[rabbitmq-discuss] 30% performance degradation with user access controls enabled

Matthias Radestock matthias at rabbitmq.com
Sat Sep 22 09:55:42 BST 2012


Chris,

On 21/09/12 23:54, Chris wrote:
> Based on descriptions on RabbitMQ access control,
> http://www.rabbitmq.com/access-control.html : "RabbitMQ may cache the
> results of access control checks on a per-connection or per-channel
> basis." Based on this statement, rabbitmq access control operations
> should not impose too much load on the server performance.
>
> But in our test, we have 5000 client connections over SSL, each client
> is set with a configure/write/read permissions like this:
>
> Configure: {client_name}.*|^amq\.gen.*|^amq\.default|^vsm.*
> Write: {client_name}.*|^amq\.gen.*|^amq\.default
> Read: {client_name}.*|^amq\.gen.*|^amq\.default|^vsm.*
>
> One publisher publishes 10 message per second to an exchange, message
> size is 1KB,  each client/consumer creates a queue and bind to this
> exchange, and we have a total of 5000 unique clients with its own queues.
>
>  From the client side, we can receive about 8000 messages per second.
>
> But if we do not set any user permissions, simply use rabbitmq system
> default guest user, and create 5000 unique connections/queues to the
> server, our client side message rate is 12000 messages per second.
>
>
> My question is: what contributes to the 30% performance degradation?

I suspect there is some difference other than the permissions between 
the two tests. Please send us the output of 'rabbitmqctl report' from 
each test.

Where did you get the rates from? The management UI? If so a screenshot 
of what you are seeing in each case would be good.


Regards,

Matthias.


More information about the rabbitmq-discuss mailing list