It would seem the protracted deliberation around cloud verses edge is no longer such a conundrum.
In the data-hungry and ever-expanding scope of the IoT, where projects underpinned by wide spread and convoluted sensor networks are pre-requisite, there is no contest; the edge, quite simply, has the edge for faster, more secure and cost-efficient data capture and analysis.
The myriad smart city projects that are now starting to come to fruition are a case in point, having long outgrown the cloud and the bandwidth requirements of existing infrastructure for their data processing requirements. Here, the sheer sum of parts and complexity of the eco-system that needs to connect, demands nothing less than the agility and steady flow of device data on the network edge to manage assets, spanning requirements as diverse as smart traffic systems and intelligent meters to real-time air quality reports.
Here, penetrating the most remote part of the network to make data more accessible to the user - where the activity of the smart device is happening - means easier data capture in real-time and provides the foundation for immediate operational data decisions. In fact, it results in an endless list of positives. Positives that include: slashing the time data spends travelling across the network bandwidth and thereby reducing the network load, to enabling localised data processing and making connected apps more resilient and responsive.
But the potential does not stop there; in an ever-evolving digital transformation agenda where every technological advancement is ripe for more improvement, we are seeing how the capabilities of edge computing are further enhanced. This includes the addition of real-time analytics and machine learning capabilities that can run and perform on even low powered devices.
The two are in fact natural bedfellows, with machine learning benefitting from the ‘freshness’ of real-time edge data to drive not only more frequent observations, but more accurate predictions for nuanced and sophisticated interventions.
It’s a mix that is set to play a central role in securing the more mainstream traction of autonomous cars, rightly cited as a central facet of the smart city revolution. In essence, a machine learning-infused edge is best-placed to address perhaps the central technical challenge presented in the context of the driverless car. That is, handling the anticipated terabytes of data churned out, without incurring the latency that can compromise the fast responses demanded in constantly changing driving conditions, to include the weather, roadworks or other driver behaviour.
Here, there is simply no time to send information back to the cloud for analysis and the reasons go beyond convenience and efficiencies. When even a fraction of delay can mean the difference between having a serious collision or avoiding that accident altogether, being able to detect patterns in sensor data to aid real-time decisions at local nodes, is essential.
As these real-life applications amply demonstrate, we are now on the cusp of an era in which the convergence of machine learning and the network edge will have a transformative effect across all facets of the digital enterprise. What will be the real game-changer is making these capabilities available via solutions which take the inherent complexity out of algorithmic modelling and make it a far more accessible and intuitive proposition for developers of varying abilities.
Perhaps one of the most prevailing messages that underpins success in the digital enterprise, is the criticality of opening up digital opportunities, enabling a hands-on collaboration as widely as possible across an organisation. This not only drives efficiencies, such as improved scalability and best practice, but generates more ideas that build on the existing progress and take it to the next level.
It is why collaborative platforms need to be at the heart of this innovation. Ones which, via code-free workflows, automated models and shared interfaces, enable citizen developers, data engineers and business experts to create, without the deep programming knowledge that has traditionally been a prerequisite.