Re: [squid-users] Squid usage

From: <BLATS@dont-contact.us>
Date: Mon, 14 May 2001 16:50:35 +0200

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