Zenoh vs DDS: A Technical Comparison for Robotics Middleware

Wire efficiency: Zenoh's protocol overhead is as low as 5 bytes, compared to DDS's RTPS header of 36+ bytes. This drastically reduces bandwidth usage in large-scale, distributed systems. Peer-to-peer discovery: Supports decentralized, multicast-free discovery, making it ideal for dynamic environments like autonomous vehicle fleets or swarm robotics where nodes frequently join and leave. Cloud-to-edge bridging: Natively bridges from embedded microcontrollers to cloud data centers through a single protocol, eliminating the need for complex gateways or protocol translators.

Network agnostic: Operates seamlessly over TCP, UDP, QUIC, Bluetooth, and serial, allowing a single middleware stack from a microcontroller to a cloud server. Resource constrained: The micro-zenoh client runs on bare-metal and RTOS devices with minimal memory footprint, something traditional DDS implementations struggle with. Use case fit: Choose Zenoh for greenfield projects requiring cloud-to-microcontroller connectivity, multi-modal transport, and bandwidth-constrained links like satellite or underwater communication.

Proven determinism: DDS offers a rich set of 22 Quality of Service (QoS) policies, including deadline, latency budget, and ownership strength, providing fine-grained control for hard real-time systems. Safety certifications: Implementations like RTI Connext Drive are certified to ISO 26262 ASIL-D, making DDS the default choice for safety-critical automotive and avionics systems. Ecosystem maturity: With 20+ years of deployment in naval combat systems, medical devices, and industrial control, DDS provides battle-tested reliability and extensive tooling for system introspection.

OMG standard: Governed by the Object Management Group, DDS ensures vendor interoperability through the standard RTPS wire protocol, allowing you to mix implementations like RTI Connext and Eclipse Cyclone DDS. Data-centric model: Uses a global data space abstraction where applications publish and subscribe to topics, decoupling applications in time and space. This is powerful for complex, modular systems like large-scale industrial automation. Use case fit: Choose DDS for brownfield, safety-critical systems requiring hard real-time guarantees, formal certification, and integration with existing ROS 2 or industrial control ecosystems.

Specific advantage: Zenoh achieves a wire overhead of only ~5 bytes per data message, compared to DDS implementations which can exceed 30+ bytes due to RTPS protocol headers. This matters for bandwidth-constrained teleoperation and large-scale sensor data streaming where every byte counts, such as transmitting LiDAR point clouds from a robot to a remote control station over a 4G/5G link.

Specific advantage: Zenoh's protocol is designed to run uniformly from microcontrollers to cloud data centers without requiring a bridge or gateway translation layer. A single Zenoh router can transparently bridge MQTT, DDS, and Zenoh-native traffic across LAN, WAN, and cloud networks. This matters for multi-robot fleets that need to communicate with centralized cloud analytics without deploying separate middleware stacks at each tier.

Trade-off: While Zenoh has a growing community, its ecosystem of pre-built tools, vendor-supported libraries, and commercial support contracts is significantly smaller than the DDS ecosystem, which is backed by over 15 years of adoption in defense, aerospace, and industrial IoT. This matters for risk-averse teams in regulated industries that require long-term support guarantees, certified safety profiles, and a large pool of experienced developers.

Trade-off: DDS implementations like RTI Connext are deeply integrated with real-time operating systems (e.g., QNX, VxWorks) and offer deterministic, priority-based scheduling for safety-critical control loops. Zenoh, while lightweight, does not yet offer the same level of certified real-time determinism out-of-the-box. This matters for hard real-time motion control and safety-rated systems where microsecond-level jitter is unacceptable.