I have a quick question. we’re currently running Openfire 3.9.1 and Spark 2.6.3. Thus far, it’s been a great solution for us (Spark 2.7.0 not withstanding). I’ve had clients request the ability to do voice chat over Spark as in some lab locations we have network drops but no phones. During testing, voice chat does work but its insanely choppy. I’m thinking it is because my server isn’t that powerful. I was wondering what the minimum recomended requirements are to allow voice chat to run over Spark?
i think Spark 2.7.0 beta has voice chat issues reported, and i don’t know the state they are in.
but to answer your question, it depends on a lot of things, server load, server resources, number of concurrent users, how active/chatty users are, your network overhead, etc.
best thing i can recommend is to troubleshoot your bottlenecks.
Start by making sure your openfire is not running out of memory, open your admin ui page --> Server tab --> Server Manager sub-tab --> Java memory. It will give you a green/red bar and a MB usage out of total MB’s and a %. Make sure you are not getting too close to the max usage, otherwise you will have problems. Allocate more memory if needed. Then check your server cpu time, is it being overworked? check your disk i/o, is it too busy? network speed tests to support enough bandwidth? Basically, you will have to look until you discover what your bottlenecks are.
Voice chat in Spark is peer-to-peer and server’s performance is not related. Voice chat is just broken since the migration to FMJ library. This is same for 2.6.3 and current 2.7.0. There is a ticket for fixing it, but i’m not sure when it can be completed. cstux have made a first prototype patch, but since then i haven’t heard anything about it SPARK-1563
Only in 2.5.8 version voice chat should work ok.
it appears he’s been busy:
don’t know the status of his work, but hopefully it’ll be ready to do a pull request soon.
weird problem…if we upgrade the version on spark the latest version should be better than the later version,but on the latest version voice doesn’t work well…
Have you heard about regression? This happens in software. Though in this case it’s a bit different. Spark used proprietary library before. But as Spark is open source now, it is better to use open source libraries. Of course this is not so easy as to just switch libraries and done. Also library is on development too and needs to be upgraded, like smack and other libraries in Spark.
if only i were programmer…i’m lack of coding but the features on spark is better than my previous software like pidgin,many features of spark which i haven’t learnt yet…but in the future i wish i could write their functions on microsoft word.