IBM applies predictive analysis to IT operations
Turning its considerable expertise in analytics to the study of IT operations, IBM has assembled a software package to help system administrators better pinpoint potential problems and performance issues, using many of IBM's tools for business intelligence, machine learning and data mining.
"We've applied the IBM investments in analytics to IT," said Matt Ellis, vice president of software development for performance management.
The new product, called Predictive Insights, was designed to predict and detect anomalies in the IT environment.
Since 2005, IBM has invested US$16 billion in boosting its analytics capabilities, both through acquisition and research and development. Now, the company is applying some of this expertise to helping administrators better manage IT operations.
Predictive Insights can tap into operational data from a number of areas. It can use log files, changes to configuration files, network traffic, performance metrics from application APIs (application programming interfaces), and alerts and alarms from the applications. It can also work with data from various monitoring systems, such as application performance management (APM) and event management software.
The software uses the IBM InfoSphere Streams platform to intercept traffic from these sources and establish a baseline for typical system performance. The software then can recognize and highlight emerging patterns across these different data sources, which is important for pinpointing problems across a stack of software.
For instance, a banking application may be performing slowly, but it would be difficult for a network administrator to determine the problem may be with the Java Virtual Machines (JVMs), which are executing garbage collection duties at inopportune times, slowing their output.
Predictive Insights was designed to catch such problems, Ellis said. The software also draws from IBM's Cognos business intelligence software and the company's SPSS algorithms for predictive analysis and machine learning. It is designed to predict when problems will occur, given past behavior.
The software also offers administrators the ability to search across data sources. In addition, it also is able to produce documentation and information for trouble tickets.
IBM is not alone in its pursuit of the emerging field of applying analysis to IT operations data. Hewlett-Packard is also planning to launch its Operations Analytics platform by the end of this year.