[CF List] CF 0.95.6 Client discards commands under "load" - feature, not bug

Frank McKenney frank_mckenney at mindspring.com
Wed Jul 5 14:07:49 CDT 2000


Reply to message from Mark Wedel <
     
     mwedel at scruznet.com
     
     > on Fri, 30 Jun
2000 20:16:25 -0700:
>
     
       Times that will consume a lot of bandwidth:
     
     >
     
     
     >
     
      1) entering a new map
     
     >
     
      2) Stepping on a square with lots of objects on it (piles of sold
     
     >
     
      stuff in shops
     
     >
     
      for example, or if you have a narrow hallway and are killing a bunch
     
     >
     
      of monsters
     
     >
     
      on the same space)
     
     >
     
      3) Picking up that large pile of stuff.
     
     
I've seen (2) and (3) after some combat in the Hall of Bones in the
"Ruins of Euthville" (can't believe it took me six _months_ to get the
pun!).  A guy can get killed while the server is still sending millio...
thousan...  hundre...  well, anyway, a _lot_ of "you can't pick
something up because you're already over your limit" messages.  Too bad
these can't be combined into one big "You failed to pick up 6,324 lbs of
assorted junk because you forgot to eat your Wheaties for breakfast"
message.

>
     
      > Would some of the lurkers out there care to offer some examples of
     
     >
     
      > "slow", "acceptable", "really snappy", and "probably unplayably slow"
     
     >
     
      > hardware, specifically with respect to Crossfire 0.95.6?
     
     >
     
     
     >
     
       Note that load on the server will vary this.  A server with 10 active
     
     players
>
     
      certainly has different load than a single player server.
     
     >
     
     
     >
     
       I would say 32 mb of ram is on the low side in modern standards.
     
     >
     
     
     >
     
      One thing you can do use use the 'time (or is it times) command.  This
     
     >
     
      displays > tick information - more importantly, how many times it was
     
     >
     
      not able to complete > all the desired stuff within the 120 ms time
     
     >
     
      constraint.  You will certainly get > some long times, since map 
     
     >
     
      loading will happen in one tick, so that can make for > some very long
     
     >
     
      ticks here and there.
     
     
Okay...  Hm.  Initial logon/movement don't seem as bad as I had epected.
Here's what I saw after I restarted the server, started the (X11)
client, and logged in:

    Total time:
    ticks=256  time=2.489
    avg time=9ms  max time=420ms  min time=-3ms
    ticks longer than max time (120ms) = 2 (0%)
    Time last 100 ticks:
    avg time=11ms  max time=420ms  min time=-3ms
    ticks longer than max time (120ms) = 1 (1%)

Here are the numbers after leaving my apartment:

    Total time:
    ticks=1645  time=11.398
    avg time=6ms  max time=420ms  min time=-3ms
    ticks longer than max time (120ms) = 4 (0%)
    Time last 100 ticks:
    avg time=8ms  max time=278ms  min time=0ms
    ticks longer than max time (120ms) = 1 (1%)

Then, after triggering the "Run On" problem and slamming myself silly
into the outer East Gate (see Bugzilla for details):

    Total time:
    ticks=2551  time=13.979
    avg time=5ms  max time=420ms  min time=-3ms
    ticks longer than max time (120ms) = 6 (0%)
    Time last 100 ticks:
    avg time=8ms  max time=278ms  min time=0ms
    ticks longer than max time (120ms) = 1 (1%)

The numbers _do_ get bigger.  Last time I checked, after 36732 ticks
(221.821) the longest tick was still only 1914 (2 secs).  I may hack the
client to report any ticks longer than (say) 500 ms to the "info" window
just so I can get a feel for when they occur.

>
     
      > That is, I know my own system is a bit slow-ish, but I'm new enough
     
     to
>
     
      > Crossfire that I have no sense of what hardware its servers (and
     
     >
     
      > clients) are running on.  If everyone by me now has a 256Mb 
     
     >
     
      > Athlon/P-III
     
     >
     
      > 600MHz box, my three-hour compilation (I'm slow and pedantic) may
     
     only
>
     
      > be of use to one or two people.
     
     >
     
     
     >
     
       3 hour?  I would really look at adding some more ram.  a 486 dx4-100
     
     >
     
      > shouldn't
     
     >
     
      be that slow.  Compiling on a 100 mhz sparc 10 (hypersparc) only took 
     
     >
     
      >about 8
     
     >
     
      minutes.  On a p3-500, it took about 1 min 45 seconds.
     
     
Um.  Er.  Actually, the 486DX4-100 only takes about 20 minutes to build
the server (configure & makes), which puts it half the throughput of
your Sparc 10.  Now, if only I could write _documentation_ that fast
(e.g.  my "compilation" of information on the command window).

Then there are the two Sun3/60s in the basement I never found the time 
to get running...

>
     
      > Do you (does anyone) have any idea how often users are actually
     
     >
     
      > exceeding (for whatever reason) the default command_window seting of 
     
     >
     
      >10?
     
      
>
     
      Just to be clear - I don't think exceeding it is necessarily a bad 
     
     >
     
      thing.
     
      
>
     
      I do at times exceed that when playing.  Most often beacause I am
     
     >
     
      running in > some direction and enter a building or the like, so there
     
     >
     
      is that brief pause.  > I generally find 10 about the right size -
     
     >
     
      enough so that I can type a few > things in advanced, but not so long
     
     >
     
      that I get stuck against a wall waiting for > all those movements to 
     
     >
     
      get resolved.
     
     
I hear what you're saying, and I need to think about it a bit.  While I
have a current _preference_, that could just be a result of the time
I've spent with 0.93.x where _everything_ got stacked up incredibly.  I
need to really try your "limited window" approach for a bit to give me
some experience for an actual "opinion".



Frank McKenney, McKenney Associates
Richmond, Virginia / (804) 320-4887
E-mail: 
     
     frank_mckenney at mindspring.com
     
     
    


More information about the crossfire mailing list