Not that this will help much but Gary's post had me wanting to do a little looking at stuff myself.
I threw up a quick server.. box is an AthlonXP 2800, 2gb DDR400 Running FC4 cat /proc/version = Linux version 2.6.19-rc1KC2 ([EMAIL PROTECTED]) (gcc version 4.0.2 20051125 (Red Hat 4.0.2-8)) #2 Sun Oct 8 01:35:36 PDT 2006 The KC2 at the end of the kernel is my own stuff for testing tracking. The server I ran was a tick 100 css lan server with -nomaster set map was dust2. I connected with my client.. strace output.. Process 2730 detached % time seconds usecs/call calls errors syscall ------ ----------- ----------- --------- --------- ---------------- 98.72 0.001999 2 827 nanosleep 1.28 0.000026 0 169 sendto 0.00 0.000000 0 1 restart_syscall 0.00 0.000000 0 2 read 0.00 0.000000 0 2 open 0.00 0.000000 0 2 close 0.00 0.000000 0 2 time 0.00 0.000000 0 3514 gettimeofday 0.00 0.000000 0 2 munmap 0.00 0.000000 0 2 sysinfo 0.00 0.000000 0 828 select 0.00 0.000000 0 2 mmap2 0.00 0.000000 0 2 fstat64 0.00 0.000000 0 171 171 accept 0.00 0.000000 0 206 171 recvfrom ------ ----------- ----------- --------- --------- ---------------- 100.00 0.002025 5732 342 total Stats reading: stats CPU In Out Uptime Users FPS Players 3.67 1539.02 6238.08 25 0 494.32 1 Restarted the server as an internet server, same settings/cfg/client connected.. stats CPU In Out Uptime Users FPS Players 3.50 1538.82 6230.13 3 0 494.07 1 Process 2800 detached % time seconds usecs/call calls errors syscall ------ ----------- ----------- --------- --------- ---------------- 99.90 0.040984 8 5261 nanosleep 0.10 0.000040 0 1172 sendto 0.00 0.000000 0 1 restart_syscall 0.00 0.000000 0 12 read 0.00 0.000000 0 12 open 0.00 0.000000 0 12 close 0.00 0.000000 0 12 time 0.00 0.000000 0 22916 gettimeofday 0.00 0.000000 0 12 munmap 0.00 0.000000 0 12 sysinfo 0.00 0.000000 0 5262 select 0.00 0.000000 0 12 mmap2 0.00 0.000000 0 12 fstat64 0.00 0.000000 0 1184 1184 accept 0.00 0.000000 0 1416 1184 recvfrom ------ ----------- ----------- --------- --------- ---------------- 100.00 0.041024 37308 2368 total Are there any other things to check or look at that you guys would want too see. I know that you guys probably want info from a loaded server that has been up for a day+... I am not much help having my dual xeon box sitting here on a table and not in a data center :P I can fire it up as well, would need to install an OS on it to do some baseline checks for you.... --kennycom ----- Original Message ----- From: "Gary" <[EMAIL PROTECTED]> To: <[email protected]> Sent: Sunday, November 05, 2006 5:53 PM Subject: RE: [hlds_linux] srcds performance > At 01:42 PM 11/5/2006, - 4ster!x - wrote: >>Very sorry to add some experience data without any graphs, but I haven't >>had >>time to set em up yet. >>I thought my experiences might be of some interest anyway : >> >>RootServer with the following configuration : >> >>AMD Athlon64 X2 4800+ "Windsor" (as such, it comes with 2048kByte 2nd lev. >>Cache) >>2048 MB Kingston DDR2-667 Ram >> >>Following are the results of the version checks : >> >>Ubuntu 6.06.1 LTS dapper >>Linux 2.6.17.13-K8 #3 SMP PREEMPT >> >>1000 Hz support is active. >> >>So far, we are running 2x CS:Source normal with 20 slots and 1x CS:Source >>GunGame with 14 slots. >> >>All 3 servers are running with Tickrate 100 and have 500 FPS (as u know, >>it >>always depends on when you are giving the server the stats command, but >>most >>of the time it will show 500 or 499). >> >>I haven't been reported any real lag-problems, although the server will >>get >>performance issues if running longer than 3 days without a restart (that >>should be something we all know about too, I guess). >> >>TOP is reporting as usual very weird CPU stats, being the one server at >>82% >>and the other two at around 60% ... connect to any of the servers and you >>won't notice any performance issues tho ... so my guess is that top still >>is >>showing totally wrong rates (and it has, imho, always been like this ;-) >> >>Memory is no issue at all ... maximum is one of the 20 slots server using >>about 20% ... >> >>Lemme know if u wanna know anything else (BESIDES graphs, as I don't have >>any yet !). >> >>Greetinx, matteo. > > After reading your reply, I was curious and attached ktrace to source > to see what it was doing. (16 man source server, full, no mods etc) > > 11753 srcds_amd RET old.recv 90/0x5a > 11753 srcds_amd CALL mincore(0xbfbfc5c0,0) > 11753 srcds_amd RET mincore 0 > 11753 srcds_amd CALL mincore(0xbfbfc5e0,0) > 11753 srcds_amd RET mincore 0 > 11753 srcds_amd CALL mincore(0xbfbfc5d0,0) > 11753 srcds_amd RET mincore 0 > 11753 srcds_amd CALL old.recv(0x5,0xbfbfb4d0) > 11753 srcds_amd RET old.recv -1 errno -11 Unknown error: -11 > 11753 srcds_amd CALL old.recv(0xa,0xbfbfb520) > 11753 srcds_amd RET old.recv -1 errno -11 Unknown error: -11 > 11753 srcds_amd CALL old.recv(0xc,0xbfbe4da0) > 11753 srcds_amd RET old.recv -1 errno -11 Unknown error: -11 > 11753 srcds_amd CALL mincore(0xbfbfb520,0) > 11753 srcds_amd RET mincore 0 > 11753 srcds_amd CALL getdomainname(0xbfbfc860,0) > 11753 srcds_amd RET getdomainname 0 > 11753 srcds_amd CALL old.gethostid(0x1,0xbfbfc6a0,0,0,0xbfbfc618) > 11753 srcds_amd RET old.gethostid 1 > 11753 srcds_amd CALL #175(0,0xbfbfc620,0,0x8) > 11753 srcds_amd RET #175 0 > 11753 srcds_amd CALL old.gethostid(0x1,0xbfbfc6a0,0,0,0xbfbfc610) > 11753 srcds_amd RET old.gethostid 1 > 11753 srcds_amd CALL read(0,0xbfbfc60f,0x1) > > Curious to why it's using older syscalls, maybe it's just the linux > emulatin layer on freebsd. > > A rundown of syscalls for a 1 second kdump (number of calls per > second (estimated)) > > old.recv 12162/s > mincore 9592/s > old.gethostid 3720/s > getdomainname 1875/s > > gethostid() according to gethostid(3) says it has been deprecated > > Does source really need to call getdomainname 1800 times a second? :) > > > > > > > > _______________________________________________ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
