[ https://jira.jboss.org/jira/browse/JBMESSAGING-180?page=com.atlassian.jir... ] Attila Szegedi commented on JBMESSAGING-180: -------------------------------------------- I don't have issues with starvation - it is really a problem if it causes a priority inversion in the application, which, honestly, is an application level problem and shouldn't be a concern of the messaging middleware. If the app suffers from priority inversion problems, an anti-starvation solution will only "help" it insofar as it'll allow the application to still run, but more slowly, instead of deadlocking (which happens without anti-starvation). I might actually prefer a deadlock (I can detect and analyze those nicely as all the participants are standing still) to a hard to diagnose slowdown. That said, a simple yet effective anti-starvation algorithm would be if you'd choose a random integer between 0 and 9 every now and then (every second?), and then only deliver messages with that priority or higher (as long as there's actual contention, of course). This'll make sure priority 9 messages are always delivered, priority 5 messages are suppressed half the time, and priority 0 messages are suppressed 9/10th of the time. (A different distribution can be achieved by nonlinear mapping of a wider random number range to the priority range). The solution only needs a random number generator local to the server-side processing component, and thus should scale nicely. Anyway, just my 2c. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira