Intelligence at the Edge: The Potential of Machine Learning to Optimize Content Distribution
Telecoms.com periodically invites expert commentators to share their insights into the most pressing industry issues. In this piece Anuj Bhatnagar, Senior Director of Product Management at network AI specialists B.Yond, explains why we need intelligence at the network edge.
We live in a connected world that is constantly streaming—video, games, music—the demand for content is always on. And, with emerging technologies presenting massive potential, including virtual reality, augmented reality, autonomous transportation, and mobile infotainment, there will be an unprecedented level of demand on the networks. Internet traffic on content delivery networks (CDNs) will more than double from 73 exabytes to 166 exabytes (Cisco VNI 2017) in the next three years.
In addition to the data demand, these applications will require lower latency, higher reliability and better fidelity than current networks deliver. This will require a significant change in network infrastructure from a centralized to a massively distributed architecture. To truly manage the volume and demand on the new network infrastructure and provide an optimized consumer experience with content at the edge, we need software-driven solutions that are focused on significantly reducing operational cost. We need intelligence at the edge.
Bringing Content to the Edge
To meet increasing content volumes, networks must effectively and intelligently manage massive and divergent amounts of data, be available anywhere with the capability to respond instantaneously, and have extensive security capabilities to support privacy concerns. However, the current industry-standard, based on centralized network infrastructure, cannot meet these requirements.
A network with a massively distributed architecture leveraging cloud, Network Function Virtualization (NFV), and Software-Defined Networking (SDN) technologies to employ edge computing, improves operational efficiency, reduces CAPEX, and creates opportunities for new revenue streams. By significantly reducing the distance between the mobile user and content, edge computing enhances network security, improves scalability and responsiveness, and supports low-latency applications.
With enhanced opportunity for content delivery through edge computing, there are further opportunities for growth and revenue. In CDNs, we are seeing a trend as companies build their own private servers on the edge and move away from distributing content through a shared CDN provider. For content providers, this shift to privatization is lowering the cost of handling increasingly high-definition videos, improving the user experience, and enhancing security.
There is a prime opportunity for operators and cable provides to capitalize on this by creating private or shared CDN servers. This can be achieved by repurposing central offices and adding nodes to cell sites and virtual Customer Premise Equipment (vCPE). Operators enable new revenue streams by building private CDNs using their wireless and wireline networks. With 5G and network slicing the costs can be further reduced.
An Intelligent Approach to Managing and Optimizing Content Delivery
As content is pushed to the edge, the automated, intelligent management and optimization of the network becomes essential. By applying Machine Learning (ML) and Artificial Intelligence (AI) to a distributed infrastructure, operators can proactively identify network traffic patterns and proactively respond appropriately to communications traffic demand with focus on improved customer experience. This process works by operators gathering real-time performance data from the software-defined core and access networks, then using ML and AI algorithms to provide guidance instantaneously. By applying this to video applications, service providers can optimize the end-to-end Quality-of-Experience (QoE) to reduce start-up delay, eliminate freezes and improve video quality.
Imagine a customer who is watching the latest episode of “Stranger Things” deployed through the closest local server in “central office one”. However, as traffic on the network begins to increase, the ML platform would proactively identify the potential impact to content delivery and automatically respond. In this case, by making a copy of “Stranger Things” in another central office. It may not be as close physically, but with more availability for transport. For customers, it means never again having their Netflix binging disrupted.
Because of the scale and reach of their networks and their ability to access full end-to-end infrastructure data, operators have an advantage over content providers distributing over the top today. To leverage the opportunity, operators need to build a virtualized CDN infrastructure with a next-generation ML- and AI-based management solution. Though necessary to effectively and dynamically manage an increasingly complex network, an intelligent management solution will also deliver enhanced quality of experience, and new revenue streams.
Intelligence is Necessary for Progress
With the explosion of content, there is no question that a move to the edge is required to support a new wave of increasingly demanding content-based applications. But, the move to a distributed infrastructure is not enough. Without the use of proactive intelligence—the complexity of a massive edge network and the demands of the content become unmanageable and turn into an operational nightmare.
The optimal customer QoE requires the application of ML and AI to network performance data in order to guide the CDN infrastructure and video applications. Operators and content providers must work together to bring intelligence to the edge to progress the capabilities of content delivery.
Anuj is responsible for bringing B.Yond’s service products portfolio and suite of advance use cases to service providers and Global 500 enterprises. He leads the strategic development of services products to deliver low-touch network management through automation and intelligence. Anuj received a MBA from Rutgers University in Management and Global Business; a MS from Drexel University in Computer Engineering; and a BE in Electrical Engineering from Stevens Institute of Technology.