IoT Protocol Comparison: MQTT vs CoAP vs HTTP

IoT Protocol Comparison

Payload size 100 B

Message rate 1.0 msg/s

Network packet loss 5%

Avg. latency (ms)

MQTT –CoAP –HTTP –

Energy per message (rel. units)

MQTT –CoAP –HTTP –

Overhead sent so far (bytes)

MQTT –CoAP –HTTP –

How it works

Three constrained "devices" send the same payload, at the same rate, over the same lossy link — but each protocol wraps that payload in a different amount of overhead before it reaches the broker/server. HTTP opens a request/response over TCP (plus a handshake if the connection isn't kept alive), MQTT publishes over one already-open persistent TCP connection, and CoAP fires a compact binary datagram over connectionless UDP. Each packet you see is drawn as two segments: a dark block for protocol overhead and a colored block for your payload, scaled to their actual byte size.

energy ≈ (payload + overhead) × E_byte + E_connection
latency ≈ L_base(protocol) + lossProb × T_retransmit

Payload size — bytes of real sensor data per message; overhead stays fixed while this grows, so its share of the total shrinks.
Message rate — how often a new packet is sent on all three lanes at once, for a fair side-by-side comparison.
Network packet loss — chance a packet needs a retransmit; every protocol here is used reliably, so a lost packet costs extra latency and energy instead of vanishing.
Persistent connections — toggles whether HTTP reuses its TCP connection (keep-alive) or reopens one per message; MQTT and CoAP are unaffected since they're already persistent/connectionless.

Real-world relevance: this is why battery-powered sensors default to MQTT or CoAP — the per-message overhead and connection energy of plain HTTP add up fast on a coin-cell budget, even though all three protocols can carry the exact same reading.

MQTT CoAP HTTP Protocol overhead

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.