Taking it to the Edge: the new ‘super platform’

Oct 31, 2015

As billions of connected devices start joining the Internet of Things (IoT), businesses will experience exciting transformations in some areas and unwelcome disruption in others. Not least, the network environment itself will transform into an astonishingly complex hybrid of distributed intelligence, virtualised functions and multiplicity of access technologies.

This is broadly the 5G vision. But the disruption lies in the threat to the traditional gatekeeper role of the network operator. Ownership of spectrum licenses and hyper-scale data centres is only part of the story in this dynamic and decentralised world, where resources are allocated and shared ad hoc as applications request them.

Mobile Edge Computing (MEC) has been described as ‘the new super platform’. Compute and storage embedded in network assets such as base stations and routers enables granular control over service delivery, caching and other performance factors. As a complement to Network Functions Virtualisation (NFV) and Software Defined Networking (SDN), MEC gives telcos unprecedented ability to optimise the end-customer experience on behalf of demanding clients such as OTT content providers.

But the question is then, who will profit from these flexible resources? When context-aware applications grab bandwidth, compute power and storage they need on the fly, the providers could be telcos or other network owners, consumers selling spare smartphone capacity, or IoT devices with on-board intelligence. In effect, the Operator as Gatekeeper is deposed by a rather nebulous ‘ecosystem’.

This matters because the success of the IoT depends on the Edge Computing paradigm, particularly for its most performance-critical applications.

Moving nodes such as autonomous cars and drones, for example, are real-time safety hazards. They need to recognise their neighbours and work out where they are going, as well as negotiate around pedestrians and other unpredictable obstacles. Computer vision calculations and sub-second decision-making rely on exceptionally low latency. So compute and storage must be located very close to the vehicle, ideally on board.

The cost of wireless connectivity is also driving processing power out to the edge. Qualcomm’s new reference design for a ‘conscious’ IP surveillance camera includes a six-core, 64-bit ARM Cortex processor to run some intense embedded analytics. The Snapdragon 618 is able to pick out footage most likely to be valuable – using facial recognition of a known perpetrator, for example. Unwanted video is discarded and the customer minimises their wireless bill, just updating the search algorithm as needed over the air.

Additionally, algorithms can help customers comply with privacy regulations by ensuring any non-compliant video is discarded before it ever leaves the camera. This reassurance is worth a lot to companies who need surveillance but don’t need the legal headache. It could even be a good trade off against the costs of the high-end camera and the cyber- and physical security required to protect it. Hmm, perhaps another camera to monitor the camera….  

Dell is also talking up localised analytics. Its Edge Gateway 5000 with Statistica middleware is built for context-aware decision making. Building energy management provider Elm Energy uses it to select between renewable energy sources based on real-time price and supply. End customers are advised when to store excess generation from volatile sources like solar or wind, and when to feed into the grid for direct payment.

While Edge Computing is coming into its own (or Fog Computing to use Cisco’s term), in reality the norm will be a hybrid of local and centralised resources. Voice recognition analytics for example, requires less stringent response time and runs just fine from the cloud. And while some devices will pack in the processing power, others will stay dumb – perhaps reaching out to smarter neighbours for intelligent decisions.

Some pretty sophisticated orchestration is required for all this, as ultimately even a single application will split its requirements between edge and cloud, based on payloads at a particular time.

The implications for the ICT industry are profound. Intelligence at the edge suggests commercial power at the edge. As ever, the challenge will be to align the infrastructure investments needed with the financial rewards that flow from them.

Fatal error: Call to a member function getUserUserBio() on null in /home/frost/public_html/digitaltransformation/packages/problog/helpers/blogify.php on line 362