With 18 months of managed cloud services under its belt, ZeroStack says it's ready to start feeding what it's learned into machine learning algorithms, aiming to further automate the private cloud.

The goal is to let application developers run jobs without worrying about things like storage capacity or the optimal placement of applications, says Steve Garrison, the startup's vice president of marketing.

"A lot of people are wanting to refrain from managing anything on-premises any more. This is why public cloud is taking off," Garrison says. What ZeroStack offers is infrastructure that can be consumed as if it were a public cloud, but that remains under control of the enterprise — a "neutral hybrid" alternative for work that the enterprise would prefer to keep closer to home, he says.

Platform9 and Stratoscale similarly offer prefab enterprise clouds that fit that hybrid model.

Because ZeroStack's cloud is a managed service — a shrink-wrapped cloud, in a sense, that can be plopped onto the customer premises — the company has been able to observe how these clouds are used and glean best practices. By now, the company has amassed more than 1 million data points to feed into machine learning algorithms, Garrison estimates.

The first stage of the company's machine learning offerings will include capacity planning and automated decisions about the sizing of virtual machines. These are relatively easy questions to answer, but those answers could help avoid the overprovisioning that's common in clouds and data centers, Garrison says.

Future capabilities will include optimized workload placement and, later, automated troubleshooting, which could include the flagging of potential security problems.