Can JMSPriority be used to distinguish between real-time, near real-time workloads?

Question

IHAC that needs to throttle work passed to a backend system, i.e. limit
          the # of concurrent requests, as well as prioritize between different
          kinds of workloads including real-time, deferred real-time
          (asynchronous) and batch.
          I'm proposing to use a pool of MDBs reading of a single JMS queue. The
          max-free-beans-in-pool would be used to limit the concurrent access to
          the backend system.
          Can JMSPriority be used to distinguish between the different kinds of
          workloads? (real-time, near real-time and batch)
          How is the JMSPriority used when consuming messages from a queue in WLS
          JMS? Are messages dequeued in strict priority order or is there an
          algorithm to ensure lower priority messages are periodically consumed,
          i.e. to ensure they are not starved-out? Or do you have to create a
          message selector to dequeue based on priority?
          Is the an appropriate use of JMSPriority or should I set-up dedicated
          JMS queues for each of the different workload types?
          Cheers
          Simon
          

Answer 1

          Tom,
          How do you create custome connection factories in JMS? And also how can u
          change the MessageMaximum to 0? I thought the minimum is 1 and that u can only
          change in the config.xml file??
          Any code snippets on how to write custom connection factories would be helpful.
          Thank You,
          Sankas
          Tom Barnes <[email protected]> wrote:
          >Hi Simon,
          >
          >Answers in-line.
          >
          >Simon Pearce wrote:
          >
          >> IHAC that needs to throttle work passed to a backend system, i.e. limit
          >> the # of concurrent requests, as well as prioritize between different
          >> kinds of workloads including real-time, deferred real-time
          >> (asynchronous) and batch.
          >>
          >> I'm proposing to use a pool of MDBs reading of a single JMS queue.
          >The
          >> max-free-beans-in-pool would be used to limit the concurrent access
          >to
          >> the backend system.
          >>
          >> Can JMSPriority be used to distinguish between the different kinds
          >of
          >> workloads? (real-time, near real-time and batch)
          >>
          >> How is the JMSPriority used when consuming messages from a queue in
          >WLS
          >> JMS? Are messages dequeued in strict priority order or is there an
          >> algorithm to ensure lower priority messages are periodically consumed,
          >> i.e. to ensure they are not starved-out?
          >
          >To get priority sorting you must configure the destination to sort by
          >priority.
          >It sorts FIFO by default.
          >
          >In addition, you may want to tune-down the pipe-line size between
          >the JMS server and the MDB (create a custom connection factory, and
          >set MessagesMaximum to 0, default is 10), as the pipe-line is unsorted
          >and consequently messages will not "jump ahead" of each-other in the
          >pipe-line.
          >
          >> Or do you have to create a
          >> message selector to dequeue based on priority?
          >
          >No.
          >
          >>
          >>
          >> Is the an appropriate use of JMSPriority or should I set-up dedicated
          >> JMS queues for each of the different workload types?
          >
          >The latter, as long the app can tolerate concurrent execution of all
          >workload
          >types due to having an MDB per queue. The latter will perform better
          >than selectors,
          >and you won't have to worry about the unsorted pipe-line.
          >
          >>
          >>
          >>
          >> Cheers
          >>
          >> Simon
          >
          >Tom, BEA
          >