Low power wide-area networking alternatives for the IoT
Wireless network technologies such as WiFi, ZigBee, and Bluetooth are fine for consumer applications of the Internet of Things (IoT), but many civic, industrial, and other IoT applications need to operate over vastly greater territory than these technologies can handle. Cellular and satellite machine-to-machine (M2M) technologies have traditionally filled the gap, but cost, power, and scalability concerns make these choices less appealing for the future. A number of low-power, wide-area networking (LP-WAN) alternatives have arisen that need careful consideration by developers looking to address these wide-ranging IoT applications.
The uses for wide-area IoT technology are legion. Civic infrastructure systems such as parking resources, traffic control, utilities monitoring and distribution control, and environmental monitoring are only a beginning. Agricultural uses such as monitoring of crop conditions and livestock movements need wide-area coverage. Asset monitoring and tracking, from taxicabs to refrigerated produce shipments need regional, national, or even worldwide coverage. Transportation infrastructures such as rail lines and roadways need wide-area monitoring. Even consumer applications such as health monitoring could benefit from having an alternative to cellphones for their wide-area connectivity.
While the applications are diverse, they have many common attributes on their network wish lists. These include:
Low cost – Most wide-area IoT applications anticipate a need for many hundreds or thousands of end-node devices for each installation. In some cases, such as city-wide parking space/meter monitoring, the numbers can get into the millions. With such high volumes, unit price is a major consideration in determining the return on investment (ROI) for the application.
Low energy consumption – Few of the applications for wide-area IoT have the luxury of a local power generator. Most will depend on batteries and some may even need to use energy harvesting. For those with batteries, replacing depleted batteries can represent a major logistical challenge as well as a substantial cost. The longer the battery life in the end node device, the better.
Extended range – All wireless networks connecting to the Internet need to work through an access point (AP) of one kind or another: gateway, concentrator, or the like. So an IoT design needs to consider both the endpoint cost and the cost of the access point infrastructure needed to support the application. The network's operating range, or allowable distance from an end node to its access point, can have a significant impact on that infrastructure cost. Range dictates the number and location of access points needed to cover the application's operating area, so in general the longer the range the lower the infrastructure cost.
Scalability – A given installation using a wide-area wireless IoT network may work well and the network may well have the capacity to handle any anticipated single user. But over time it's reasonable to expect that many different installations will be made in the same geographic area. If these different installations share common access points, like cellphones share towers, then the number of devices an access point can support can become a limiting factor and require increases infrastructure to overcome. Even if they don't share access points but do share the frequency spectrum, an increase in installations can erode the operational range of application through increased noise levels. In the worst cases, available channel capacity can fill and prevent new installations from operating at all.
Among the more established wireless networking technologies, only cellular and satellite communications offer the extended ranges that these applications require. Mesh networks such as ZigBee can potentially cover large areas but have limited scalability due to the need to forward traffic.
Unfortunately cellular and satellite communications technologies short in the other attributes. Their radio requirements involve higher energy use and complex protocols that lower battery life and increase cost beyond what many applications can sustain. This arises in part from their history; they were originally designed to handle voice traffic. The networks are ill adapted to handling short data messaging.
Still, some IoT applications and services – often called machine-to-machine (M2M) – did arise to leverage cellular and satellite communications networks. Many of them were based on the CDMA, or "2G" cellular technology. Unfortunately, those networks are now starting to be phased out by service providers in order to free spectrum for more advanced cellular technologies. However, the cellular community has made some strides toward improving the situation for M2M. The most recent specification for LTE (release 12) defined communications Category 0 designed around the needs of M2M traffic. Energy use and cost still remain concerns, however.
This situation has opened a door for alternative approaches to wide-area wireless networking for the IoT, approaches that focus on the low-power, low-cost requirements. At least six different approaches are currently defined with network deployment growing or getting started, and three more are in development. While all these approaches seek to provide the same key core attributes, they have different takes on numerous other system attributes that can affect their suitability for various IoT applications.
Desirable LP-WAN Attributes
These other attributes that vary in importance among applications but still need consideration include:
Roaming – Many applications call for end nodes to be fixed in their position, but others may require that nodes operate while moving within and even across sectors served by different access points. Most wide-area IoT networking alternatives allow movement of nodes from one sector to another, but they can vary in how quickly the network adapts to the altered relationships.
Penetration – Some applications call for the end node to be located inside a building or underground while the access point is in another room or outside and above ground. In these applications the network's range can be considerably reduced by the absorption of walls and dirt. Such absorption is frequency-dependent, with lower frequencies generally offering better penetration than higher ones.
Short message handling – While some IoT applications will need to send substantial amounts of data frequently, many will need to send only brief messages, often infrequently. The ability of a wireless network to handle short messages efficiently can have a beneficial effect on the network's scalability and the end node's energy consumption. Such handling includes any overhead for connection setup, interrogation, acknowledgement, or the like.
Bidirectional communications – Some end nodes may only have a need to report data, not receive commands, so a unidirectional link may seem adequate for such applications. A bidirectional link, however, allows for such things as handshaking with the access point to improve the reliability of data transfers, authentication exchanges for greater security, and with sufficient bandwidth allows for remote software updates and management of end nodes.
Secure communications – Sensitive data will need a secure communications link between end node and access point, but even if the data is not sensitive, security may still be a concern. Without a secure link an IoT application is more vulnerable to such attacks as spoofing, where a fraudulent end node injects false data into the network or a fraudulent access point hijacks end node data.
Higher level services – A given wide-area IoT networking alternative may define any number of levels in the OSI model, from just the physical and data link layers through the application layers. In some cases the network itself is operated and managed by a service provider that leases time on the network to users running their protocols and provides users with cloud services. Other alternatives define only the lower layers and have their access point connect to the Internet or to a private network, leaving the higher OSI layers to the user's choice. In such cases an ecosystem of higher-level service providers usually becomes available over time.
The various low-power, wide-area networking schemes on offer address these many needs and considerations in a variety of ways. Each has made a different choice of tradeoffs among interacting attributes such as battery life, data rate, operating frequency, achievable range, and scalability. Further, they have made different choices around attributes such as security, OSI levels defined, and roaming support. This diversity makes it impossible to provide a comprehensive side-by-side comparison, but it is possible to provide a start.