Using Learning Algorithms to Monitor IT

I spoke with Daniel Heimlich, vice president at Netuitive, today. Heimlich told me that there's so much data to track in a modern environment that it just isn't possible for people to do it all.

"It just defies human analysis," Heimlich said. "It makes no sense."

For example, you might have automatic load balancers that shift your VMs around when their hosts get too busy. But before it can do that, it has to know what "too busy" means, exactly, and that means you have to synthesize the data from all your different systems.

Netuitive 5 can automatically learn the behavior or your systems—it might figure out that your systems get a lot busier Monday morning at 8 and calm down during the weekend, for example. Other vendors are including this kind of smarts in their products, such as System Center Operations Manager's self-tuning thresholds. Between storage, processing, networks, hypervisors, and everything else in IT, it seems the trend toward products that do a little thinking for themselves is likely to continue.