Seven point one (7.1)

Dr. Bashir Mohammed

On July 5, 2019 the U.S. Geological Survey (USGS) reported an earthquake of magnitude 7.1 in southern California. The USGS has about 2,000 sensors used to monitor earthquakes in the U.S. alone. These sensors are connected to a network and they transmit traffic in the form of sensor data in real time.

Imagine a scenario where there is traffic congestion on a network, and these readings do not get to the USGS on time and they are unable to give warnings of an earthquake. Imagine the devastation to human life.

And before you ask, let me tell you now, my research is not primarily about earthquakes. It is focused on developing an AI algorithm to avoid network traffic congestion at ESnet.

Today, we live in an interconnected world, which is heavily dependent on the internet. Of the seven billion people that populate our world, four billion are already online. This means network traffic will continue to grow exponentially day by day.

Imagine waking up in the morning and you cannot connect to the internet with your cell phone, tablet, or PC because the network is congested or down” OMG! A lot of research has been done in the area of network traffic congestion management, but there is a problem. Most research has looked at the issue from a reactive standpoint, meaning it has focused more on addressing the problem after the traffic congestion has occurred. Clearly, this is risky and not a good solution, as evidenced by the earthquake example I referenced earlier.

This is where I come-in. My team and I have addressed this problem from a proactive point of view. Basically, we are utilizing the concept of basic statistics to analyse and explore this massive amount of network data, plus the power of advanced computer simulation and artificial intelligence to predict way in advance, the real-time network behavior of the next level of traffic congestion. In simple terms, we have used historic time-stamped network data to predict the future network pattern.

This, we believe, will provide network engineers with prior information of the network behavior, which will help them avoid data traffic congestion, network downtime, and above all, save DOE millions of dollars on equipment costs. And when we are successful, it will answer what is arguably one of the most challenging open research questions in network engineering research.