Java Broker Design - Flow to Disk commented on by Aidan Skinner (Feb 02, 2009).

Comment:

Copy-per-queue is incredibly expensive when dealing with topics, and changes our memory usage there from O(<QueueEntry>*subscribers) to O(<size of message>*subscribers). It also involves a lot (N = subscribers) of copies in memory, which aren't nearly as cheap as you'd hope.

We don't need to actually implement copy-per-queue to use that as a simplifying assumption when calculating queue usage, and I totally agree that trying to acurately gauge queue memory usage within the overall context of the broker is futile.

--------------------------------------------------------------------- Apache Qpid - AMQP Messaging Implementation Project: http://qpid.apache.org Use/Interact: mailto:commits-subscr...@qpid.apache.org

Reply via email to