The Age of Small Data
The 2010’s – the Big Data decade – spawned the term of data gravity: the concept that the greater the volume of data, the more that processing will be pulled towards it. Major analytics platforms like Splunk, Hadoop and TensorFlow needed to own the data; and so, migration of monolithic data became a precursor to running analytics.
Today, we are in the Small Data age; processing is moving closer and closer to the source of data, thereby minimizing the transport costs and maintaining privacy and data sovereignty; and analytics are being run against smaller and smaller data sets; with only conclusions and exceptions being sent to the cloud for central processing. It’s as if the polarity of the data has been reversed, and the gravity is now at the edge.
Where that edge is located is relative; the term covers a range of locations, architecture and application environments, each located closer to the source of the data than the alternative of centralized cloud processing. Topio’s Edge Computing Landscape, which is harmonization with the LF Edge’s taxonomy on edge computing, characterizes the space as split into four segments: the Access and Regional Edge, the On-Prem Edge Device, the Smart Device Edge & the Constrained Device Edge.
Still Early at the Edge
This migration of processing to the edge is an inevitable long term trend that will take a decade or two to play out fully, and we are still very early. In the first phase of the edge we have seen the highest volume of deployments in environments like factories, supply chains and retail stores. Here companies have deployed the edge to enhance their visibility of the shop floor or business environment, and gain insight into areas of potential improvement, such as increasing the reliability of machines in a production line, early detection of faults, or closely monitoring the movement of inventory in a store or supply chain.
The majority of these deployments are in the On Prem Edge and Smart Device Edge categories, in that they utilize compute resources attached to machines, or in the retail premise or factory, which then communicate learning and expectations to the cloud; and the cloud then communicates resulting new AI models back to the edge.
This provides excellent results for these customers with all the known benefits of edge computing – much reduced data migration costs, inherently low latency due the proximity of the devices, and, important for many customers, guaranteed privacy and control over data. Yet there are of course trade-offs of these benefits with the inherent inefficiency of dedicating resources to one, a few, or even in the best cases, a dozen or so applications.
Multi-Tenancy to the Fore
In the next phase of Edge growth, the Access & Regional Edge will be more prominent, and applications will be deployed in shared intelligent infrastructure that offer economics, flexibility and ease of use that is familiar to users of the cloud. One such environment would be the Public Infrastructure Network Node being built out by the Autonomy Institute, which brings together sensors, connectivity and edge compute – initially from EDJX – in one pole, deployed along smart corridors. Another would be Vapor.io’s INZones, deployed around major cities in the US; another still would be the Wavelength Zones being deployed by Amazon Web Services in conjunction with carriers such as Verizon. Each of these environments offer differing edge capabilities in different flavors that will appeal to different customers and use cases. But all offer the same fundamental multi-tenancy economics.
Capability the Killer
Much like there is no one killer app for the smartphone or for the cloud, there will be no one killer app for the edge; instead, like those examples, it is the flexibility and capability of a multi-tenancy environment that will ultimately drive the next hockey stick growth phase of the edge. Just like those examples, there will be different flavors on offer; although likely over time multiple environments will become fully interoperable, or consolidate into a handful of options. The proliferation of available edge nodes will drive utility and opportunity; and we believe Metcalf law applies – the value of a network is proportional to the square of the number of its user (n2). This opportunity gives rise to the explosion in Edge Native Applications – apps designed with edge capability in mind – that will define the next decade of application development.
Gavin Whitechurch, Principal Edge Analyst, Topio Networks