Get Smart (Or Get Really Confused)

The need for self-learning performance tools is being driven by a number of factors, the growth in the adoption of ITIL (IT Infrastructure Library) and Business Service Management (BSM), and the rise of virtualization prominent among them.

In these environments, where organizations must understand the dependencies between systems and how inter-related performance degradations impact the user, traditional rules-based and threshold-based systems management tools can’t be effective, argues one vendor.

“The core challenge in systems management is this: IT folks are up against key performance indicators (KPIs) for each system or server. How do I baseline each, what thresholds should I set, and how do I do that for an environment that is in constant change, evolving in terms of the number of users, the types of applications, even seasonal types of changes?” says Daniel Heimlich, vice president of IT performance management software vendor Netuitive.

“It all comes down to the core problem, that it’s humanly impossible to understand the health and availability of all systems in real time when you are talking about thousands and tens of thousands of KPIs. For an automated environment, you need automated tools that can learn those environments," Heimlich says. "It’s great to be talking about IT process and business alignment, but if there’s no accurate and timely data underneath that, you can’t support those business objectives in a meaningful way.”

Like a doctor at a marathon

Service Analyzer 2.0, the latest version of the company’s self-learning performance management system that ships this week, aims to address these issues. The software creates behavior profiles against relevant KPIs and presents the data points in context with one another, alerting IT when things begin to deviate from the norm. (And understanding that the norm can be different depending on the time and day of the week — for instance, it might be very common for a business to see consumers abandon online shopping carts 50% of the time on a Monday morning, but highly unusual if that happens on a Saturday afternoon.)

The system is designed to catch the real problems rather than raise false alarms — to use an analogy, it would be like a doctor at a marathon raising a red flag because one runner’s body temperature has increased while his heart rate has decreased, which together might indicate a problem, yet not calling a halt to the race simply because another runner’s heart rate spiked very briefly.

“What our software does is learn the performance dependencies between all the elements of a mission-critical application or server that you tell it are part of the service,” says Heimlich, in comparison with other tools that require IT staff to create scripts and rules against what they understand to be the baselines of each performance indicator. Service Analyzer can import this information in from a configuration management database (CMDB).

“Any time a new application set is added or acquired, the characteristics of the environment change, and our software automatically adapts," he says. "It’s impossible to do that with rules-based tools.”

Next page: Another side of virtualization