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
