5 Replies Latest reply on Aug 9, 2007 5:24 PM by mike soto

    BisocketServerInvoker and random port issue

    mike soto Newbie

      Can anyone please clarify if there is any way to configure the BisocketServerInvoker to not use a random port when starting the secondary server? If this is true, am I correct in assuming that it is impossible to operate an application (jboss-4.2.0.GA + jboss-messaging-1.3.0.GA) that utilizes it behind a firewall unless all ports are open? If this is true, was there a reason for it?

      Is the thought here to use the multiplex transport instead of the bisocket? Would there be any performance impacts by using the different transport? I am assuming bisocket would give the best performance, is that correct?

      thanks.
      Ray.

        • 1. Re: BisocketServerInvoker and random port issue
          Ron Sigal Master

           

          Can anyone please clarify if there is any way to configure the BisocketServerInvoker to not use a random port when starting the secondary server?


          There is a JIRA issue for this feature: JBREM-749 "BisocketServerInvoker: Make configurable the address and port of secondary server socket". It is scheduled for the next Remoting release.

          If this is true, am I correct in assuming that it is impossible to operate an application (jboss-4.2.0.GA + jboss-messaging-1.3.0.GA) that utilizes it behind a firewall unless all ports are open? If this is true, was there a reason for it?


          Until the fix for JBREM-749 is available, you're correct. The motivation for the bisocket transport was to support *clients* behind a firewall.

          Is the thought here to use the multiplex transport instead of the bisocket? Would there be any performance impacts by using the different transport? I am assuming bisocket would give the best performance, is that correct?


          I'm recommending, whenever possible, bisocket over multiplex. It's faster (we don't have any statistics yet, but I'd guess at least 50% faster). It's also, in my opinion, more reliable. It's used by JBossMessaging, which means it's being used in production in many places. It's relatively new, and we're still shaking out some wrinkles, but I'm not aware of any comparable use of multiplex.

          • 2. Re: BisocketServerInvoker and random port issue
            Ron Sigal Master

             

            Can anyone please clarify if there is any way to configure the BisocketServerInvoker to not use a random port when starting the secondary server?


            There is a JIRA issue for this feature: JBREM-749 "BisocketServerInvoker: Make configurable the address and port of secondary server socket". It is scheduled for the next Remoting release.

            If this is true, am I correct in assuming that it is impossible to operate an application (jboss-4.2.0.GA + jboss-messaging-1.3.0.GA) that utilizes it behind a firewall unless all ports are open? If this is true, was there a reason for it?


            Until the fix for JBREM-749 is available, you're correct. The motivation for the bisocket transport was to support *clients* behind a firewall.

            Is the thought here to use the multiplex transport instead of the bisocket? Would there be any performance impacts by using the different transport? I am assuming bisocket would give the best performance, is that correct?


            I'm recommending, whenever possible, bisocket over multiplex. It's faster (we don't have any statistics yet, but I'd guess at least 50% faster). It's also, in my opinion, more reliable. It's used by JBossMessaging, which means it's being used in production in many places. It's relatively new, and we're still shaking out some wrinkles, but I'm not aware of any comparable use of multiplex.

            • 3. Re: BisocketServerInvoker and random port issue
              Ron Sigal Master

               

              Can anyone please clarify if there is any way to configure the BisocketServerInvoker to not use a random port when starting the secondary server?


              There is a JIRA issue for this feature: JBREM-749 "BisocketServerInvoker: Make configurable the address and port of secondary server socket". It is scheduled for the next Remoting release.

              If this is true, am I correct in assuming that it is impossible to operate an application (jboss-4.2.0.GA + jboss-messaging-1.3.0.GA) that utilizes it behind a firewall unless all ports are open? If this is true, was there a reason for it?


              Until the fix for JBREM-749 is available, you're correct. The motivation for the bisocket transport was to support *clients* behind a firewall.

              Is the thought here to use the multiplex transport instead of the bisocket? Would there be any performance impacts by using the different transport? I am assuming bisocket would give the best performance, is that correct?


              I'm recommending, whenever possible, bisocket over multiplex. It's faster (we don't have any statistics yet, but I'd guess at least 50% faster). It's also, in my opinion, more reliable. It's used by JBossMessaging, which means it's being used in production in many places. It's relatively new, and we're still shaking out some wrinkles, but I'm not aware of any comparable use of multiplex.

              • 4. Re: BisocketServerInvoker and random port issue
                Ron Sigal Master

                How's that for service? You get three answers for the price of one. :-)

                • 5. Re: BisocketServerInvoker and random port issue
                  mike soto Newbie

                  Thanks very much for the repsone. much appreciated.