MSP Today Expert Feature
March 07, 2013

CA to Boost Performance with Machine Learning


Computer Associates hopes to get an edge in the hotly competitive world of performance management with the addition of machine learning to its CA (News - Alert) Application Performance Management (APM) tool.

The addition is thanks to an OEM deal with Prelert, a performance management vendor with some rather unusual technology. Prelert’s “machine learning predictive analytics” is, in a sense, two steps above expert systems, and is considered an artificial intelligence machine learning system.

The software pores over the data gathered by existing monitoring tools looking for patterns and anomalies that impact performance. This data includes usage reports, transactions, SNMP polls, time series performance information, SNMP traps, etc.

Based on this analysis, the tool “learns” about your systems, ultimately focusing more and more on the data that actually points to the causes of performance problems. This is the notion of machine learning. In the case of Prelert, the tool can save 90 percent of the costs of finding and fixing problems, the company argues.

“Prelert self-learns the behaviors and inter-relationships of tens of thousands of metrics and log notifcations reducing deployment and tuning efforts by 90 percent. Prelert automatically identifies anomalous activities and problems as they develop, often predicatively, and before they can impact users,” the company explains.

According to an analysis by Enterprise Management Associates (EMA (News - Alert)), Inc., the algorithms behind Prelert came out of the world-changing human genome project. 

The software has proven its worth, EMA argues. One customer used Prelert to find out that one of its servers was dropping its connection to a key client. Instead of a million dollar network upgrade, the problem, once identified, was simply fixed. 

This kind of software is not only useful to large enterprises, but Managed Service Providers (MSP) as well. Huge MSP networks are hard to analyze given their girth and complexity. Problems not only slow client performance, but can force the MSP to shell out money when service level agreements (SLA) are not made. Also, end-user experience is a key factor in whether a customer sticks with a provider, or goes shopping for a new one. 

Prelert believes the complexity of today’s IT environments make proper performance analysis impossible to conduct by IT staffs alone – there is simply too much data. “Prelert provides machine learning predictive analytics software that complements and extends IT capabilities by automatically mining huge amounts of data to identify developing issues and their linkage to the complex application environment,” the company said in its announcement. 

As part of CA APM (News - Alert), Prelert will cross correlate information from CEM, SOI, NetQos, and Introscope. "Our customers manage some of the world's most complex application delivery environments, and CA Technologies (News - Alert) is committed to continuously provide them with new capabilities that make their jobs easier," said Rick Fitz, senior vice president, product management, CA Technologies. "We look forward to working with Prelert to integrate its robust and leading technology as an integral part of our Enterprise Management offerings."




Edited by Brooke Neuman




Comments powered by Disqus