Most IoT architectures in the business world are complex. At a minimum, they usually require some sort of intelligent gateway to filter and aggregate data coming in from sensor networks or to take real-time action based on that data. The most sophisticated IoT environments have even more tiers to handle complex business rules and events, security, high volumes of data and redundancy.
The distributed computing architecture and the manner in which networks tie that architecture together matter a lot. This is where edge computing and 5G come in.
What of the edge?
Edge computing covers many industries and use cases including IoT. Edge is about distributing application data and services in the most effective way. This often means provisioning closer to users or edge devices. In an IoT context, that might mean locating intelligent gateways or even clusters of computers outside of traditional datacenters. It also includes more traditional forms of distributed computing such as remote branches and distributed retail.
One way that edge computing has evolved from narrower IoT concepts is the recognition that edge computing must be a part of a broader hybrid cloud strategy for an organization. This strategy requires some consistency throughout hybrid clouds, from edge devices to the network to the centralized datacenter.
Designing architectures for edge computing requires thinking about parameters such as bandwidth, latency, network reliability and cost. That’s always been the case with non-trivial IoT projects. But considering tradeoffs in the context of an enterprise IT architecture means thinking about a broader set of operational concerns beyond a specific IoT silo.
A recent discussion at a virtual MIT Enterprise Forum Connected Things event in May illustrates this broader thinking nicely. Joe Biron, CTO and General Manager of IoT at PTC, noted that IoT often involves the convergence of IT and OT. This might imply putting computers near the equipment that’s being monitored and controlled.
“Do I want to have all that gear out in the factory? Maybe do training centrally? It’s a big slider. The elephant in the room is that the distributed architecture has to be automated as much as possible,” said Biron.
The convergence of IT and OT at the edge presents additional challenges as well. Biron notes that something such as an edge gateway might get swapped out every three years in a similar manner to regular IT equipment. As a result, it doesn’t make sense to build in capabilities that aren’t required over that time horizon. On the other hand, the longer life cycle of dedicated hardware, such as an ATM, should make you think about future-proofing the design with communications links — such as 5G — and other key technologies whether or not you can take full advantage of them immediately.
IoT and 5G are often conflated with edge computing. As we’ve seen, IoT is best thought of as an important use case for edge computing, while 5G is one of the most important associated enabling technologies.
“The idea is that 5G means you don’t have to put as much stuff at the edge. And one argument for better connectivity is that you maybe don’t need to rethink distributed architecture,” said Frank Gillett, Vice President and Principal Analyst with Forrester.
The reasoning is that 5G can speed up communications between edge devices and central systems, which enables IoT architectures that aren’t as dependent on heavy-duty processing at the distributed edge.
“Of course, it’s never zero latency and infinite bandwidth,” said Gillett.
This is meant to serve as a warning about putting too much faith in 5G or any other technology as a magic bullet for distributed computing challenges. Hardware in remote locations will fail and network links will be interrupted. You’ll probably need to curate the data that’s transmitted back to a datacenter.
Fast local actions in response to events probably require decisions that are also local. Finally, there is so much data being generated locally — much of it redundant — that it’s just not useful or economically feasible to send it all home to a centralized location where much of it will sit unused.
The general hybrid cloud trend has been leaning towards spreading computing out. Many companies are now spending more on the network, computing and storage resources in edge locations than on upgrades in core datacenters.
5G implementations are often themselves edge architectures that support software defined networking and other functions. Edge is the next step in the transformation journey of telecommunications suppliers that started with OpenStack and are continuing with container platforms such as OpenShift as they move to distributed architectures.
Wireless networking technologies have always affected the choices we make when designing architectures to solve business problems, and 5G is no exception. And it might even have a bigger impact than many other technologies. However, it’s best viewed as one technology element among many that go into an edge architecture.
The evolution of edge as a concept will also influence IoT. Many edge examples involve IoT at least in part, which highlights how edge is pulling in distributed computing concepts as part of many organizations’ hybrid cloud strategy.