I am trying to develop an application in which a publisher can occasionally move from one group to another and subscribers will follow, resuming the stream from the old group with minimal delay.
Right now I have significant delay. Even with the publisher waiting for the new stream to fully connect, getting dumped back to pull mode can cause subscribers to experience multiple seconds of waiting for the stream to resume.
To me it seems I will need to "seed" the new stream with redundant data to force the subscribers to develop an optimized push connection before actually making the switch. Michael Thornburgh states in http://forums.adobe.com/thread/565124 that "the multicast system builds and optimizes push trees relatively slowly to keep duplicated stream fragments to a minimum" -- how slowly is this, exactly?
I would like to come up with some programmatic way to make a good estimation as to how much delay, and how much "seed" data, I would have to give to a new stream in order to "bootstrap" most peers into push mode and provide a smooth transition experience. What kind of things trigger the development of the push network?