January 26, 2016 | Written by: Matt Rodkey
Share this post:
Cognitive systems can understand the world using senses and interaction, reason using hypotheses and arguments and learn using experts and data.
That definition sounds both incredible and scary in a SkyNet sense (to use a Terminator reference). When I think about cognitive IT I often imagine a personal digital assistant attached to the computers in a network operations center. Her voice sounds a lot like Siri and she says things like, “You should probably take a look at the online banking application, I believe there is a memory leak.”
The reality is that today we are not that far off from that vision in the world of predictive analytics for IT. Today we can use machine learning to understand “normal” operational behavior across application, infrastructure and network. From there we can identify emerging changes before someone even knows to look for them. We can help find and fix problems before they impact service. Here’s another blog post where author Denny O’Brien lays out the vision for cognitive IT.
Introducing a new vision
Let’s contrast traditional IT operations with a more cognitive approach.
Take a simple metric like end-user response time for a web application. In traditional IT operations a set of statically-set thresholds would alert you if the response time took too long. This can work, but with a variable metric like response time it is hard to find the right place to set that threshold to avoid having too many or too few alerts.
On the other hand in the world of cognitive IT, we know that response times are a little longer around 9AM when everyone is arriving at the office and signing on, and much faster after about 6PM. When you receive an alert from a cognitive system it will not only tell you that the response time is abnormal, but that it has also found several other correlated changes related to the same problem. A cognitive system will also point you to error messages in the logs and offer some expert advice about how this type of problem has been solved in the past.
All of the above is possible today, and perhaps in the future that IT personal digital assistant will be possible as well.
You can take a look at what we at IBM are doing with cognitive IT by clicking here.