Since a overwhelming variety of Internet-connected devices can be envisioned and even existing today, some kind of classification of constrained devices is needed. Bormann, Ersue & Keränen suggested a three tier classification in their IETF document that reached RFC status in 2014 and has been referred since as a baseline classification. This classification is illustrated in Table 1. They based their classification on distinguishable clusters of commercially available chips and design cores available for constrained devices at the time of writing the document. These boundaries of these classes are expected to move over time, but not as fast as in larger scale of computing.
Moore's law tends to be less effective in embedded space and the gains made available by increasing transistor count and density will likely be invested in reduction of cost and power than in increases in computing power. (Keranen et al., 2014)
TABLE 1 Classes of Constrained Devices
Name Data size (RAM) Code size (ROM/Flash)
Class 0, C0 < 10 kB < 100 kB
Class 1, C1 ~ 10 kB ~ 100 kB
Class 2, C2 ~ 50 kB ~ 250 kB
Class 0 devices are very constrained sensors, with so severe memory and processing constraints that they are unable to communicate directly with the Internet in a secure manner. Class 0 devices need the help of larger devices acting as proxies, gateways or servers to participate in Internet communications. Generally they cannot be secured or managed comprehensively in the traditional sense, but they will likely be preconfigured and will be rarely reconfigured, if at all. (Keranen et al., 2014)
Class 1 devices are quite constrained in code space and processing capabilities. They are not able to employ a full Internet protocol stack and not able to communicate to other nodes using HTTP, Transport Layer Security (TLS), other related security protocols and XML-based data representations.
Instead Class 1 devices are capable enough to use a protocols stack designed for constrained nodes including CoAP over UDP and special implementations of Datagram Transport Layer Security (DTLS). This enables them to communicate without the help from a gateway node, so they can be integrated as fully developed peers of an IP-network. But their state memory, code space and often also power expenditure set limits to protocol and application solutions.
(Keranen et al., 2014)
Class 2 devices are less constrained and so capable of supporting most of the protocol stacks of normal Internet nodes. However even in this level the devices can often benefit from lightweight and energy-efficient protocol usage and from consuming less bandwidth. The use of protocol stack defined for more constrained devices on Class 2 device leaves more resources available for applications, since they will be using fewer resources for networking. This might also reduce development costs and increase interoperability. Devices significantly beyond minimum level of Class 2 are less demanding on the protocols used, but can still be constrained by a limited energy supply.
(Keranen et al., 2014)
4.1.1 Classifications based on energy limitation
As mentioned earlier the available power and energy is also a limiting factor for constrained devices. The power and energy available to a device can differ from kilowatts to microwatts and from unlimited to hundreds of microjoules. Watts determine the sustainable average power available for the device over the time of it is functioning. Joules determine the total electrical energy available before the energy source is exhausted. Devices can be limited both in available energy and available power. Bormann, Ersue & Keränen (2014) describe a four level
classification for energy limitations that is illustrated in Table 2. (Keranen et al., 2014)
TABLE 2 Classes of energy limitation
Name Type of energy limitation Power source example E0 Event energy-limited Event-based harvesting
E1 Period energy-limited Periodically replaced or recharged battery E2 Lifetime energy-limited Non-replaceable primary battery
E9 No limitations to available energy Mains-powered
Devices classified as E0 have limited amount of energy available for a specific event, such as a button press in an energy-harvesting light switch. E1 classified devices have a energy limitation based on a specific period. Examples of this kind of devices are a solar powered device with limited energy stored for night, device that is manually connected to a charger or a device that needs it's battery replaced in certain intervals. E2 device has an total energy limitation for its usable lifetime and it may be discarded when its non-replaceable primary battery runs out. When no relevant limitations to energy exist the device is classified as E9. (Keranen et al., 2014)
In the case of wireless devices the radio transmissions cause a big portion of the total energy consumption of the device. The parameters of the radio transmissions influence the power consumption during transmission and reception. These parameters include the available spectrum, desired range and the bit rate. The duration and number of transmission and reception including waiting for incoming messages influence the total energy consumption of a device. Depending on the energy source and communication frequency different strategies for power usage and network connectivity may be used.
(Keranen et al., 2014)
There are three strategies in the device level for power usage and they can be described as follows.
Always-on: No need for power saving measures, so the device can stay on and connected to the network all the time.
Normally-off: The device sleeps long periods and reconnects to the network when it wakes up. In this strategy the main area of optimization is to minimize the effort needed for the reattachment process and resulting application communications. If the device needs to communicate infrequently, the increase in energy expenditure during reattachment may be acceptable.
Low-power: This strategy is suitable when devices need to operate on small amount of power, but still need to communicate in relatively frequent basis. This strategy requires that low-power solutions are also available in the hardware and and link-layer mechanisms. These devices retain their attachment to the network in some form, despite they may have a relatively short sleep period between transmissions. This strategy minimizes the power usage
needed for reestablishing communications. An example of this strategy is duty cycling where components are switched on and off in a regular cycle.