Hey, Sorry that I had to miss the meeting. I was tied up with work for a client at that time. I thought I might be able to get out of it, but alas, I had to miss the meeting.
Anyway, I do appreciate the assistance that the Terracotta guys are doing for this effort. Thanks also to Steven Gong for being there to represent the Red5 development team! I look forward to seeing what you guys come up with. -Chris On 6/15/07, Steven Gong <[EMAIL PROTECTED]> wrote: > Orion, > Glad to receive your mail. It's a great pleasure to meet you all from > Terracotta team via skype and it's a fruitful meeting indeed. :-) > > I am happy to see that my original idea to make the global scope as a shared > root is in a right direction. Tim's idea to make the children of global > scope as root is ingenious and saves a lot of trouble to cluster spring. > > I believe it's a right path to follow and I will continue with the TC > configuration and refactoring. Keep in touch! :-) > > > On 6/15/07, Orion Letizi <[EMAIL PROTECTED]> wrote: > > > > We (Steven Gong, Steve Harris, Tim Eck, Taylor Gautier, and Orion > > Letizi) had a good discussion this morning (California time) about how > > to approach clustering Red5 with Terracotta. > > > > The upshot is it seems like the best thing to start with is to cluster > > the relevant tree of Scope objects in the origin servers and to fire a > > Terracotta "distributed method call" whenever a shared object in a Scope > > object is updated. That way, all the shared objects in the Scope tree > > will be available (via Terracotta) to all origin servers on demand and > > all origin servers will be notified when the state of a relevant scope > > object changes. > > > > In order to do that, we'll need to figure out: > > > > * what to use as a Terracotta clustered root object. A likely candidate > > is the children of the org.red5.server.GlobalScope object. There's a > > small wrinkle in the way roots work in Terracotta that would require the > > children of the GlobalScope object to be kept in a separate field. > > Right now, GlobalScope extends org.red5.server.Scope and inherits the > > Scope.children map. We discussed adding a separate field to GlobalScope > > that we can use as the Terracotta root instead of the inherited > > Scope.children field. > > > > * what to make transient in the scope tree. There is some JVM-specific > > stuff in the scope tree (e.g., connections, streams, listeners, etc.) > > that should not be clustered by Terracotta. These things need to be > > declared transient so that they don't become clustered. > > > > * iron out the details of the Spring GlobalScope initialization. This > > seemed like a fairly minor issue, but there may be some stuff we have to > > figure out there. > > > > All in all, this seems very doable, although it will take some tinkering > > and maybe some minor refactoring in the Red5 server code to get it all > > to work properly. > > > > Thanks very much to Steven Gong for taking the time to meet with us. > > This is really exciting stuff. > > > > Cheers, > > Orion > > > > _______________________________________________ > > tc-dev mailing list > > [email protected] > > http://lists.terracotta.org/mailman/listinfo/tc-dev > > > > > > -- > I cannot tell why this heart languishes in silence. It is for small needs it > never asks, or knows or remembers. -- Tagore > > Best Regards > Steven Gong > _______________________________________________ > Red5devs mailing list > [EMAIL PROTECTED] > http://osflash.org/mailman/listinfo/red5devs_osflash.org > > _______________________________________________ tc-dev mailing list [email protected] http://lists.terracotta.org/mailman/listinfo/tc-dev
