I'm working on a small embedded sensor project that needs to run on battery for several months. For the network layer, I'm torn between three widely used IoT protocols: MQTT, which is lightweight and based on publish/subscribe; CoAP, designed for constrained environments with a request/response model; and HTTP/REST, which is very common but has more overhead. Which protocol do you think is best suited for low-power sensors, and why?
What IoT protocol should I choose for low-power sensors?
👁️ 109 views💬 2 replies❤️ 0 likes
2 Replies
For battery-powered sensors that need to run for months without maintenance, I’ve preferred CoAP in several projects. The protocol runs over UDP, has a header of just 4 bytes, and avoids TCP handshake overhead—especially beneficial for short, regular sensor readings, where it significantly improves the energy balance. CoAP also integrates easily with DTLS without drastically increasing packet size, and its *Confirmable* messages provide reliable delivery feedback without requiring a full broker like MQTT.
If you’re already using an MQTT ecosystem or prefer Pub/Sub architectures, MQTT with QoS 0 and a very short keep-alive interval can also work well—I tested this with a battery-powered temperature node that ran continuously for 6 months with 5-minute intervals. Still, CoAP remains the most energy-efficient choice when you primarily need point-to-point communication and minimal header size.
In my DIY weather station project, I finally went with CoAP: its very low overhead and request/response model allowed my battery-powered sensors to last three months without frequent wake-ups, whereas MQTT required an always-on broker and consumed a bit more power. In practice, I found that 10-byte CoAP messages are enough to transmit measurements without bogging down the radio!