
Open source software stack for the development of robotic applications.
Office Address
Plaza de la Encina 10-11, Núcleo 4, 2ª Pl.
28760 Tres cantos - Madrid (España)
eProsima recently undertook a thorough analysis of the main open-source DDS implementations and their performance in terms of latency and throughput.
Fast DDS is the default open-source middleware for the vast majority of ROS 2 distributions over the years, including the long-term supported versions Foxy and Humble. Vulcanexus opted for this specific DDS middleware as the one and only option for communication in the stack. The reason is not only the long list of features, and components that integrate seamlessly with the middleware but also its performance.
Performance is a crucial aspect especially when it comes to demanding systems that rely on real-time characteristics and use safety-critical applications.
Using the Apex.AI performance tests, the eProsima benchmarking shows that Fast DDS is a very stable performer, beating Cyclone DDS and OpenDDS in latency intra-process, and performing similarly in throughput.
The following graphs show only a small sample extracted from the complete report which can be found here.
Latency:

Fig 1.1a Latency: Fast DDS 2.8.0 intraprocess vs Cyclone DDS 0.9.0b1 intraprocess & Open DDS 3.13.2 intraprocess

Fig 1.2a Latency: Fast DDS 2.8.0 interprocess vs Cyclone DDS 0.9.0b1 interprocess & Open DDS 3.13.2 interprocess
Throughput:

Fig 2.1a Throughput: Fast DDS 2.8.0 intraprocess vs Cyclone DDS 0.9.0b1 intraprocess & Open DDS 3.13.2 intraprocess

Fig 2.2a Throughput: Fast DDS 2.8.0 interprocess vs Cyclone DDS 0.9.0b1 interprocess & Open DDS 3.13.2 interprocess
Furthermore, Fast DDS was also tested on its own using various configuration options like shared memory (zero copy), loans and UDP. The results are equally good showing a stable performance in all cases with high throughput and low latency. In this case, it can be observed how Fast DDS has a much higher throughput than in the previous graphs, as these tests allow to push the throughput limits further than the Apex.AI tests.
The following graphs are also only a small sample of the complete report which can be found here.


Please contact us.

Open source software stack for the development of robotic applications.
Plaza de la Encina 10-11, Núcleo 4, 2ª Pl.
28760 Tres cantos - Madrid (España)