[論文レビュー] Zigbee vs. Matter over Thread: Understanding IoT Protocol Performance in Practice

Massimo Nobile, Fabio Palmese+3 more|arXiv (Cornell University)|Mar 4, 2026

IoT Networks and Protocols被引用数 0

ひとこと要約

論文は同一のハードウェアテストベッド上で Zigbee と Matter over Thread を経験的に比較し、 Zigbee の局所ルーティングは速いがスケーラビリティが劣る一方、Matter over Thread はマルチホップ展開においてより良いスケーラビリティとロバスト性を提供することを明らかにします。

ABSTRACT

The widespread adoption of the Internet of Things (IoT) has positioned smart homes as paradigmatic examples of distributed automation systems, where reliability, efficiency, and interoperability depend critically on the underlying communication protocol. Among the low-power wireless technologies available for this scenario, Zigbee and Matter over Thread have emerged as leading contenders. While Zigbee represents a mature, non-IP mesh networking solution, Matter over Thread introduces an IP-based architecture designed to unify device interoperability across different ecosystems. However, despite extensive documentation of their design principles, there is a lack of empirical, comparative performance data under realistic network conditions. This paper presents a comprehensive experimental comparison between the two protocols, conducted on a testbed built from commercially available hardware. The proposed methodology focuses on different key performance dimensions, such as scalability, responsiveness, and fault tolerance. The results reveal that Zigbee achieves a lower baseline overhead and faster route recovery, making it more responsive in static small-scale deployments. Matter over Thread, conversely, exhibits superior scalability and robustness, maintaining stable throughput and predictable latency across multi-hop scenarios. Overall, we highlight that Zigbee and Matter over Thread embody distinct trade-offs between agility, efficiency, and scalability.

研究の動機と目的

• Assess scalability and stability of Zigbee vs. Matter over Thread in realistic smart-home topologies.
• Evaluate responsiveness and efficiency for user-facing operations and OTA-style tasks.
• Quantify fault-tolerance and route recovery performance under node failures.
• Provide data-driven insights to guide protocol selection for large-scale deployments.

提案手法

• Construct a controlled hardware testbed with identical ESP32-C6 devices for both protocols.
• Configure Zigbee as a router and Matter over Thread as a full Thread device on the same hardware.
• Use two topologies (fully connected mesh and chain) to test scalability and routing behavior.
• Separate protocol overhead from application traffic by classifying traffic in Wireshark.

実験結果

リサーチクエスチョン

• RQ1 Which protocol maintains greater stability and responsiveness as the number of devices grows and network depth increases?
• RQ2 Which protocol provides superior responsiveness under user-facing operations and handles heavy tasks efficiently?
• RQ3 Which protocol offers better fault tolerance and faster route recovery after node failures?

主な発見

• Zigbee has lower baseline overhead in small networks but overhead rises sharply with size and can exceed Thread in six-node meshes.
• Matter over Thread shows more stable, linear-like growth in traffic with increased hops, indicating better scalability.
• In chain topologies, Zigbee experiences severe congestion and reliability issues beyond three hops, while Thread remains operational across depths tested.
• Thread achieves higher and more predictable end-to-end latency that scales with hop count, whereas Zigbee latency increases non-linearly and degrades with more hops.
• Zigbee delivers higher single-hop throughput (~75 kbps) but loses performance rapidly in multi-hop settings; Thread maintains higher throughput and supports TCP/UDP, with TCP offering stable, repeatable performance.
• Fragmentation is worse for Zigbee (85–95+ bytes range depending on conditions) than for Thread due to 6LoWPAN efficiency, contributing to higher latency in Zigbee under fragmentation.
• Thread with TCP provides stable throughput without per-hop tuning, while Zigbee requires careful pacing to avoid congestion.