[CF-Devel] Status update

[Mark Wedel]

Andreas Vogl wrote:

>
     
      > Not much has changed recently with the socket code that I am aware of.
     
     >
     
     
     >
     
      New cmds have been created for PR, metaserver stuff has been added.. etc.
     
     >
     
      There have well been changes to the socket code.
     
     
 Actually, the PR code basically just sends more data with the existing stats
command.

 metaserver is sort of seperate from normal socket code.  Also, there is no
reason to expect metaserver updates to vary at all on load - if there were
periodic pauses for no good reasons, I might look closer at the metaserver code
to see if its blocking.

 IF this is a strong suspicion, it is easy enough to turn off the metaserver
update code.

 The problem is that I'm not positive the change really happened since 0.95.7. 
As said in another message, I went into the dragon hangar and was there for 5-10
minutes before performance started getting bad.  This suggest to me that there
could be some trigger event which then causes this performance problem, and
perhaps some servers are jsut more immune to that event than others.


>
     
      I am absolutely sure that the CPU is never maxed out. I get this "disrupted
     
     >
     
      character movement" when I'm playing alone on one server with formidable
     
     >
     
      connection speed. When there are ten other players online, behaviour
     
     >
     
      is *exactly* the same.
     
     
 Number of players is sort of an odd metric.  Going into the dragon hangar
probably represents a load by that one player comparable to 10-15 other players
doing mundane things.  And multiple characters in the dragon hangar wouldn't
make much different, since it is the same set of monsters.

 10 players on 10 different dragon hangar style maps may be interesting.

 Note that with a bug in the server, one player could max out cpu time in the
server if it starts speding too much time looping over something.  The cast a
bunch of fireballs at a wall illustrates that point.

>
     
     
     >
     
      No doubt, the reason must be #3. Don't forget that this problem did not
     
     >
     
      exist
     
     >
     
      on servers of older CVS-versions. (There was lag when I entered a room
     
     >
     
      full of spellcasters, but there was definitly no diruption when I
     
     >
     
      entered a room full of trolls!)
     
     
 Note that it would not be hard to track the amount of data the client is
receiving.  I'm actually going to make a small change so that the client will
also print out the size of the packets it receives - you can then fire up the
client and see how much data the client really is getting.