The network edge is a huge component of 5G. That’s why big service providers like Verizon and AT&T are suddenly talking about how they are adding new capabilities to the edge of their networks. But adding all those capabilities also creates complexity because it often involves thousands of edge sites in every major market.
That’s where Edgility comes in. This new initiative was developed by AT&T and Cloudify to help operators manage the computing resources across the network.
According to Eden Rozin, product manager in AT&T’s Research and Development offices in Tel Aviv, Israel, Edgility isn’t an open source project but more of a “use case” that can be deployed along with Akraino, the Linux Foundation’s open source edge project, or Open Network Automation Platform (ONAP), a Linux Foundation project developed to help carriers to automate, design, orchestrate, and manage services and virtual functions. “When you have this type of [edge] deployment, you need to know how to manage it. How do you manage the resources of the infrastructure that you have?” Rozin said, adding that orchestrating all these elements at the network edge is much different than orchestrating them in the cloud.
For example, the edge architecture is different because it consists of many different types of nodes from different vendors such as Amazon’s Greengrass, or Microsoft’s Azure IoT Edge, and each vendor has its own metrics and APIs. But with Edgility, an operator can introduce a serverless functionality that will help manage the network resources so that only applications that need certain network resources like power and capacity are using them. If an application doesn’t use certain resources, they are reallocated to applications that need them. “This is a major change,” Rozin noted.
Cloudify CTO Nati Shalom described Edgility as a “bridge” that is needed between the edge and the rest of the network that makes sure that operators are able to execute a workload anywhere on the edge.
Rozin described one use case where Edgility would be particularly useful. For example, if a network operator is using a content delivery network for video processing, Edgility would be able to predict behavior (video viewing increases after 8 p.m.) and then ensure that smaller code is used at the network edge during the busy time so that not all the video traffic is going back to the central cloud and customers have a better user experience without a lot of delays.
Edgility is still in development and looking for additional members to help with the initiative. Rozin said that the group is now starting to work on the second phase, which will incorporate machine learning. He also said that Edgility would like to be able to interface with Acumos, the Linux Foundation’s open source project that provides a common framework for building and managing of artificial intelligence and machine learning tools. “We would like to interface with Acumos and provide a predictive orchestration and resource utilization.”