PARAMETRIC EVALUATION OF MQTT 5 OVER QUIC UNDER MOBILITY SCENARIOS

Abstract

MQTT remains the dominant publish/subscribe transport for the Internet of Things (IoT). Two recent developments—the OASIS MQTT 5.0 specification and the IETF QUIC transport protocol—change the design space in ways that the existing literature has only partially explored: most published evaluations either fix MQTT 5 over TCP+TLS or compare MQTT 3.1.1 over QUIC against MQTT 3.1.1 over TCP, leaving the joint behaviour of MQTT 5 features (topic aliases, message expiry, shared subscriptions) and QUIC connection migration on multi-radio-access-technology (multi-RAT) mobile links largely unmeasured. We address this gap with a parametric analytical model of three transport stacks—MQTT/TCP+TLS 1.2, MQTT/TCP+TLS 1.3, and MQTT 5/QUIC v1—across six network conditions (stable WiFi, lossy 4G, 5G, 5GWiFi handover, 4G5G handover, NAT rebinding) at three QoS levels and four payload sizes. The model’s transport arithmetic is anchored to published byte-level overheads and the round-trip counts specified by RFCs 9000/9001/9002 and TLS 1.2/1.3, and the calibration is reported alongside the headline results. The TCP+TLS predictions are validated against a Docker-based testbed running the EMQX 5.8 broker with tc/netem impairment; QUIC predictions remain unvalidated and should be interpreted with corresponding caution. With 30 independent runs per cell, bootstrap-percentile 95% confidence intervals on tail metrics, and normal-CIs on means, we find that QUIC’s advantage in the steady state is negligible on low-loss links (indistinguishable from TCP+TLS at p95 on WiFi and 5G) but grows to about 35% at p99 on a lossy 4G link, driven primarily by TCP’s 200 ms minimum RTO floor versus QUIC’s RTT-proportional PTO. On a 5GWiFi handover the QUIC advantage on event-level re-establishment time is small (5–10 ms over TCP+TLS 1.3, not statistically significant) because the L2 outage dominates, but on a 4G5G handover the QUIC advantage grows to about 32 ms because the post-handover RTT is larger and the avoided handshake is proportionally larger. On NAT rebinding QUIC reduces the re-establishment cost to a single path-validation RTT (about 100 ms saved against TCP+TLS 1.3). Topic aliases on 64-byte payloads improve uplink goodput efficiency from 0.40 to 0.44, a 9% gain. We conclude with use-case-specific guidance: QUIC is most beneficial for mobile IoT devices that undergo handovers to higher-RTT bearers or that experience NAT rebinding, while topic aliases dominate the value of MQTT 5 for constrained-payload sensor fleets.

Keywords: IoT, connection migration, multi-RAT handover, NAT rebinding, transport protocol, publish-subscribe, EMQX, analytical model.