Edge Computing on a Multichannel Router


Goodmill is part of a European wide consortium with a project called AINET. The name comes from “Accellerating digital transformation in Europe by Intelligent NETwork automation”. Our part in the project is focused on bringing edge computing to our existing multichannel router products and running partner workloads on it.

While edge computing simply means the ability to run general computing workloads (server applications) within a communications network and thus close to the user application, this isn’t quite as simple for a multichannel router. In addition to the usual topics like security and computing power of the current hardware, to make edge computing on a router really usable, we need to address complex questions regarding the management of the running workloads. If our customer has, for example, a couple of different workloads and thousands of routers spread out the country in ambulances, moving, and handling emergencies, how do we update an application or how do we monitor which application instances are performing as intended?

So, what does edge compute and intelligent network automation have to do with each other?

There are two ways in which our edge computing is related to network automation. The first one is that we believe that our orchestration challenge is big enough that we will need quite smart automation to handle it. And the second one is that many of the workloads we have identified with our customers can be considered network automation tools themselves. Our customers are looking to, for example, run so called MCx applications like Mission Critical Video or Push-To-Talk on the router, cybersecurity solutions like custom encryption or packet inspection solutions, and even network monitoring solutions that may interact heavily with the core multichannel routing. For example, a network monitoring solution might bring more information to the router on which networks should be used in a given situation, if the networks are, for example, under attack or otherwise degraded. Other uses include video data solutions and even small code snippets handling our general-purpose IO interface that are currently implemented more hands-on within the product.

Further complexity in automation stems from the fact that from a cellular network perspective, the router is a user device and not part of the network. We therefore need to find ways in which the management can still be uniform and coherent and how we can, for example, avoid instability from the network and the user device both trying to automatically adjust the same topic.

AINET comprises of three subprojects all of which tap heavily into using artificial intelligence for the automation:

  • ANIARA, which focuses on high performance services on the network edge.
  • PROTECT, which focuses on the automation of network resiliency and security.
  • ANTILLAS, which focuses on autonomous network operations including edge compute (MEC) and Internet of Things.

Goodmill is part of the ANTILLAS subproject but works together with all the relevant members of the consortium across the subprojects.

AINET’s funding was approved late last year, and it is a part of the Celtic-NEXT. Celtic-NEXT being a billion-euro R&D effort looking at next generation telecommunications and consisting of more than a hundred different projects. In AINET there are consortium members from seven countries. In addition to Finland, partners have their home bases in Germany, France, Sweden, Netherlands, Poland, and the United Kingdom. Some notable partners in addition to us innovative startups include both industry behemoths Nokia and Ericsson, as well as national research organizations VTT, RISE, and Fraunhofer.

Our project started on two fronts. We’ve been looking at how to bring more hardware power to the routers and what sort of edge computing needs our customers and consortium partners are able to identify. AINET will run for several years, but we hope to bring innovations into Goodmill products already earlier. For example, it looks like we are able to bring out more processing power and thus increased throughput as fruits of the efforts done in the project.