At 02:52 AM 2/20/2006 -0000, you wrote:
>Sounds like Valve should just instrument the games to provide better data,
>like reporting min, max and mean FPS etc. as well as per-mod data like
>player numbers and anything else that's relevant, on every map change.
>That'd provide much more relevant data than the stress tests.
You might be on to something here. Rather than rely on something
like EtherReal and do a packet capture, perhaps it would be possible in
some future build do something similar minus the actual frame data?
It would seem from the server side all we need is a timestamp for the
packets to and from a specific client, and perhaps a tag in them if
they're of a global refresh type or incremental type, maybe a note that
we expected an update from the client and missed it? Turning this on and
off could then be done via server-side variables so easily done remotely
through rcon:
sv_capture_file "joe_lagging.cap"
sv_capture_ID "player_name" or "steam_id"
sv_capture "1" or "0" to start|stop
The downside is this is going to generate a fair amount of disk I/O, but it
has to be done from the server engine to be at all accurate. Rather than
using server variables I suppose an external app similar to hlds could
be enabled, but again that's one more point of difference between what
the server sees and what the external app sees. The only way I can see
that working is if the external app is a glorified syslog service so the
engine can just stream data to it and the external app is responsible for
allocating buffers and getting that data to disk. Either way, it's going
to take something server-side to start and stop the logging no matter if
it's done to disk or to a syslog service.
Personally, in the future I'd like to see Source done so it
can be installed as a service and thus monitored via perfmon. That's
going to be a major divergence between the Windows and Linux versions
of the engines, but running it under FireDaemon or in a disconnected
screen session is a cludge of the first order anyway. Better to make
it run like a service|daemon and publish some snmp MIB's we can
query for status, but I suspect that's down the road quite a ways.
- Dan
* Dan Sorenson DoD #1066 A.H.M.C. #35 [EMAIL PROTECTED] *
* Vikings? There ain't no vikings here. Just us honest farmers. *
* The town was burning, the villagers were dead. They didn't need *
* those sheep anyway. That's our story and we're sticking to it. *
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds
