Re: [squid-users] Tuning Squid for large user base

On Wed, 10 Mar 2004, Henrik Nordstrom wrote:

> On Wed, 10 Mar 2004, James MacLean wrote:
>
> > Broke out 4 squids on the one computer and divided the traffic of our B
> > class over them. Each Squid got a 3G disk cache and 64M cache_mem. From
> > everyone's feedback, some settings adjusted were :
> One drive per Squid is a little low and will become a significant
> bottleneck when the cache is full. You will not notice this bottleneck
> very much until Squid starts to recycle cache space.
> Regards
> Henrik

Thanks again Henrik. We will investigate that too. Did an strace count of
a busy squid this AM :

% time seconds usecs/call calls errors syscall
------ ----------- ----------- --------- --------- ----------------
 32.97 1.110003 108 10245 10 write
 25.52 0.859270 11 80062 77630 msgrcv
 13.27 0.446672 1823 245 poll
  9.57 0.322006 51 6364 2092 read
  3.94 0.132480 88 1497 close
  3.29 0.110635 139 796 796 connect
  3.15 0.105986 18 5893 fcntl64
  1.84 0.062008 78 793 socket
  1.52 0.051205 21 2425 msgsnd
  1.10 0.036997 48 765 85 accept
  0.78 0.026131 33 793 bind
  0.77 0.025970 33 781 getsockopt
  0.64 0.021694 27 793 setsockopt
  0.52 0.017630 339 52 select
  0.36 0.012283 102 121 sendto
  0.36 0.012200 18 680 getsockname
  0.15 0.005089 30 167 43 recvfrom
  0.15 0.004972 16 302 gettimeofday
  0.06 0.001994 1994 1 open
  0.03 0.000997 166 6 brk
  0.00 0.000149 75 2 getdents64
  0.00 0.000055 55 1 munmap
  0.00 0.000029 29 1 old_mmap
  0.00 0.000017 17 1 getrusage
  0.00 0.000015 15 1 fstat64
------ ----------- ----------- --------- --------- ----------------
100.00 3.366487 112787 80656 total

Lots of errors in msgrcv. Is that typical/ok ?

take care,
JES

-- 
James B. MacLean        macleajb@ednet.ns.ca
Department of Education 
Nova Scotia, Canada