Great Connections Happen Here.™

Machine Learnings Pair with Optical Transport Networks

One of the big developments that occurred during the 1980s and 1990s was the replacement of traditional phone cabling with fiber optics.
maximalfocus-VT4rx775FT4-unsplash

One of the big developments that occurred during the 1980s and 1990s was the replacement of traditional phone cabling with fiber optics. That major change in infrastructure by the phone companies put in place the building blocks for the Internet and all the data communications we enjoy today for massive data movement.

The Next Network Evolution

Today, machine learning is clearly showing the potential to move optical telecommunication to an exponentially higher level. With the aggressive move toward interconnected everything continuing, the demand for Optical Transport Networks or OTNs become practically essential for high capacity telecommunications data movement while still keeping the overall infrastructure cost down. As a result, researchers have been busy pushing the capability of fiber optics even farther. The benefits are quick and immediate; activities that need high amounts of data transfer quickly are becoming more efficient with the same improvements, the most visible example being self-driving cars and homes that manage themselves or intensive video/VOIP networks.

The other feature advantage of an OTN is its self-regulation processes. High data demand causes significant fluctuation from one side of the network to the other very quick, depending which user is running a given program. With an OTN, the network is automatically allocating resources to the demand where it is needed, drawing away from where the resources are not being actively used. The result is a far more responsive network instead of users having to be prioritized and less ranking players getting the “still buffering” message instead of immediate service.

The Challenge

However, OTNs are also very vulnerable. Lots of data means lots of exposure and vulnerability to problem data, especially when its buried in seemingly innocuous traffic. However, the latest research is solving this gap with machine learning. First, there is the classic reinforcement training that teaches the machine what is correct and what is not allowable. Then, the machine moves to the next level of application, learning and seeing by example how to apply rules of protection. And instead of the standard case reference so common in traditional coding (and also easily circumvented by new risks), machine learning utilizes a neural approach to learning, adapting and seeing similarities of risk versus strict definitions in code snippets. In short, the machine is learning how to protect like a human building experience.

Solution Finders In Play

One of the lead researchers from Universitat Politècnica de Catalunya, Albert Cabellos-Aparicio, sums up the combination of machine learning with OTNs as a new development, but with a tremendous benefit potential. Their research is intended to push support for further development to the forefront instead of continuing to rely on data networks as they are under current traditional designs. The first practical level of application has been in resource allocation within an OTN. However, the logic of the computer is still trying to get past the exposure to new, unplanned condition changes, i.e. novel events. Research realized that the issue may have very well been insufficient reference exposure to conclude a new event was a problem. So, they ramped up the simulation exposure and the machine learning improved significantly. The same way athletes practice 10,000 touches of a sports ball, the machine had to do the same to realize its learning breakthrough. The results started outperforming classic algorithms in performance under the same conditions and test event.

The above might seem minor in the grand scheme of things currently, but each application of machine learning in the OTN field has dramatic improvement potential in reducing operational costs from coding to contingencies as well as reducing lag and data delays. The learning machines once they have their rhythm perform and move data management far faster. Albert’s team is already looking to expand into the visual side of data performance, but they unique push they’ve already achieved is a disruptor to how networks should behave with data. And what’s going to come out of these changes in just a few years’ time is going to be as dramatic as man launching to the moon in the late 1960s.

Partnering With The Future Now

PS LIGHTWAVE provides high-speed, fiber Internet for public and private commercial entities in the Greater Houston and surrounding areas.

Through our high-quality infrastructure, innovative technology and expert, locally based support, we deliver not only the best in connectivity and reliability but in scalability and redundancy. We invite you to learn more about our services, our history and our dedicated team.

Share This Blog Post:

Facebook
Twitter
LinkedIn
Pinterest
PS LIGHTWAVE Blog

PS LIGHTWAVE Blog

PS LIGHTWAVE, a leading telecommunications service provider headquartered in Houston, Texas, provides managed Ethernet Data Circuits, Internet, private network solutions and Voice over IP (VoIP) over one of the nation’s largest facilities-based private Metropolitan Area Networks (MANs). The switched Layer 2 network, backed by 24/7/365 Network Operations Center (NOC) support, encompasses approximately 5,500 route miles and 1,400 on-net locations and connects 100+ fault-tolerant multi-gigabit Ethernet rings for built-in redundancy, security, low latency, and high-availability. At PS LIGHTWAVE Great Connections Happen Here™.

For more information, please visit https://www.pslightwave.com or call 832-615-8000.