I have several questions
1) could I add another CPU (bi processor)
2) what are the interresting item in the stats
3) I have reach 105 requests/sec what is the potential number I can reach
4) I have some core files but nothing with dbx, what is the debug level
must I put in squid.conf
I have increased cache_mem to 128 Mb
here the new stats
sample_start_time = 989848750.989909 (Mon, 14 May 2001 13:59:10 GMT)
sample_end_time = 989849052.387055 (Mon, 14 May 2001 14:04:12 GMT)
client_http.requests = 105.548445/sec
client_http.hits = 0.082947/sec
client_http.errors = 0.009954/sec
client_http.kbytes_in = 42.920114/sec
client_http.kbytes_out = 458.421726/sec
client_http.all_median_svc_time = 0.583090 seconds
client_http.miss_median_svc_time = 0.764075 seconds
client_http.nm_median_svc_time = 0.018474 seconds
client_http.nh_median_svc_time = 0.000000 seconds
client_http.hit_median_svc_time = 0.304593 seconds
server.all.requests = 88.056574/sec
server.all.errors = 0.000000/sec
server.all.kbytes_in = 438.846226/sec
server.all.kbytes_out = 44.569102/sec
server.http.requests = 84.894633/sec
server.http.errors = 0.000000/sec
server.http.kbytes_in = 413.895757/sec
server.http.kbytes_out = 42.691181/sec
server.ftp.requests = 0.129397/sec
server.ftp.errors = 0.000000/sec
server.ftp.kbytes_in = 13.580089/sec
server.ftp.kbytes_out = 0.006636/sec
server.other.requests = 3.032544/sec
server.other.errors = 0.000000/sec
server.other.kbytes_in = 11.367062/sec
server.other.kbytes_out = 1.874603/sec
icp.pkts_sent = 177.317538/sec
icp.pkts_recv = 163.335986/sec
icp.queries_sent = 161.945794/sec
icp.replies_sent = 15.371745/sec
icp.queries_recv = 15.371745/sec
icp.replies_recv = 147.964241/sec
icp.replies_queued = 0.000000/sec
icp.query_timeouts = 0.000000/sec
icp.kbytes_sent = 13.666354/sec
icp.kbytes_recv = 12.057181/sec
icp.q_kbytes_sent = 12.548228/sec
icp.r_kbytes_sent = 1.118126/sec
icp.q_kbytes_recv = 1.181166/sec
icp.r_kbytes_recv = 10.879333/sec
icp.query_median_svc_time = 0.000000 seconds
icp.reply_median_svc_time = 0.000000 seconds
dns.median_svc_time = 0.119264 seconds
unlink.requests = 0.000000/sec
page_faults = 0.059722/sec
select_loops = 12.548228/sec
select_fds = 765.597827/sec
average_select_fd_period = 0.001215/fd
median_select_fds = 54.998322
swap.outs = 0.000000/sec
swap.ins = 0.003318/sec
swap.files_cleaned = 0.000000/sec
aborted_requests = 7.660988/sec
syscalls.polls = 54.665415/sec
syscalls.disk.opens = 0.009954/sec
syscalls.disk.closes = 0.009954/sec
syscalls.disk.reads = 0.003318/sec
syscalls.disk.writes = 483.425281/sec
syscalls.disk.seeks = 0.000000/sec
syscalls.disk.unlinks = 0.000000/sec
syscalls.sock.accepts = 52.286494/sec
syscalls.sock.sockets = 46.327579/sec
syscalls.sock.connects = 44.366711/sec
syscalls.sock.binds = 44.496108/sec
syscalls.sock.closes = 86.009441/sec
syscalls.sock.reads = 432.711463/sec
syscalls.sock.writes = 353.885236/sec
syscalls.sock.recvfroms = 200.585841/sec
syscalls.sock.sendtos = 188.432441/sec
cpu_time = 281.430000 seconds
wall_time = 301.397146 seconds
cpu_usage = 93.375138%
DISKD STATS
sent_count: 8173
recv_count: 8088
max_away: 140
max_shmuse: 129
open_fail_queue_len: 13
block_queue_len: 8980
OPS SUCCESS FAIL
open 537 530 0
create 759 758 0
close 1270 1268 0
unlink 2294 2171 63
read 821 812 0
write 2492 2486 0
Squid Object Cache: Version 2.4.STABLE1
Start Time:
Mon, 14 May 2001 13:18:02 GMT
Current Time:
Mon, 14 May 2001 14:10:46 GMT
Connection information for squid:
Number of clients accessing cache: 1732
Number of HTTP requests received: 316237
Number of ICP messages received: 494808
Number of ICP messages sent: 532502
Number of queued ICP replies: 0
Request failure ratio: 0.0000000000000%
HTTP requests per minute: 5997.4
ICP messages per minute: 19482.9
Select loop called: 56271 times, 56.223 ms avg
Cache information for squid:
Request Hit Ratios: 5min: 15.9%, 60min: 1.4%
Byte Hit Ratios: 5min: 1.5%, 60min: 4.7%
Request Memory Hit Ratios: 5min: 8.6%, 60min: 7.6%
Request Disk Hit Ratios: 5min: 34.8%, 60min: 39.0%
Storage Swap size: 6136628 KB
Storage Mem size: 31192 KB
Mean Object Size: 11.90 KB
Requests given to unlinkd: 4015
Median Service Times (seconds) 5 min 60 min:
HTTP Requests (All): 2.02066 0.49576
Cache Misses: 2.64709 0.68577
Cache Hits: 0.61549 0.55240
Near Hits: 2.79397 2.50793
Not-Modified Replies: 0.55240 0.52331
DNS Lookups: 0.22276 0.09117
ICP Queries: 0.00000 0.00000
Resource usage for squid:
UP Time: 3163.731 seconds
CPU Time: 2869.960 seconds
CPU Usage: 90.71%
CPU Usage, 5 minute avg: 67.81%
CPU Usage, 60 minute avg: 90.74%
Maximum Resident Size: 671488 KB
Page faults with physical i/o: 2539
Memory usage for squid via mallinfo():
Total space in arena: 166464 KB
Ordinary blocks: 148843 KB 2103615 blks
Small blocks: 0 KB 0 blks
Holding blocks: 0 KB 0 blks
Free Small blocks: 0 KB
Free Ordinary blocks: 1186 KB
Total in use: 148843 KB 89%
Total free: 1186 KB 1%
Memory accounted for:
Total accounted: 100039 KB
memPoolAlloc calls: 45409705
memPoolFree calls: 43531740
File descriptor usage for squid:
Maximum number of file descriptors: 32768
Largest file desc currently in use: 1660
Number of file desc currently in use: 1536
Files queued for open: 0
Available number of file descriptors: 31232
Reserved number of file descriptors: 100
Store Disk files open: 0
Internal Data Structures:
522117 StoreEntries
11574 StoreEntries with MemObjects
5588 Hot Object Cache Items
515851 on-disk objects
_______________________________________________________________________
Joe Cooper <joe@swelltech.com> on 14/05/2001 13:46:55
Please respond to Joe Cooper <joe@swelltech.com>
To: Stephane Blat/France/IBM@IBMFR
cc: squid-users@squid-cache.org
Subject: Re: [squid-users] Squid usage
Hmmm, can the CPU be upgraded reasonably? You'll be hitting the roof on
375MHz at your load (even if it is a proper processor, and not an x86).
Your large amount of RAM can make up for a lot, but CPU still can be a
limiting factor.
Also, adding a third and possibly fourth disk will be helpful.
Giving Squid more cache_mem will help marginally. Give it about 128MB,
and it will probably lower CPU usage slightly (almost immeasurably--but
every little bit helps, I guess).
Another important item, because you have so much RAM, there is no reason
why you can't up the size of your cache_dirs by a quite a bit. I would
say 24GB for your current drives, or 30GB if you add one or two more,
would be a nice total capacity for your object store. More than that is
overkill at your load (and will eat into the free RAM you have for
buffering disk i/o--thus /hurting/ performance), less will lead to an
unnecessarily large amount of disk activity in cleaning up--removing
files is probably a definite limiting factor on your system, because you
have such a small object store.
If none of those things helps lower your hit response time, you can
revert to some disk I/O tricks to help things along a little:
Set your minimum object size to be written to disk to something like 2k.
This will cause a little less work for the disks and will only
marginally impact hit ratio.
Set the low and high water marks farther apart... 80 and 100 are good
numbers for this trick. This leads to slightly less aggressive deletion
when your cache approaches 'full'.
Neither of these has a huge impact on performance (and can have slightly
negative impacts on hit ratio in some circumstances), but again, every
little bit helps.
BLATS@fr.ibm.com wrote:
>
> I use squid 2.4stable1 on AIX 4.3.3 , 375 MHZ, 2 GB RAM , 2*18GB for
> cache1 and for cache2
>
> here some options
>
> cache_mem 16 MB
>
> cache_dir diskd /cache1 3500 32 256
> cache_dir diskd /cache2 3500 32 256
> cache_store_log none
> emulate_httpd_log on
> log_icp_queries off
--
Joe Cooper <joe@swelltech.com>
Affordable Web Caching Proxy Appliances
http://www.swelltech.com
Received on Mon May 14 2001 - 08:50:49 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:59:59 MST