Major airlines often are forced to cancel flights because of relatively minor computer-related glitches and outages, stranding customers and costing the airlines countless millions of dollars in refunds and vouchers. Such unforeseen failures provide a textbook argument for smarter systems management software, whose machine learning technology and ability to process huge amounts of data can identify and resolve system problems before they adversely affect a company’s revenue and reputation.
Moe Fardoost, a senior director of product marketing for Oracle Management Cloud, says it’s a natural progression for the company to be adding its machine learning and data processing expertise to its systems management offerings, after decades of gleaning insights and advising customers.
“We are a data company: We mine data, we store data, we help our customers analyze their data, we know how to deal with data, and data is—ultimately—intelligence,” Fardoost says.
Machine learning is a perfect partnership between that data, the computer, and the human coder. Programmers write algorithms that allow the computer to learn as it processes data that describe routine operations. The more data the ML software processes, the more it learns and, therefore, the more accurately it identifies which operations are running normally, and which are not. And before an abnormal operation becomes serious enough to affect customers, the system can notify its human operator to take corrective action, or fix the problem automatically.
Machine learning powers the integrated portfolio of monitoring, maintenance, and analytics services offered in Oracle Management Cloud. The product set, an Oracle platform-as-a-service offering, helps enterprises improve the stability of their IT systems and prevent outages.
A Smart Idea
Consider the experience of IDEA Cellular, the third-largest telecom provider in India and early adopter of Oracle Management Cloud. Because the systems management software evaluates many millions of metrics collected from the full range—rather than just a subset—of monitoring tools plus related activity logs, IDEA Cellular now can more effectively pinpoint the causes of problems.
“It has helped us correlate sources of data and logs from multiple components that previously were not looked into,” says IDEA CIO Prakash Paranjape.
Safexpress, an India-based logistics company, is using Oracle Management Cloud to manage its IT infrastructure more efficiently and detect problems before they impact customers, says Sanjay Bakshi, an IT senior manager at the company. “For example, we were able to detect issues like slow transactions and resolve them proactively and faster,” he says.
Predicting What’s Next
Oracle Management Cloud incorporates four major categories of machine learning, says Dan Koloski, Oracle senior director of product management and business development:
- Anomaly detection algorithms tell the operator when observed behavior falls outside normal expectations. “When things are operating as expected,” Koloski says, “we generally don’t have a problem. When improving performance or preventing outages, we want to manage by exception.”
- Clustering algorithms identify the patterns in huge volumes of data to focus the operator’s attention on the small pieces of information that indicate a departure from the usual pattern.
- Correlation algorithms identify metrics that change in tandem with one another—for example, the time it takes for a web page to load often increases with the number of users on that page. When correlation doesn’t follow typical patterns, the algorithms will flag a potential problem.
- Prediction algorithms draw informed conclusions about what might happen in the future based on historical observations and current context. This insight allows companies to intervene to prevent problems instead of just reacting to them.
It’s an exciting time to be applying new technologies to systems management software, both within Oracle and at customer locations, Koloski says. “Traditionally, systems management has been all about looking backward at what happened in the past,” he says. “But now IT professionals have enough confidence in machine learning predictions that instead of just reacting to what happened, they can prevent what might happen.” Just in time for this summer’s holiday air travel.
Linda Currey Post covers science and technology advances as a senior writer at Oracle.