Hello everybody!
I run squid 1.1.18 on FreeBSD 2.2.2. It is linked against GNU malloc.
Now my cache swap contains about 700,000 objects, and after squid
restart I get the following mstats :-)
=====
...
Maximum Resident Size: 187672 KB
Page faults with physical i/o: 78
Memory usage for squid via mstats():
Total space in arena: 196928 KB
Total free: 105803 KB 54%
...
Accounted Memory Usage:
StoreEntry 673445 x 52 bytes = 34198 KB
URL strings = 31425 KB
IPCacheEntry 42 x 36 bytes = 1 KB
FQDNCacheEntry 0 x 56 bytes = 0 KB
Hash link 1 x 12 bytes = 0 KB
Pool MemObject structures 76 x 100 bytes = 7 KB ( 0 free)
Pool for Request structur 70 x 4536 bytes = 310 KB ( 0 free)
Pool for in-memory object 13 x 4096 bytes = 52 KB ( 0 free)
Pool for disk I/O 23 x 8192 bytes = 184 KB ( 0 free)
NetDB Address Entries 0 x 76 bytes = 0 KB
NetDB Host Entries 0 x 8 bytes = 0 KB
NetDB Peer Entries 0 x 24 bytes = 0 KB
ClientDB Entries 1 x 292 bytes = 0 KB
Miscellaneous = 1674 KB
Total Accounted = 67853 KB
=====
I was observing the same thing with standard FreeBSD malloc (that
squid's size in memory is twice its accounted size)
It seems that squid's behaviour causes a heavy fragmentation of the arena...
I went doing RTFS...
SY, Yar
Received on Thu Dec 18 1997 - 11:41:47 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:37:59 MST