Can your pet predict an earthquake?
For generations, people claim the behavior of their dog or cat changes noticeably before an earthquake. Scientists have long sought to prove the theory by linking changes in the atmosphere, or in the electro-magnetic spectrum, to a change in an animal’s normal behavior. The hope is to build a prediction tool.
At AT&T’s (GNOC) Global Network Operations Center, in Bedminster, New Jersey, we also look for changes in the normal usage patterns in our network which presage something abnormal — like an earthquake! Our goal is to predict and then pre-empt problems before they impact our customers.
Take, for example, these graphs which track phone calls on mobile phones. On the left, we count the number of cell phone calls. On the right; we track text messages. The yellow lines represent the usage of one week ago. The green and blue is the usage of the current day.
See how the voice usage (left) peaks in the late afternoon as people leave work and call home to check on the kids. Note how the messaging usage (right) is consistently high for most of the day, but drops quickly when those kids (!) are going to bed. Indeed, there is a rhythm to the network. It mimics the ebb and flow of human behavior.
Any deviation from the normal will impact the otherwise predictable inflow of data we see in the network.
So, can the GNOC predict an earthquake?
Below are the cell phone call and text messaging graphs on the day of the unusual earthquake experienced in the northeast on August 23, 2011. The dramatic spike in voice calling and text messaging are obvious. But, notice the drop in traffic just prior to the spike.
From pro-active to pre-emptive
From the network’s point-of-view, there is a delayed reaction. The big increase in calling and texting is preceded by a period of lower than normal usage. That serves as a forewarning to the network managers here in the GNOC. They can use the time to initiate a series of controls which will make additional capacity available and create a buffer that will absorb the inevitable stress of the surge in calling and texting which follows.
Network managers can predict outcomes based on subtlest of variations from the otherwise predictable stream of data that flows before us here at the GNOC. Practically speaking, we have been able to move beyond being reactive or even proactive to being predictive and pre-emptive.
And we can do the same for you. We can monitor the operation of your network and get in sync with the rhythms of your business. We’ll build predictive models based on these rhythms and be ready to predict and preempt service (or security) issues before they affect your enterprise.