These benchmark results illustrate the industry-leading performance of RTI Connext DDS (formerly RTI Data Distribution Service), the lowest latency, highest throughput Java Message Service (JMS) publish/subscribe messaging solution. Connext DDS performance is at least 10x higher than other JMS and enterprise messaging implementations.
These benchmarks were conducted in the following environment:
To benchmark RTI Connext DDS on your hardware, download a free trial now.
The above graphs show the average one-way latency in microseconds for message sizes ranging from 128 to 8,192 bytes. The producer was sending approximately 8,000 messages per second to the consumer, which would echo back every 50th message. This allowed the roundtrip latency to be measured on the sending machine, avoiding clock synchronization issues. The roundtrip latency was divided in half to get the one-way latency that is shown.
The above graph is calculated from the same data as the latency graph. It shows the same average latency (in red) but adds several measures of jitter—the variation in latency from message to message. A system is more deterministic if it exhibits lower jitter.
The red vertical “error bars” show standard deviation and the two blue series show the minimum measured latency and the 99.99% latency (the latency below which 99.99% of the samples fell).
As can be seen, standard deviation is typically just a few microseconds and the range between the minimum, average and 99.99% latency measurements is very narrow. This shows that Connext DDS exhibits very low jitter and very high determinism, making it suitable for time-critical applications.
This graph shows sustainable one-to-one throughput (measured in megabits per second) as a function of message size for message sizes between 32 and 8,192 bytes. It was measured between a single producing and consuming thread, each using a single Gigabit Ethernet port.
Accounting for Ethernet, IP and UDP overhead, the maximum bandwidth available for message data (and metadata) is slightly over 950 megabits. As can be seen, Connext DDS is able to fully utilize all of this available bandwidth when sending messages larger than 512 bytes—meaning throughput is limited by the network and not by the CPU or Connext DDS.
Because RTI uses true peer-to-peer messaging—with no message brokers, servers or daemon processes—there is no inherent limit on aggregate messaging capacity. It is limited only by the network infrastructure.
RTI's one-to-one throughput is more than 10x higher than other JMS and enterprise messaging brokers can do in aggregate. RTI's overall capacity is several orders of magnitude higher.
The above graphs show the same throughput data in terms of the message rate (measured in messages per second). For message sizes larger than 512 bytes, the number of messages per second drops proportionately to message size since the network is saturated.
© Copyright Real-Time Innovations. 2007-2012. All rights reserved.