Quorum queues vs classic mirrored queues resource usage

[Vilius Šumskas]

Hi,

 

we have started to test our testing RabbitMQ cluster with quorum queues instead of classic mirrored queues. The cluster contains 3 nodes.

 

For the same amount of queues (~2000 of them) under idle load we observe 2x of RAM and 2x of CPU requirement increase. Just wondering if this is expected behaviour?

 

--

    Vilius

 

[Michal Kuratczyk]

Hi,

Yes, quorum queues tend to use more under steady workload, but they offer higher throughput, lower latency and should behave better

(more predictably) in extreme conditions. Also, they actually offer data safety. :)

We've just updated this section as there was some outdated info, but it did mention higher resource usage before as well:

https://www.rabbitmq.com/quorum-queues.html#resource-use

Best,

--
You received this message because you are subscribed to the Google Groups "rabbitmq-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rabbitmq-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/rabbitmq-users/AM8PR01MB7699682EDC8379707C7A65FF92DB9%40AM8PR01MB7699.eurprd01.prod.exchangelabs.com.

--

Michał

RabbitMQ team

[Vilius Šumskas]

Hi,

 

thank you for your answer. I just rechecked our monitoring and it seems like I have looked at it wrong the first time.

CPU load is comparable, however RAM usage is still 50-60% more when compared to classic mirrored queues.

 

By the way this is not under steady load, but under completely idle system with one or two very small (heartbeat) messages comming every few minutes.

 

--

    Vilius

To view this discussion on the web, visit https://groups.google.com/d/msgid/rabbitmq-users/CAA81d0s8v53ovakiQrmEp9Rffa73f5tuXawVkgYM6t_LmCvBMQ%40mail.gmail.com.