Today we are happy to announce better performance and throughput on new instances created in Azure! Our benchmark test showed around 100% better throughput than before! We have changed the underlying server types for CloudAMQP servers created in Azure, and at the same time we moved from Azure Classic to Azure ARM.
Your RabbitMQ instance will be located on a Standard F8 server if you create the single node Awesome Ape plan in Azure. You will end up on Standard D15 v2 with 20 cores and 140GB in RAM if you create a new single node Loud Lion plan.
We could easily get more then 600.000 messages per second in throughput on a single node Loud Lion instance!
Benchmark testing of Awesome Ape (single node)
The RabbitMQ PerfTest tool, which is part of the RabbitMQ Java client has been used for benchmark testing.
We created one single node Awesome Ape on Azure Classic and one single node Awesome Ape on Azure ARM. We used different flags, different number of queues, consumers, producers and different exchanges etc. Almost every combination gave us around 100% better throughput. Below follow two examples.
Azure Classic
$ runjava.sh com.rabbitmq.examples.PerfTest --uri $uri --consumers 12 --producers 4 --autoack
time: 52.004s, sent: 6999 msg/s, received: 81630 msg/s
time: 53.004s, sent: 8272 msg/s, received: 86533 msg/s
time: 54.004s, sent: 9544 msg/s, received: 86482 msg/s
time: 55.004s, sent: 5089 msg/s, received: 85393 msg/sAzure ARM
$ runjava.sh com.rabbitmq.examples.PerfTest --uri $uri --consumers 12 --producers 4 --autoack
time: 52.005s, sent: 18178 msg/s, received: 207524 msg/s
time: 53.010s, sent: 15192 msg/s, received: 202283 msg/s
time: 54.010s, sent: 24179 msg/s, received: 223324 msg/s
time: 55.010s, sent: 15269 msg/s, received: 234518 msg/sGive it a try and do some benchmark testing! As usual, all plans are billed by the second so you can try it out and delete it when you have finished benchmarking. Please email us at contact@cloudamqp.com if you have any suggestions or feedback.
