Ancestry spent two years executing its shift to the cloud. In that time, the genealogy company migrated a database of over 20 million members away from data centers and into Amazon Web Services.

After the move the company turned attention to optimizing its presence in the cloud, manually adjusting workload settings to improve efficiency and reduce costs.  

"Due to the fact that it's a manual process, we iterated very slowly," said Darek Gajewski, principal infrastructure analyst at Ancestry, in an interview with CIO Dive. "It takes time for us to do performance testing and then being able to get it out into our production environment safely so that we're not affecting our customers at the end of the day."

But there was still room for improvement. An initial proof of concept from Opsani — an AIOps company that relies on machine learning and artificial intelligence — pointed the way to areas to optimize cloud use without hampering user experience.

"We let the system run its course, give us feedback, and then we implemented the recommendations that Opsani's tool gives us without having to spend a copious amount of time trying to come up with the most efficient path for that application," said Gajewski.

Driven by the threat of cloud cost overrun, cloud-based companies such as Ancestry are turning to AI to increase infrastructure efficiency and reduce the length of cloud receipts. 

How machines can help

Machine learning platforms have an advantage for optimizing cloud performance: they can continuously reassess what works and what doesn't. 

Opsani's technology evaluates the resources and settings of an application to identify ways to improve it.

"We use a technique called reinforcement learning," said Ross Schibler​, CEO at Opsani, in an interview with CIO Dive. "We basically have a neural network which learns from making small adjustments on what the best settings should be."

The platform automatically makes changes to a small subset of Ancestry's network, measures that group's performance against the rest of the network, and suggests the optimized settings to Ancestry's leadership, which can then manually deploy the changes to the system as a whole.

"We put a human in front of that one, because those are larger scale systems," said Gajeski. "And it's not that the AI couldn't do it. We just have too many other interaction points with other systems that we need to be watchful for." 

Trusting AI paid off for the genomics company. After deploying Opsani, Gajewski said, Ancestry saw a 50-100% increase in resource utilization and up to a 50% decrease in cost from its initial migration to AWS deployment to today.

AI's upper hand

At the end of last year, in its Market Guide for AIOps Platforms, Gartner estimated the size of the AIOps platform market is between $300 million and $500 million per year. 

AIOps platforms can address infrastructure and operations leaders' "need for operations support by combining big data and machine learning functionality to analyze the ever-increasing volume, variety and velocity of data generated by IT in response to digital transformation," according to Gartner.

Optimizing cloud costs is just one of the outcomes these platforms can achieve.

"In infrastructure, the emerging field of AIOps uses machine learning to predict and prevent system outages, while also monitoring power and cooling to reduce operating costs," said Murli Thirumale, CEO and co-founder at container storage company Portworx, in an email to CIO Dive.

But AIOps can also play a role in software development, where the technology helps programmers automate code development to improve speed and efficiency and reduce human errors.

"The sky is really the limit when it comes to integrating AI with other enterprise IT operations," Thirumale said.