PingTimeout and ValidationFrequency

[Hunternif]

Hello.

My question essentially is: does setting these two parameters in Application.xml lower than default impose significant load on WMS? What minimum would you suggest?

In my video conference, when someone gets disconnected from WMS (due to network issues), I want to inform subscribers that he is temporarily offline (but still listed in the conference - in hopes of connecting back soon). But this can only be done after my WMS module's 'onDisconnect' call, which happens after a ping timeout. The subscribers are seeing an empty NetStream during the timeout.
The resulting pause equals <PingTimeout>+<ValidationFrequency> [milliseconds]
Hence the question. What if, for instance, I set both PingTimeout and ValidationFrequency to 1000?

Thank you in advance.

[rrlanham]

The affect of shortening these settings on overhead will increase with server-load. At the high end I think it would be significant. Of course you would have to test and measure to see actual difference. If all clients are Flash RTMP, and not RTMPT, RTMPS, etc, then I think you have more flexibility, i.e., you can make these values shorter. If some clients are RTMPT, etc, making these values too short might disconnect clients that shouldn't be.

Richard

[Hunternif]

Thank you, Richard.

Could you please elaborate a bit on the protocols? Does it take longer for RTMPE to respond to pings from WMS than RTMP?

[rrlanham]

No, there should be no difference like that, in fact I there might not be any difference in this regard. RTMPT can be more fragile on the client-side, more easily affected by network difficulties that would not affect RTMP. I'm not sure if that affects the server-side PingTimeout mechanism.

Richard

[rrlanham]

here is other info:

PingTimeout and ValidationFrequency only apply to RTMP connections not RTMPT connections. Lowering these values will through the onDisconnect event earlier when there are network problems. Lowering these values could lead to more false disconnects

Sorry for the confusion.

Richard