adding these analytics capabilities to their products. In fact, the new breed of hyperconverged and composable servers that the big hardware vendors are bringing to market presuppose the use of extensive automation.

The basic concepts are already a staple of many IT organizations, which have been analyzing log data for years to spot performance anomalies, catch intruders and anticipate provisioning needs. But these existing processes are often manual, particularly at resource-challenged midsize businesses. The complexity of tomorrow’s multicloud and hybrid cloud environments won’t allow humans to be a bottleneck.

The AIOps process begins with ingesting historical information to create baseline reference points. It then folds in streaming and real-time data to compare to the baseline and anticipate the need for future changes. Machine learning can detect patterns in large amounts of data that humans would miss, and the quality of its findings improves over time.

Simplifying Operations; Creating Efficiencies

The goal is to automate as much of IT operations as possible. That’s what makes it possible for web-scale giants like Microsoft* to run data centers containing tens of thousands of servers with just a few dozen employees (nyti.ms/2Cytkan). And McKinsey has estimated that more than 70 percent of IT spending can be targeted for automation (mck.co/2QQnPZp). The IT operations of any company planning a digital transformation need to aspire to dramatic new efficiencies.

The goal isn’t to eliminate people but to enable scale. Take the analogy of the stock markets, which migrated from paper trading forms to electronic trading beginning in the 1980s. The switch enabled the New York Stock Exchange to grow daily trading volume from a few million shares 30 years ago to more than 3.5 billion today. Programmatic trading didn’t replace humans but enabled the kind of massive growth that would have been impossible with strictly manual processes. Web-scale companies build their businesses on the same economics.

Transitioning to AIOps

For all of the efficiency gains AIOps promises, the transition won’t be easy for most IT organizations. The challenges will be more cultural than technical. IT operations centers have traditionally been structured around technical disciplines such as systems,

networks, security and the help desk. AIOps unifies these functions around a shared goal of making the right information available to the people who need it as quickly as possible. Data becomes the unifying force, rather than bandwidth or memory.

This transition will require many people to broaden their skill sets in areas like data science and cloud management. But those are the skills that will be in greatest demand in the future. AIOps should be seen not as a way to cut costs but to improve and grow the business.

Delivering the latest technical information to your inbox.