RE: [squid-users] TCP response time | Proxy efficiency | Paging .. pending questions

From: <vincent.blondel_at_ing.be>
Date: Wed, 26 Aug 2009 08:25:42 +0200

no I do not. I simply manage some 30 squid servers and I included in
this mail some statistics from one of these proxies because I think I
should increase the cache_mem value for some of them but I am not really
sure about my suggestions ... so if some of you have deep experience in
it I would be happy they help me.

do not hesitate to read the 5 questions in my mail.

many thks
Vincent

_______________________________

        From: Gerard Leonardo [mailto:gerard.leonardo_at_gmail.com]
        Sent: Tuesday, August 25, 2009 3:06 PM
        To: Blondel, V. (Vincent)
        Cc: squid-users_at_squid-cache.org
        Subject: Re: [squid-users] TCP response time | Proxy efficiency
| Paging .. pending questions
        
        
        Hi, Do you mind sending how to tune or tweak squid+dansguardian
with 300 concurrent users.
        
        Thanks in advancE!
        Gerard

        On Tue, Aug 25, 2009 at 6:45 PM, <vincent.blondel_at_ing.be> wrote:
        

                Hello all,
                
                when looking on the good working of one of my squid
proxies, I see some
                values I think somewhat relevant for increasing the
cache_mem.
                
                next to my investigations I can see this machine
receives some 130req/s,
                sends some 80 req/s, caches some 1.2 million objects on
disk. this is
                the very basic part of it.
                
                $ squidclient mgr:info
                Internal Data Structures:
                       1182755 StoreEntries
                        70431 StoreEntries with MemObjects
                        70365 Hot Object Cache Items
                       1174685 on-disk objects
                
                $ squidclient mgr:5min |egrep 'http|fault'
                client_http.requests = 127.388482/sec
                client_http.hits = 45.899453/sec
                client_http.errors = 0.000000/sec
                client_http.kbytes_in = 95.962190/sec
                client_http.kbytes_out = 1820.344974/sec
                client_http.all_median_svc_time = 0.021898 seconds
                client_http.miss_median_svc_time = 0.042766 seconds
                client_http.nm_median_svc_time = 0.001789 seconds
                client_http.nh_median_svc_time = 0.020695 seconds
                client_http.hit_median_svc_time = 0.004626 seconds
                server.http.requests = 78.912393/sec
                server.http.errors = 0.000000/sec
                server.http.kbytes_in = 1285.901343/sec
                server.http.kbytes_out = 66.245877/sec
                page_faults = 0.043333/sec
                
                we can also see a correct hit (36%) and byte (20%) rate
on this server
                but ...
                
                * what does it mean "Proxy efficiency 46.01" ??
                * and what does it mean "Average speed increase 24.27% "
??
                * what does it mean "TCP response time of 100%%
requests" ..
                        ? the maximum time taken to make the SYN SYN/ACK
ACK when
                asking for a socket on the parent cahe
                        ? the maximum time taken for getting an object
from the parent
                cache
                        ???
                
                Summary
                Calamaris statistics
                
                lines parsed: lines 15855657
                invalid lines: lines 0
                parse time: sec 4145
                parse speed: lines/sec 3825
                
                Proxy statistics
                
                Total amount: requests 15855657
                Total Bandwidth: Byte 165G
                Proxy efficiency (HIT [kB/sec] / DIRECT [kB/sec]):
factor 46.01
                Average speed increase: % 24.27
                TCP response time of 100%% requests: msec 1900
                
                Cache statistics
                
                Total amount cached: requests 5627335
                Request hit rate: % 35.49
                Bandwidth savings: Byte 33781M
                Bandwidth savings in Percent (Byte hit rate): % 19.96
                Average cached object size: Byte 6294
                Average direct object size: Byte 13885
                Average object size: Byte 11191
                
                the machine squid is running on is a Sun Solaris 8 V210
with 2Gb memory.
                
                $ prtconf
                System Configuration: Sun Microsystems sun4u
                Memory size: 2048 Megabytes
                System Peripherals (Software Nodes):
                SUNW,Sun-Fire-V210
                
                concerning the memory usage this becomes much more
unclear for me ... if
                I good understand I see the process size is some 700mb
and get some
                600mb allocated. when going deeper in the vmstat stats
pi and po column
                respectively give 825 and 236 at some time ...
                
                
                * is this not some too big value ??
                * cache_mem is currently defined at 400mb, don't you
think increasing
                this value to a bigger value (let's say some 500-600 mb)
would not be
                better ??
                
                
                $ squidclient mgr:info
                Squid Object Cache: Version 2.7.STABLE4
                Start Time: Mon, 17 Aug 2009 02:00:38 GMT
                Current Time: Thu, 20 Aug 2009 10:48:06 GMT
                Connection information for squid:
                       Number of clients accessing cache: 2264
                       Number of HTTP requests received: 11355087
                       Number of ICP messages received: 11553839
                       Number of ICP messages sent: 11564676
                       Number of queued ICP replies: 0
                       Number of HTCP messages received: 0
                       Number of HTCP messages sent: 0
                       Request failure ratio: 0.00
                       Average HTTP requests per minute since start:
2342.5
                       Average ICP messages per minute since start:
4769.2
                       Select loop called: 94346313 times, 3.083 ms avg
                Cache information for squid:
                       Request Hit Ratios: 5min: 38.8%, 60min: 30.9%
                       Byte Hit Ratios: 5min: 21.8%, 60min: 29.7%
                       Request Memory Hit Ratios: 5min: 27.4%,
60min: 27.1%
                       Request Disk Hit Ratios: 5min: 23.4%,
60min: 18.0%
                       Storage Swap size: 29196377 KB
                       Storage Mem size: 409632 KB
                       Mean Object Size: 24.85 KB
                       Requests given to unlinkd: 114
                Median Service Times (seconds) 5 min 60 min:
                       HTTP Requests (All): 0.02190 0.01035
                       Cache Misses: 0.04047 0.03622
                       Cache Hits: 0.00463 0.00379
                       Near Hits: 0.01955 0.02317
                       Not-Modified Replies: 0.00179 0.00179
                       DNS Lookups: 0.00190 0.00190
                       ICP Queries: 0.00221 0.00221
                Resource usage for squid:
                       UP Time: 290848.297 seconds
                       CPU Time: 26579.690 seconds
                       CPU Usage: 9.14%
                       CPU Usage, 5 minute avg: 30.92%
                       CPU Usage, 60 minute avg: 32.65%
                       Process Data Segment Size via sbrk(): 693792 KB
                       Maximum Resident Size: 0 KB
                       Page faults with physical i/o: 19875
                Memory accounted for:
                       Total accounted: 603121 KB
                       memPoolAlloc calls: 1390282241
                       memPoolFree calls: 1384735607
                File descriptor usage for squid:
                       Maximum number of file descriptors: 8192
                       Largest file desc currently in use: 801
                       Number of file desc currently in use: 697
                       Files queued for open: 0
                       Available number of file descriptors: 7495
                       Reserved number of file descriptors: 100
                       Store Disk files open: 0
                       IO loop method: poll
                
                $ vmstat 5 5
                 procs memory page disk
faults
                cpu
                 r b w swap free re mf pi po fr de sr m0 m1 m3 m4
in sy cs us
                sy id
                 0 0 0 2614824 634360 62 277 76 29 29 0 0 3 0 3 5
4294967261 276
                176 10 9 81
                 0 0 0 2442720 414768 145 128 568 236 236 0 0 0 0 0 60
3418 15771 2959
                31 30 39
                 2 1 0 2442304 418104 296 851 739 75 75 0 0 27 0 1 20
2862 58159 2199
                41 45 14
                 0 0 0 2442720 423744 229 0 825 180 180 0 0 0 0 0 8
2749 36179 2911
                36 28 36
                 0 0 0 2442720 428832 146 0 452 96 96 0 0 0 0 1 4
3825 13153 2585
                26 30 45
                
                $ vmstat -i
                interrupt total rate
                --------------------------------
                clock 1582676343 100
                --------------------------------
                Total 1582676343 100
                
                
                $ vmstat -s
                       0 swap ins
                       0 swap outs
                       0 pages swapped in
                       0 pages swapped out
                119188729 total address trans. faults taken
                 63414018 page ins
                 33261045 page outs
                150679474 pages paged in
                 57581487 pages paged out
                985466858 total reclaims
                985358256 reclaims from free list
                       0 micro (hat) faults
                119188729 minor (as) faults
                 34878720 major faults
                1518806042 copy-on-write faults
                655083663 zero fill page faults
                  187523 pages examined by the clock daemon
                       0 revolutions of the clock hand
                 57477049 pages freed by the clock daemon
                 47526220 forks
                  149015 vforks
                 49086268 execs
                3171912755 cpu context switches
                1401743919 device interrupts
                3042943801 traps
                2429831033 system calls
                1065243178 total name lookups (cache hits 87%)
                151392974 user cpu
                148588621 system cpu
                1165931599 idle cpu
                116763714 wait cpu
                
                many thks to help me and do not hesitate to give me any
comment about
                the files give above.
                
                Vincent.
-----------------------------------------------------------------
ATTENTION:
The information in this electronic mail message is private and
confidential, and only intended for the addressee. Should you
receive this message by mistake, you are hereby notified that
any disclosure, reproduction, distribution or use of this
message is strictly prohibited. Please inform the sender by
reply transmission and delete the message without copying or
opening it.

Messages and attachments are scanned for all viruses known.
If this message contains password-protected attachments, the
files have NOT been scanned for viruses by the ING mail domain.
Always scan attachments before opening them.
-----------------------------------------------------------------
Received on Wed Aug 26 2009 - 06:25:51 MDT

This archive was generated by hypermail 2.2.0 : Wed Aug 26 2009 - 12:00:04 MDT