There’s much talk in industry circles about the potential of machine learning to automate and improve IT infrastructure, but the discussion can seem distant to the majority of administrators working for traditional organizations that are still grappling with the everyday challenges of virtualization. That’s the divide that SIOS Technology Corp. has set out to bridge with the latest addition to its portfolio, which was the topic of the most recent episode of CUBEconversations.

Setting the tone for his first appearance in the SiliconANGLE studios, Jerry Melnick, the COO of the San Mateo-based software maker, told host Dave Vellante that the transition from the old way of running applications directly on the physical layer has proven difficult for many of its customers. Relatively late adopters, their staff is only now coming up against the challenges to seizing the full benefits of virtualization.

And they’ve “ran into a wall,” Melnick said. “This is not the deterministic environment of a physical server, there’s lots of things going on, lots of complex interactions, and we saw the opportunity to help them by applying intelligence to how they’re operating.”

SIOS iQ, as the company’s new solution is called, can automatically infer those relationships from the traffic flowing through the network on which it’s installed without so much as requiring administrators to set up local tracking agents. As a result, Melnick said, the software can get fully up and running within about five minutes of the initial download.

Its data aggregation component records activity from everything between the hypervisor – in this case VMware Inc.’s ESXI – up to the applications running above and passes on the information to a patented correlation engine called vGraph. According to Melnick, the technology pieces together the individual components of the environment into an integrated view of operations, complete with insights into the dependencies and interactions that are bogging SIOS customers down.

That information is then used to identify inefficiencies such as underutilized virtual machines and redundant copies. Neither requires any special skills to uncover, but can prove immensely time-consuming and difficult to find manually in large environments with upwards of dozens of different workloads with changing resource consumption patterns.

Meanwhile in the background, the built-in machine learning algorithms incorporate all of that activity into a model that serves as a reference for the normal behavior of the environment, helping to isolate more subtle operational issues such as long-term performance hiccups. “We’re continuously learning, observing patterns and looking for anomalies,” Melnick explained. “We then report on those and helping identify why there’s a problem and what you need to do to fix it.”

Those insights are made available through a homegrown dashboard that can be accessed through mobile devices and avoids most of the manual tinkering needed to enable such monitoring, the executive said, notably setting fixed operational thresholds. “You can look at that with one click and we will show you the objects that are impacted by the anomaly, the root cause of what happened and why. So you can see what happened there.”

On launch, SIOS iQ is geared primarily towards the operations professionals interacting directly with the infrastructure. But Melnick sees the insights gleaned through the platform potentially also finding use with CIOs later down the road in strategic tasks such as determining how much additional hardware needs to be acquired in order to accommodate data growth over a certain period.

That reflects the fact that SIOS iQ is designed to run solely  on-premise, which he attributed to customer concerns over storing operational data from their environments to the cloud. However, Melnick hinted that other deployment options may be added with time as its user base continues to embarce the new realities of IT operations.