[SQU] Performance?!

Hi all,

My proxyserver has an performance problem. If there are more than 30
clients connected, then the cpu ussage of the squidprocess is going up to
100%
and
the response of the server is very very slow :-(
If I am running squid only with a few clients connected, there is no
problem (see below).

Resource usage for squid: (tested with a few clients)
UP Time: 72892.155 seconds
CPU Time: 3310.624 seconds
CPU Usage: 4.54%
CPU Usage, 5 minute avg: 0.02%
CPU Usage, 60 minute avg: 6.14%
Maximum Resident Size: 83408 KB
Page faults with physical i/o: 0
Request failure ratio: 0.00%
HTTP requests per minute: 10.6

# ps -efl | grep squid
    8001 I 201 713 1 0.0 44 0 120K wait 15:20:45
    0:00.02 /usr/local/squid/bin/squid
80008001 S 201 714 713 0.0 44 0 84M event 15:20:45
   59:50.02 (squid)
80808001 S + 0 8184 7623 0.0 44 0 184K 5d8d5e94 12:03:21
    0:00.01 grep squid

I am using an AlphaServer 2100A 4/275 with 1 Alpha EV4.5 (21064) 275 MHz
processor (4Mbyte b-cache) and 2048 Mbytes of RAM
with 9 Harddisks - (SCSI Wide16) and stripped volumes (LSM) for the
cache-dirs and log-dirs ;

Filesystem 1024-blocks Used Available Capacity Mounted
on
/dev/rz1a 503455 119379 333730 27%
/
/dev/rz8a 3729138 771745 2584479 23%
/syssav
/proc 0 0 0
       100% /proc
usr_domain#usr 1543240 677212 840176 45% /usr
cachedom1#cache1_fs 17770496 337413 17397304 2% /cache1
cachedom2#cache2_fs 17770496 321530 17414088 2% /cache2
logdom#logs_fs 17770496 306123 17455792 2%
/logs
 dev/rz8f 4860829 1 4374745
  1% /logsav

# swapon -sv
Swap partition /dev/rz2b (default swap):
    Allocated space: 523755 pages (4091MB)
    In-use space: 1 pages ( 0%)
    Free space: 523754 pages ( 99%)

Total swap allocation:
    Allocated space: 523755 pages (4091MB)
    Reserved space: 27245 pages ( 5%)
    In-use space: 1 pages ( 0%)
    Available space: 496510 pages ( 94%)

Virtual Memory Statistics: (pagesize = 8192)
  procs memory pages intr cpu
   
  r w u act free wire fault cow zero react pin pout in sy cs us
sy id
  2 88 28 52K 182K 21K 2M 160K 1M 320 142K 0 13 138 289 6 2
92

My Ethernet Interface is an DEC De500-BA TULIP (10/100)(DECchip 21143:
Rev. 3.0) card (100 Mbit/s)

My Software:
Digital UNIX V4.0F (Rev. 1229);
Squid Cache: Version 2.3.STABLE4;
Storage Management: AdvFs / LSM

I tried this (but no success):
11.46 Squid uses 100% CPU
There may be many causes for this.
Andrew Doroshenko reports that removing /dev/null, or mounting a
filesystem with the nodev option, can cause Squid to use 100% of CPU. His
suggested
solution is to ``touch /dev/null.''

My squid.conf (using a lot of ACLs):
(behind a firewall)
------ snip ----

http_port 8080
icp_port 0

cache_peer xxx.xxx.xxx.xxx parent 8080 0 no-query default

dead_peer_timeout 15 seconds

hierarchy_stoplist cgi-bin ?

acl QUERY urlpath_regex cgi-bin \?
no_cache deny QUERY

cache_mem 16 MB

cache_swap_low 90
cache_swap_high 95

ipcache_size 4096
ipcache_low 90
ipcache_high 95
fqdncache_size 4096

cache_dir ufs /cache1 15500 16 256
cache_dir ufs /cache2 15500 16 256

cache_store_log none

emulate_httpd_log off
log_mime_hdrs off
log_fqdn off

redirect_children 15

refresh_pattern ^ftp: 1440 20% 10080
refresh_pattern ^gopher: 1440 0% 1440
refresh_pattern . 0 20% 4320

quick_abort_min 128 KB
quick_abort_max 4096 KB
quick_abort_pct 75

negative_ttl 5 minutes
positive_dns_ttl 6 hours
negative_dns_ttl 5 minutes
peer_connect_timeout 90 seconds

half_closed_clients off

miss_access allow all

cache_effective_user squid
cache_effective_group squid

memory_pools off
log_icp_queries off

client_db on
netdb_low 900
netdb_high 1000

query_icmp off
test_reachability off
buffered_logs on
prefer_direct off
ignore_unknown_nameservers on

---- snip ----

Any ideas?
Please help.

yours,
Mike

-- 
Sent through GMX FreeMail - http://www.gmx.net
--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html