Open flaviojs opened 10 years ago
I understand you fixed (well patched, it still has issues with pps) the displayed issue by bumping up the effective "tx ring" size to 16 instead of 1 in the specific driver affected, but this model of operation itself will present pathology in a few network designs and features (low protocol timers, bfd, ntp, jitter simulation (this basically creates jitter by design), etc). Would it be possible to make this more realtime as opposed to the current 10ms cycle sleep based queue processing?
My intention is to make the emulated behavior closer to the real thing. The fact that it's sharing the single ptask thread is enough to break that. =~~ So far I know some chips send on command, while others do memory polling and others depend on the configuration. I want to read the datasheet of each relevant hardware chip and emulate what they do.
Packet transmission is limited because:
Links:
http://forum.gns3.net/topic10476.html