[rabbitmq-discuss] the cost of creating a queue?

Alex Shneyderman a.shneyderman at gmail.com
Wed Oct 17 12:12:17 BST 2012


Thanks, guys!

Cheers,
Alex.

On Wed, Oct 17, 2012 at 12:20 PM, Simon MacMullen <simon at rabbitmq.com> wrote:
> On 17/10/12 11:16, Francesco Mazzoli wrote:
>>
>> 100-200 queues won't be a problem, at all.  I cannot give you exact
>> numbers, but
>> creating a queue basically amounts to creating a record in the mnesia
>> database,
>> and the record is quite tiny:
>>
>> <http://hg.rabbitmq.com/rabbitmq-server/file/c6fbb723f718/include/rabbit.hrl#l49>.
>
>
> The memory cost of the process is higher, although this is not huge either.
> To the OP: a few hundred queues is no big deal at all. I would start to
> worry when we're talking hundreds of thousands :-)
>
>
>> There will be an overhead when creating a queue in a clustered rabbit
>> (since the
>> mnesia transaction has to be synchronised across the nodes) but there
>> won't be
>> much creating an HA queue (the slaves are contacted asynchronously when
>> creating
>> the queue).
>
>
> This sounds like HA queues are cheaper to create than non-HA ones. Which is
> not true. To the OP: high rates of queue creation / deletion might be a
> problem, but it sounds like you are not planning on doing that.
>
> Cheers, Simon
>
> --
> Simon MacMullen
> RabbitMQ, VMware


More information about the rabbitmq-discuss mailing list