User-agent: Microsoft Outlook Express Macintosh Edition - 5.01 (1630)
The number of "RTP connections" literally means how many RTSP Session IDs
have been allocated by the server. "RTSP connections" is the number of TCP
connections currently active on the server. Because RTSP TCP connections do
not have to be maintained through the course of a session (a client could
theoretically do the DESCRIBE, SETUPs, and PLAY on separate TCP
connections), these counts must be separate. BUT, with SpamPro as a client,
you will see 1 RTP connection (because audio and video streams are SETUP
using a single RTSP session ID), and 1 RTSP connection per client.
If you are in the process of connecting and disconnecting clients, these
numbers may not be the same. After all clients have connected, these numbers
should be IDENTICAL (save 1 RTSP connection SpamPro maintains for control
purposes) In your test, you should be sure that clients are not in the
process of connecting or disconnecting. Otherwise, there is a slim chance
that the server is disconnecting your RTP connections for some reason. Could
it be that the maximum connections value or the maximum bandwidth value in
the server configuration is being exceeded? In that case, the server will
close the "RTP connection" for that client immediately, which will stop
streaming and make the RTSP session ID invalid. But, the TCP connection may
take awhile to time out. That may explain the oddity of your test. Is that
the case?
Btw - "HTTP connections" is the number of HTTP tunneled TCP connections
currently active on the server. If you are using SpamPro in HTTP tunneling
mode, instead of an RTSP connection, you will see 1 HTTP connection for each
client.
Hope that helps.
on 11/10/00 12:59 PM, Zimmerman, Don at email@hidden wrote:
> This is why Larry expects the "RTP connections" to be double the count of
> RTSP connections. If you expect the RTP connections to be "not that far
> apart" from the RTSP count, then exactly what does the RTP connections count
> mean? How would analyze the problem of descrepancy in the counts? (With
> some clues we can investigate this better).
--
Denis Serenyi
QuickTime Streaming Server Engineering
email@hidden
