Massive IoT Interop Fuels Protocol Battle

Wireless standards are plentiful, but most are not capable of being scaled to the level of a smart city.

As a result, such networks have been built application-by-application using proprietary stacks, often with non-interoperable network layers. That, in turn, has slowed the proliferation of dense wireless connectivity at scale.

“In a hyper-connected world, connectivity choices are driven by use cases,” said Frank Schirrmeister, senior group director for solutions and ecosystem at Cadence. “We find a sliding scale of requirements for these use cases, from smart meters and asset management requiring low power with low throughput on one end, to public safety and patient monitoring requiring high bandwidth combined with low latency on the other end.”

This has a significant impact on the choice of communications technology. “When we talk about a smart city, you cannot say, ‘Hey, I’m building one application on my own network, but it’s going to be targeted only for us,’” said Abhjit Grewal, director of marketing for smart cities at Silicon Labs. “For a single application, the only way you can actually scale the network is if the network layer is interoperable.”

Ongoing efforts have resulted in a new protocol vying for dominance in massive IoT deployments. It’s designed both to operate at scale and to allow multiple applications to interoperate on a single IP-based network. This creates an opportunity for smoother development of efficient wireless machine communications in urban environments and other spaces that require nodes numbering in the tens of thousands and higher.

But this opportunity, regardless of protocol, hasn’t translated into excitement or even much new silicon in the semiconductor world. That raises questions about how attractive the IoT market is for hardware makers.

Unmet needs for the IoT at scale
There’s no shortage of wireless protocols that are suitable for use outdoors. Most of them, however, can’t handle massive deployment, since the full complement of nodes and traffic could bog them down.

“Zigbee Smart Energy and Thread, etc., are good for a few hundred devices,” said Phil Beecher, president and CEO at Wi-SUN Alliance. “But we’re looking at maybe up to 4,000 devices for every border router connection into a backhaul network.”

Grewal agreed this represents an unmet need. “Most of the technology stacks are not geared toward smart-city networks,” he said. “None of these stacks is going to support hundreds and thousands, or even millions of nodes.”

The Wi-SUN Alliance notes a number of companies offering proprietary technologies in this application today. They include Arch Rock (part of Cisco), Elster, Itron, Landis+Gyr, and Silver Springs Networks (now part of Itron). Their approaches are said to be similar, but they aren’t interoperable.

In addition, Silicon Labs has been involved in projects that were proprietary not to a networking company, but to the customer. “The majority of our smart-city solutions were proprietary, which meant that the customers were using our radios and then writing their own technology or protocol stacks on top of that,” said Grewal.

Meanwhile, technologies like LoRaWAN target low-bandwidth applications. Smart-city applications, in the aggregate, can require transport of far more data than such a network could handle.

What’s ideally desired in such environments is, at the very least, the ability to take a shared network based on something well known, like IP, and then have it carry the traffic for all of the applications being deployed. The stack, up to the transport layer, would be common across multiple applications.

“The application needs for a manhole detector or parking meter are going to be very different from the application needs of an electric meter or a streetlight, but they need to share the same network,” said Grewal.

This situation convinced a number of companies to come together to develop a new protocol stack called Wi-SUN. It is centered on an IPv6 network, which will allow easy interoperation of traffic from other networks that are also IP-based. The stack is defined up through the transport layer. Above that, application writers are free to do what makes the most sense for their application.

Further examination of the choices made in the Wi-SUN specification can help to illustrate the challenges of these kinds of deployment.

Fig. 1: A simplified Wi-SUN stack. A single protocol allows for multiple profiles for different applications. They will all be interoperable. Source: The Wi-SUN Alliance

Modifying for massive deployment
While it may have been easy conceptually to rally around an IP-based network, there were still many concerns that needed to be addressed in order for things to operate smoothly on a large scale. “802 15.4g is the underlying radio technology,” said Beecher. “And then we took a number of IPv6-type standards from the IETF, including the routing layer and security standards, and we integrated all of them into a single specification.”

The first architectural item is the fact that Wi-SUN defines a mesh network. That distinguishes it from some of the alternatives. “Wi-SUN’s mesh network structure, vs. the more typical star-structure used by LoRaWAN and NB-IoT, plays a role in making connectivity decisions,” noted Schirrmeister.

In a smart city, some nodes may be battery-powered, while others will be line-powered. Wi-SUN addresses this with two kinds of node. The line-powered nodes act as always-on anchor nodes, with some acting as gateways to other networks. The battery-powered nodes then have a duty cycle, powering up on their own schedule and connecting to the neighboring line-powered nodes as leaf nodes.

Fig. 2: A Wi-SUN mesh network. Blue arrows indicate links between line-powered nodes. Gray arrows show links to battery-powered leaf nodes. Source: The Wi-SUN Alliance

That a leaf node must be within reach of a line-powered node may sound like a limiting network-design requirement. But the Wi-SUN Alliance notes that it’s no different from standard cellular networks, where a battery-powered handset is connected only if within reach of a line-powered base station.

In fact, they say it’s stronger than the cellular system in one way. A cell phone can be connected to only one tower at a time. With Wi-SUN, more than one tower can participate in transporting data to and from a leaf node that may be located in a geographically challenging spot.

There are two fundamental versions of Wi-SUN, which define a smaller home-area network (HAN) and a larger field-area network (FAN). HAN implementations are much simpler, involving fewer options. Most of the discussion that follows applies to FAN installations, since they have the massive scale-out issues that a home doesn’t have.

Low-level options
The Wi-SUN PHY layer is based on the same 802.15.4 standard that drives Zigbee and other protocols, but it’s been modified for Wi-SUN, yielding 802.15.4g. The frequency band is 900 MHz in the US. “That brings you much greater range and better performance,” noted Beecher.

Beecher pointed to one of the downsides of this frequency choice. “The problem with the sub-gigahertz frequency band is there is no single band available globally,” he said. So they’ve planned out some of the channel options globally to minimize the number of possible variants.

There are a number of modulation options based on the bandwidth needed for a given node in a FAN. (HANs have only one PHY option.) More options may be added in the future, but they must be backwards-compatible with existing nodes on the network.

Modulation can be adapted dynamically, as available bandwidth changes due to degradations or improvements in the wireless link – so-called “gear-shifting.” This means the choice of modulation scheme doesn’t have to be a static one for all situations, or even for a given node.

Frequency hopping has been adopted for FANs to make more efficient use of the spectrum, to help with multi-path fading issues, and to improve security by making it harder to jam or hack. “If you do have things running over a large geographic area, you want to optimize the usage of the spectrum and minimize collisions,” noted Beecher.

Each device will have a pseudo-random hopping sequence, although problem channels can be blacklisted and skipped over. That blacklisting is dynamic, so if the situation improves, a channel can come off the blacklist. “If you’ve got lots of reflections off buildings, it can cause big gaps to appear in coverage,” said Beecher. “If you can dynamically change the frequency, you can avoid that.”

While most of these networks will exist in urban environments, some will also migrate to rural areas. But, by definition, those networks (or portions of the network) will have lower node density, making it possible to use lower bandwidth links – which will have longer range.

“As you’re moving more into the rural areas, you might not be running OFDM for higher throughput, but you might be running FSK or OQPSK, which gives you a longer range,” noted Grewal. “But you don’t need the same level of load management in the rural areas, and grid management becomes relatively easy versus an urban area.”

Fig. 3: Wi-SUN interaction with WAN connections as well as an illustration of possible use cases (along the bottom). Source: The Wi-SUN Alliance

Routing and security
While using IP for routing, the algorithm has been modified to use so-called ROLL routing – Routing Over Low-power and Lossy networks. This is part of a new routing protocol called the IPv6 Routing Protocol for Low-power and Lossy networks, or RPL. The IETF decided that this was necessary when a review of the existing routing protocols on challenging channels yielded none that was adequate when considering loss response, routing state, control cost, link cost, and node cost.

“This is pretty similar to what they used in the Smart-Energy profile,” observed Beecher. “But the parameters have been tweaked to allow you to connect thousands of devices onto a single border router.”

Security also has been improved, leveraging EAP TLS – similar to enterprise WiFi. EAP, or “Extended Authentication Protocol,” enables the use of X.509 certificates. “We’re using certificate-based authentication in all of the devices,” said Beecher. “When the device authenticates, it’s pretty similar to the way enterprise WiFi works.”

This level of security alone is said to have aroused attention. “There’s huge interest in Wi-SUN primarily because of the security enhancements,” observed Grewal.

Fig. 4: A detailed view of the Wi-SUN stack. Source: The Wi-SUN Alliance

Spinning up the market
From a supply standpoint, it might seem that the number of options would create a complex mix of devices. For instance, leaf nodes would use only lower-power PHY modes, while the line-powered nodes would have to be able to talk to any other node that might be using any of the PHY options.

In reality, the Wi-SUN Alliance says that multi-mode implementations easily can be created without adding complexity or cost. They expect such an approach to predominate, simplifying component choice.

FAN certifications, meanwhile, started a couple of years ago, and roughly 45 devices from nearly 20 suppliers have been certified. “When we give a certified logo, it tells the customer that it’s going to be able to interoperate on the same network as other certified devices,” said Beecher.

As to what will happen when this rolls out in areas that already have some smart-city capabilities, it may not be a restart. “There are quite a lot of networks in North America that today are Wi-SUN-capable,” said Grewal. “That means you can do an over-the-air update and make them FAN 1.0- or FAN 1.1- interoperable.”

Will this impact the semiconductor arena?
Despite these developments, Wi-SUN seems to have made surprisingly little impression on the broader semiconductor ecosystem. Many companies that you might expect to be aware of new protocols had no familiarity with Wi-SUN. And that may be, at least in part, because of lackluster interest in the IoT space more generally.

“The IoT hasn’t seen as many new design starts as hyperscale data centers or automotive are seeing,” said Priyank Shukla, staff product marketing manager at Synopsys. “So that’s the reason you will find IP vendors not so laser-focused on the IoT. It’s very difficult to make money as a hardware provider.”

The thought here is that the majority of the IoT opportunity comes from services and higher-level cloud-based capabilities, not the underlying hardware.

This may be particularly true of the consumer IoT, but industrial IoT applications are also not attracting big investments in hardware startups. Some see this realm as a closed ecosystem, with entrenched players that make it difficult for new entries into the market.

“Whenever a large industrial control company has a new need, they’re not going to fund a $20-million startup for this,” said Shukla. “They will just give this requirement to ADI or TI, and those companies will provide their own solution based on what they have.”

Nonetheless, based on the overall opportunity, the IoT has spawned dedicated variations of a number of protocols. When it comes to who will own the network layer, Schirrmeister pointed out a wide range of candidates for municipal and industrial installations.

“Different clusters like massive IoT, broadband IoT, and critical IoT are emerging, with LoRaWan, NB-IoT, LTE-M, LPWA, Wi-SUN, and variations of 5G like 5G-NR-Lite being candidates for the network and transport layers,” he said.

While the current focus is on the network, he’s also keeping his eyes on the application layer. “Like ‘Matter’ in the smart-home area, it will be interesting to see whether a unifying application layer will emerge that presents all the data to the ecosystems and cloud.”

Wi-SUN doesn’t attempt this – and that fact is marketed as a feature. Application-layer standardization may be feasible in the future as standard profiles emerge independent of connection protocol. At this point, however, converging on a single network is seen as solving the immediate bottleneck without imposing requirements at the application level. It’s just a question of whether that network will be Wi-SUN’s or someone else’s version.