With $3.6M in fresh funding, YotaScale optimizes cloud computing for enterprises

YotaScale, a graduate of Alchemist’s enterprise accelerator, is announcing a $3.6 million venture round today from Engineering Capital, Pelion Ventures and angels Jocelyn Goldfein, Timothy Chou and Robert Dykes. The startup employs machine learning to help balance performance, availability and cost for enterprise cloud computing. Competitors CloudHealth Technologies and Cloudability have raised a combined $80 million in the hot space.

Cloud computing has rapidly become integral to businesses in just about every industry. But the quick pace of innovation has made it hard to monitor ever-evolving cloud infrastructure. Rather than dump the responsibility on humans, YotaScale is automating performance management itself.

The company combs over a myriad cloud data to ensure that a company’s infrastructure is optimized for its overarching business priorities. These priorities can be really simple, like minimizing cost, or they can be highly complex, involving multiple projects with different end-goals.

“Anybody can do the simple stuff and tell you your machine is running low on utilization and you should shut it down,” explains Asim Razzaq, CEO of YotaScale.

Razzaq’s system is able to combine usage data with billing and log data. This information serves as the underpinnings for anomaly detection against a baseline. Though it might not sound like a lot of data, it’s enough to extrapolate out things like resource consumption and CPU utilization.

But the tricky part of anomaly detection is defining normal, because normalcy is highly contextual. A spike in use might not be an anomaly at all for an e-commerce company on Black Friday. To this point, YotaScale isn’t just concerned with historical data, it actually makes forward projections. This makes it possible to contextualize data fluctuations. Instead of flagging every single change, the system compares expected performance against actual performance.

Different types of cloud infrastructure data are created in different time intervals; some hourly, others daily, etc. The challenge becomes optimizing across that differentiation. Ensemble machine learning techniques are used to improve the accuracy of analysis and to manage the many dimensions of captured data. Regression models serve as the foundation, with other semi-supervised models coming in for specific uses.

Using YotaScale, enterprises like Apigee and Zenefits can ideally rely on machines to manage their cloud computing needs, taking a load off cloud and DevOps teams. Not to mention, machines have a pretty strong compute advantage when it comes to real-time analysis.