Scalability of Temporary Destinations

Question

          Hi,
          I read a few posts in this newsgroup on temporary queues, and found that although
          the creation of those queues is not very performance-friendly it is probably a
          better idea than have multiple selectors on a queue. (topics are not an option
          in our case)
          However, I was wondering about scalability. Is their a limit to the amount of
          Temp. Destinations that can exist. And if so, what happens when the creation of
          an extra destination is attempted.
          And, what quantity of temp destinations really starts to hurt performance?
          Thanks,
          Erik
          

Answer 1

Hi Eric,
          There are no theoretical limits, but there are practical ones.
          Performance depends on hardware, how frequently the temp queues get
          created and removed, available memory, message rates, how clustering is
          used, what the application is doing outside of JMS, etc. All I can say
          is that I have seen applications with thousands of destinations, and
          tens of thousands of subscribers per server. As usual in these cases,
          I recommend writing a benchmark that models your application and running
          the benchmark on hardware similar to that which will be used in production.
          If you could use topic subscribers instead you might wish to revisit
          using that approach. WebLogic JMS has built-in optimizations for a
          certain use cases. The optimization changes CPU burn for topic selector
          evaulation from O(n), where n is the number of subscribers, to O(1).
          Tom
          Erik wrote:
          > Hi,
          >
          > I read a few posts in this newsgroup on temporary queues, and found that although
          > the creation of those queues is not very performance-friendly it is probably a
          > better idea than have multiple selectors on a queue. (topics are not an option
          > in our case)
          >
          > However, I was wondering about scalability. Is their a limit to the amount of
          > Temp. Destinations that can exist. And if so, what happens when the creation of
          > an extra destination is attempted.
          >
          > And, what quantity of temp destinations really starts to hurt performance?
          >
          > Thanks,
          > Erik