kantver - Fotolia
With emerging trends, such as 5G and edge computing, tech experts will find that advice on what technology they need quickly changes with each next big thing. In the case of IoT, 5G and edge computing offer enticing advantages, but experts warn against going all in on either technology without considering their context and longevity.
"As always, in IoT and in life, [you want] the right tool for the job," Joseph Biron, CTO of IoT at PTC, said during MIT Enterprise Forum Cambridge's Connected Things 2020: WFH Edition panel, titled "Point-Counter-Point: 5G vs Edge, Which Will Disrupt IoT More?"
The improvements 5G promises mean that storing IoT data in the cloud could fulfill organizations' needs better than previous connectivity options. But moving that data closer to the edge also offers advantages for different applications.
"There's a slider that we can move between 'everything happens' or 'almost everything happens' in the cloud versus everything happens at the edge," Biron said. "The slider should never be on any one of those poles."
To determine how much compute should reside at the edge and how much should take advantage of the cloud, IT experts must consider the pros and cons of each option specific to their use case.
"The question is how much compute do you put on a thing. We're coming out of a 10- to 15-year period where the assumption was you put skinny compute on a device because [the data] all belongs in the cloud," Frank Gillett, vice president and principal analyst at Forrester Research, said. "Now, there's a trend that you need to put compute on the edge."
Know when to add 5G to IoT deployments
To make a connected device, organizations must choose the connection option, which includes deciding how much compute should be put on the device at the edge.
"We're about to get a new generation, the once-every-decade version of the telco connection and there's a bunch of magical, wonderful, stupendous improvements in 5G that mean you shouldn't have to compromise," Gillett said.
5G offers higher bandwidth, lower latency and, eventually, lower costs, which means organizations might not necessarily need to put as much compute or storage at the edge. The connectivity option will also use less energy for IoT devices, which means batteries can last longer than with 4G LTE connectivity. If organizations choose a network from a provider, the provider also manages the network, which decreases the work for IT administrators. Organizations might also choose their connections differently if they are connecting something that moves around or is very expensive versus a cheap connected device that stays in one place, Gillett said.
With cloud computing, organizations can increase their bandwidth through 5G connectivity to prevent the need to retrofit IoT deployment hardware and distributed software architecture for the edge. But the rollout of 5G does not guarantee that the right bandwidth will be available where an organization needs it. Providers will offer three types of 5G: long-range and low-band in rural areas, mid-band in metro areas, and millimeter wave or high-band in places where having transmitters on every block makes sense. If IoT devices will be in an area with high-band connectivity and have a clear line of sight to an antenna, it wouldn't necessarily make sense to use edge computing versus cloud computing because the 5G shortcomings wouldn't be an issue, Gillett said.
"You should put 5G connectivity in if you want to bet on telco connectivity in the first place," Gillett said. "But you need to dig deeper into your requirements and context here to determine 5G or edge. Nonetheless, for anything with any legs on it, you want to bet on 5G because this will be the standard for the next 10 years."
Review current technology before committing to edge computing
IoT devices are fundamentally distributed architecture. Organizations would want edge computing when operational staff use a local app on a smartphone or an augmented reality display and need to know what's going on with a machine immediately, Biron said. Training machine learning models and scoring at the edge where the data is created gives organizations an advantage if they need real-time compute and decreased latency. But edge computing gives organizations a piece of IT that needs to be managed by operational technology experts where the edge devices reside, which can lead to IT/OT convergence challenges. The increased level of complexity from distributed architecture must be as automated as possible to keep operations running smoothly, Biron said.
"The value of edge is instant response without any delays or dependencies on the network. It gives you more control," Gillett said. "The tradeoff is you have to rewrite and rethink the software."
Organizations face another difficulty when implementing edge computing in an IoT deployment if they must retrofit hardware and software that used cloud computing. It's not very practical to take software that was built to run in the cloud and distribute it on the edge, he said.
Joseph BironCTO of IoT, PTC
Consider the durability and upgradeability
Whether the IoT use case calls for prioritizing 5G connectivity or edge computing, experts recommend choosing hardware and software based on long-term planning.
If organizations want their IoT devices to run in the field for the next 10 years without needing to retrofit them later, investing in 5G and a better radio when it fits their requirements would be the better option, Gillett said. For edge, organizations must consider if they are willing to invest more in hardware, even if the software doesn't yet meet their needs.
The sooner they wish to implement and are likely to replace gear, the better off organizations will be buying only what they need now, Biron said. For devices that organizations will use for longer periods of time, the more they should plan for future technology, including 5G connectivity. Both options for handling data depend on the context and needs of the organization.
No matter the longevity of the IoT deployment's use, being capable of upgrading security is critical.
"The reality is a bunch of stuff out there in the world is communicating with security protocols that a teenager could crack. [It doesn't] even have to be a smart teenager," Biron said. "We need to get the world onto a mindset of this stuff has to be upgradeable."
That means both overbuying hardware so it will have longevity, and the modularization of the IoT hardware package because organizations need to be able to swap it out on long-lived assets.
"Common principles for a smart connected product design would be to plan for decades, from the radio to the hardware to the software stack," Biron said. "Plan for your products' expected useful life."