From the outside looking in, the promotion of 5G and Mobile Edge Compute appears to be a re-run of an old ‘IoT is Serious’ episode, with the cast of door sensors, street lighting and automated dustbins replaced by headline VR, AR, Gaming and autonomous vehicles actors, sponsored by the public could providers terrified the edge will erode their margin levels.
The motivation for this article was one such marketing blast where the originator proclaimed ‘5G is the Edge’ at which point I chuckled and pondered a response.
What do I think is the Edge?
As I’ve already mentioned IoT it is important to recognise that not all Edge is IoT and not all IoT is Edge. A sensor waking up once a week, producing and dumping data to the cloud is not Edge.
My version of edge is a computational engine capable of local actions, operating within a 20ms latency window over fixed or wireless infrastructure to a range of connected devices.
This version comprises dedicated User Edge, controlled and operated by the user and a Service Provider Edge, shared and operated as a service. The two overlap slightly but can be divided by ‘last mile networks’ with User Edge downstream. In effect the LF Edge model
In both cases the GPU or TPU enhanced computational engines supporting micro-service based AI and ML applications and comms like LoraWAN, Wifi and 5G, deliver true edge platforms, operating in isolation if needed but more often in partnership with cloud based services.
Whilst it is easy to consider spectacular latency critical AI services like Virtual Reality as prime Edge drivers more mundane concerns have pushed early adoption. Asset Tracking for example, generates huge message flows, which when sent to the cloud equates to considerable bandwidth, ingestion, buffering and storage costs. Likewise image recognition or people counting streamed across the internet has significant GDPR/compliance issues and in volume, considerable latency concerns. Both actions also assume operators understand the complexities of deploying a cloud backend (not always the case). Too many times cloud based reporting services fall short, delivering substandard customer experience.
So the whole ingesting, processing and acting locally makes commercial and regulatory sense, with anonymised data reports sent for cloud based analytics and storage. Through transactional mapping at the edge, device blueprints dramatically reduce the security concerns corralling breaches locally without impacting the entire estate.
And yes 5G has a massive part to play in the mobile element of edge. The ability through slicing to identify, route locally and process time critical service like Gaming makes perfect sense, but so does placing edge software on a LoraWAN gateways for augmented building management or on PS-LTE servers for campus wide services straight from the get-go.
In conclusion, Edge computing is a key technology in our digital transformation strategy and whilst 5G adds significant power to the model it is a player, nothing more.