Low-Power Wide-Area Networks at the IETF

A new breed of wireless technologies has emerged under the generic name of low-power, wide-area (LPWA), and with a number of common characteristics that make these technologies uniquely suitable for Internet of Things (IoT) applications. These common characteristics include a power-optimized radio network, a simplified network topology, frame sizes in the order of tens of bytes transmitted a few times per day at ultra-low speeds, and a mostly upstream transmission pattern that allows the devices to spend most of their time in low-energy deep-sleep mode. These characteristics enable a range of several kilometers and long battery lifetimes, possibly ten years operation on a single coin-cell. It also enables simple and scalable deployments with low-cost devices and thin infrastructures. LPWA-type capabilities fit a wide variety of use cases in IoT that require only a very low rate of data reporting and for which mains power or frequent battery swaps are not an option.

Counterintuitive as it may be for networking engineers, who are used to ever increasing throughput and reliability, an interesting variety of LPWA use cases do not need a high throughput and are not very sensitive to packet loss or transmission latency. Such use cases include the following:

• Humidity measurement in soils
• Corrosion monitoring on silos and tanks
• Snow or natural water levels outdoors
• Presence and/or rough location of goods (e.g., cars in a manufacturer’s parking lot)
• Detection of vibrations in an engine (e.g., indicating an increasing chance of failure in the coming hours or days)

In these types of use cases, each reading is of low individual value since the data that is reported is usually simple and mostly constant, such as a bit expressing an “OK” status. This is why, to a large extent, that type of data was not effectively reported so far; apart from very specific industries such as oil and gas, the cost of deploying wires would largely overwhelm the value of the reported data, and the rare event of production stopping due to a failed vent or waste of irrigation water would be accepted as irremediable. Though the value of one individual datum may be negligible, the overall value of all this unmeasured information is now seen as the next goldmine for optimization of processes in all sorts of industries, and leads to the industrial Internet effort.

To enable typical LPWA use cases, the cost of a monitoring device must be kept very low, the deployment must be very secure yet as simple as peel-and-stick, and the operational expenses per device must be kept minimal during the device’s life cycle. Another challenge comes from the large amount of monitored things—possibly thousands or tens of thousands for one application—and the span of the deployment, which often spreads over a very large area, far wider than that traditionally covered by wireless LAN (WLAN) and low-power wireless PAN (LoWPAN) technologies, and roughly in the range of wireless neighborhood area networks (Wi-NAN) and cellular technologies.

Both possible approaches—evolve the low-power wireless technologies to reach farther out or evolve cellular technologies for lower cost and energy consumption—could be taken to design LPWA Networks (LPWAN). In fact, all possible paths are being explored, yielding a variety of novel technologies with a range of specific capabilities (Figure 1).

For operation in the licensed band, 3GPP has standardized a new narrowband radio technology called NB-IOT that offers low complexity, low power consumption, and long range. NB-IOT can also utilize the existing LTE and legacy infrastructure and coexist in the same frequencies for reducing time to deployment. 3GPP has also standardized a category of low-complexity low-bandwidth UEs (called Cat-M1) for LTE, and IoT related improvements to GSM/GPRS networks called EC-GSM-IoT.
Another set of LPWA technologies (e.g., LoRa, SIGFOX, INGENU) was designed to operate on the unlicensed industrial, scientific and medical (ISM) radio bands, with capabilities for reaching tens of kilometers with data rates in the order of tens of Kbps, using sometimes very diverse kinds of radios, from SIGFOX’s Ultra Narrow Band (UNB), which uses a thin peak of spectrum, to LoRa’s Chirp Spread Spectrum (CSS), which spans all the available bandwidth. Surprisingly, this variety of technical approaches appears to be beneficial to the end user as it provides an additional form of diversity, and therefore also an improved immunity between deployments competing for the same portion of spectrum. These new technologies complement capabilities that are already in operation in the ISM band with IEEE802.15.4 for applications such as the smart grid (Wi-SUN) (Figure 2).

The balance between cost, energy budget, service guarantees, and manageability indicates that an optimization has to be found for each individual application, and that different radio technologies with different capabilities and different service offerings will continue to coexist—each serving the applications for which it is best suited.

This variety of choices, as well as the potential to adopt new types of radios and services as technologies and needs evolve, is the enabler of a wide variety of ecosystems and applications, and a core value of the LPWA approach (Table 1).

Table 1. LPWA technologies compared

This diversity could also be a threat if the complexity that appears as its downside cannot be controlled—for instance, if each technology comes with a different application, identity, security, and service management model, making it impossible to migrate an application as its needs evolve, or if the design of the cloud side of the application is so very different from one technology to the next that no software or operational skillset and toolset can be factorized.

In order to avoid a combinatory explosion of complexity, there is a clear need for a convergence—an hour-glass model—at a layer above the radio and in a fashion similar to what IP provides for the Internet.

IPv6 and CoAP can act as a potential convergence layer for LPWAN, providing device reachability and yet relative isolation in a manner that can abstract the actual underlying radio technology. A new LPWAN Working Group was formed after a successful WG-forming BoF at IETF 96 in Berlin; the LPWAN WG meets for the first time at IETF 97 in Seoul. The WG will address IPv6 over LPWA Networks, and how LPWANs can join the Internet community in a mutually beneficial way.

A Generic LPWAN Architecture

At first glance, LPWA technologies appear diverse and not all requiring the same work items from the IETF. For example, Wi-SUN already supports IPv6 with 6LoWPAN, and may not need additional work in that area. But Wi-SUN may still benefit from other components, such as security and identity management, that may or may not be of interest to other parties due to their existing support.

A closer look reveals that LPWAN technologies usually share a very similar structure with both radio-layer gateways (RADIO-GWs) that connect to end devices and network-layer gateways (LPWAN-GWs) that aggregate multiple RADIO-GWs and enable connectivity to the outside world. LPWAN technologies also share a desire to enable IP technology between network applications and the applications that reside in end devices in order to promote more portable services and more generic tools (Figure 3).

For the IETF, it is critical that a generic architecture is derived that enables some common work between technologies. A large part of the IETF’s work is to identify common needs for functionalities in the LPWAN gateway (GW), and to standardize the protocols that enable these functionalities. The new architecture must focus on the core similarities of the LPWAN technologies, such as the need to optimize battery life, very sporadic traffic, and low throughput. These extreme requirements place the bar beyond what could be achieved with 6LoWPAN for low-range, low-power wireless technologies.

Leveraging this architecture, the IETF value proposition is to converge the diverse radio technologies towards a common hourglass model with an extremely compressed form of IPv6 and CoAP between the end-device and the network gateway, in order to both provide a common management of the gateway and enable secure, Internet-based services to the applications.

The Need for New Algorithms

Some LPWA technologies are highly sensitive to frame size, due to their very low data rates. As a result,  classical compression techniques may not be enough for technologies in which only an octet or two are available to signal both IP and CoAP. At the extreme, robust header compression (RoHC) may provide that level of compression, but the learning curve in traffic and the variety of flows prevents using that technique as is.

The IETF needs to do more work on this, perhaps taking the best of both compression techniques and providing the extreme compression needed to enable IP and CoAP to the device. This work may also enable the LPWAN-GW to terminate the IP flows, in which case the end device would appear as a remote I/O to the gateway, like a USB device would to a PC.

The new protocols must take into account application-level protocol compression (e.g. CoAP) and employ all possible mechanisms and LPWAN specificities to achieve optimal compression, most notably the typical star topology and the limited and known-in-advance flows per end-device.

Beyond Compression

If reachability to the device through IP was the only goal for new work at the IETF, the creation of a specific WG for LPWAN would probably not be mandated. Bringing CoAP to the end-device also allows for the use of the IETF interaction model developed in CoRE WG and the T2T RG and recently adopted by Open Connectivity Foundation (among others). But a lot more needs to happen over the Internet in order to offer safe and available services for low-power IoT devices. The recent attacks on Krebs’ blog, On Security, demonstrated that IoT devices cannot be protected and maintained like traditional PCs. Long-term guarantees that the device will not be exploited for harmful activities are needed. This may be achieved by automating posture management and software updates to patch vulnerabilities as they are discovered and by isolating devices from potential attackers and targets.

In addition to classical practices, IoT devices must be strongly protected from attackers seeking to gain information, influence, and use them for their own purposes. On the one hand, communication between the IoT device and its application should be available at all times, whether the application is in initial commissioning, normal operation, or decommissioning. On the other hand, unwanted communication to or from the IoT device should be barred at all times in order to limit the effects of compromised devices on the rest of the network and to limit the potential for remotely compromising more devices.

One simple and solid way to provide such protection is to isolate the device and its application in an overlay network that operates on unique local addresses that cannot be reached from the outside of the overlay. On-demand and scalable overlays are not the only requirement from IoT on the Internet that the IETF could help fulfill. With the advent of connected vehicles, as well as for simple use cases, such as homeowners moving with all their connected appliances, IoT devices must be mobile at the IP layer.

IETF overlay technologies, such as NEMO and LISP, enable mobility with traffic isolation, but they have different approaches to security and scalability. Matching solutions to actual needs and real-world use cases requires a combination of skills from the real-world practice of LPWA combined with expertise in Internet technologies.

IPv6 Activities thus Far

The initial work presented in the T2T RG at IETF 93 was a first draft. It showed a glimpse of the potential and the constraints of low-power, wide area networks and what could be done in these networks at the IETF, including, but not limited to, security, mobility, device management, and network and service discovery. The main motivations for the draft were to present the problems LPWANs are facing and to find a way to do the work in existing Working Groups. Multiple discussions during both IETF 93 and 94, as well as with the new industry players, indicated that we needed a new, focused place where this work could either be done, as no existing group is really chartered for it, or be coordinated from when there is an existing group. The rich discussions on the non-WG mailing list, and the active participation in the non-WG-forming BoF at IETF 95 in Buenos Aires further reinforced this view. The critical point was the specific proposal of the Static Context Header Compression draft, which demonstrated for the first time a practical way to bring the IETF stack to these ultra-constrained networks. This, along with the IETF’s robust standardization process, motivated the four major LPWAN technologies—SIGFOX, LoRa, Wi-SUN, and 3GPP—to support the creation of the LPWAN WG and to mandate the IETF to bring its stack to their LPWANs. The WG also plays an important role in bringing together different communities with many people new to the IETF.

Charter and Roadmap

First steps will be focused on new forms of IP/UDP/CoAP compression, the cornerstone of future work in this field. Milestones are on a very tight schedule (with a goal of finalizing the work by mid-2017) and correspond to the immediate demand of the four baseline technologies. This first stage will also help structure the LPWAN community and prepare the eventual extension of the work, where additional items can be addressed. These may include the Radio-GW and LPWAN-GW management protocols, end-device mobility and AAA procedures, overlays, security, and the use of DNS at the core of these networks, among others.