[squid-users] client_persistent_connections off

From: sk <suresh.kumar@dont-contact.us>
Date: Fri, 10 Dec 2004 20:41:15 +0700

Could some one help me understand why client_persistent_connections has a
good impact on cpu usage.

as read somewhere it has relation with FD
here's a snapshot.

sample_start_time = 1102685651.632932 (Fri, 10 Dec 2004 13:34:11 GMT)
sample_end_time = 1102685951.657148 (Fri, 10 Dec 2004 13:39:11 GMT)
client_http.requests = 39.986772/sec
client_http.hits = 19.465095/sec
client_http.errors = 0.000000/sec
client_http.kbytes_in = 24.074723/sec
client_http.kbytes_out = 234.997698/sec
client_http.all_median_svc_time = 0.615488 seconds
client_http.miss_median_svc_time = 1.115393 seconds
client_http.nm_median_svc_time = 0.003785 seconds
client_http.nh_median_svc_time = 1.242674 seconds
client_http.hit_median_svc_time = 0.003785 seconds

Connection information for squid:
        Number of clients accessing cache: 570
        Number of HTTP requests received: 721650
        Number of ICP messages received: 0
        Number of ICP messages sent: 0
        Number of queued ICP replies: 0
        Request failure ratio: 0.00
        Average HTTP requests per minute since start: 3343.3
        Average ICP messages per minute since start: 0.0
        Select loop called: 6008615 times, 2.155 ms avg
Cache information for squid:
        Request Hit Ratios: 5min: 48.7%, 60min: 49.9%
        Byte Hit Ratios: 5min: 18.7%, 60min: 21.1%
        Request Memory Hit Ratios: 5min: 17.9%, 60min: 15.6%
        Request Disk Hit Ratios: 5min: 31.1%, 60min: 29.9%
        Storage Swap size: 146346404 KB
        Storage Mem size: 204760 KB
        Mean Object Size: 15.07 KB
        Requests given to unlinkd: 202
Median Service Times (seconds) 5 min 60 min:
        HTTP Requests (All): 0.61549 0.61549
        Cache Misses: 1.11539 1.17732
        Cache Hits: 0.00379 0.00379
        Near Hits: 1.24267 1.24267
        Not-Modified Replies: 0.00379 0.00286
        DNS Lookups: 0.56687 0.56687
        ICP Queries: 0.00000 0.00000
Resource usage for squid:
        UP Time: 12951.166 seconds
        CPU Time: 4256.770 seconds
        CPU Usage: 32.87%
        CPU Usage, 5 minute avg: 17.84%
        CPU Usage, 60 minute avg: 23.05%
        Process Data Segment Size via sbrk(): 915349 KB
        Maximum Resident Size: 0 KB
        Page faults with physical i/o: 606
Memory usage for squid via mallinfo():
        Total space in arena: 261532 KB
        Ordinary blocks: 257878 KB 3424 blks
        Small blocks: 0 KB 0 blks
        Holding blocks: 6920 KB 6 blks
        Free Small blocks: 0 KB
        Free Ordinary blocks: 3653 KB
        Total in use: 264798 KB 99%
        Total free: 3653 KB 1%
        Total size: 268452 KB
Memory accounted for:
        Total accounted: 965689 KB
        memPoolAlloc calls: 127672944
        memPoolFree calls: 97520745
File descriptor usage for squid:
        Maximum number of file descriptors: 4096
        Largest file desc currently in use: 325
        Number of file desc currently in use: 247
        Files queued for open: 0
        Available number of file descriptors: 3849
        Reserved number of file descriptors: 100
        Store Disk files open: 0
Internal Data Structures:
        9715558 StoreEntries
         43593 StoreEntries with MemObjects
         43485 Hot Object Cache Items
        9712888 on-disk objects

ealier when the option was tunred "on" the fd usage touched 2000 with about
100 hits/sec. Then the cpu was near 98 -99. Now it has come down to 65 -70
under similar conditions.

Is it something due to my kernel limitations
linux 2.4.18 SMP.

Plan upgrading to 2.6.4 or greater. But on a test machine wccp wasn't coming
up with 2.6.9 kernel using the bundled ip_gre.

Please help.

TIA
SK
Received on Fri Dec 10 2004 - 06:41:02 MST

This archive was generated by hypermail pre-2.1.9 : Sat Jan 01 2005 - 12:00:02 MST