16hot 的博客

SQUID 2.6都出到 STABLE4了，今天花了些时间进行了升级。

2.6与2.5，在配置选项方面有些调整，需要配置好才能正常的运行。

相比2.5，在BSD下采用kqueue，性能有很大提升，另外在COSS方面也有提升。

升级后，对于CPU消耗降低了20-30%。

终于，终于在今天弄好三台在一个交换机下的服务器来测试缓存了。

现在分别是FreeBSD 6.1，Redhat AS 4，Debian 3.1。

sqiud运行在FreeBSD上，Redhat作为polysvr，Debian作为polyclt。

从运行情况来看，最高并发连接1.8万。之后没有多久，mbuf满了，中断测试。

不过命中率似乎还不太理想，内存命中80%左右，5min命中，5%。

另外还是出现内存使用太大后，提示xcalloc错误，然后就自动重启squid。

2006/06/14 16:39:34| Ready to serve requests.
2006/06/14 16:39:34| Finished rebuilding storage from disk.
2006/06/14 16:39:34| 0 Entries scanned
2006/06/14 16:39:34| 0 Invalid entries.
2006/06/14 16:39:34| 0 With invalid flags.
2006/06/14 16:39:34| 0 Objects loaded.
2006/06/14 16:39:34| 0 Objects expired.
2006/06/14 16:39:34| 0 Objects cancelled.
2006/06/14 16:39:34| 0 Duplicate URLs purged.
2006/06/14 16:39:34| 0 Swapfile clashes avoided.
2006/06/14 16:39:34| Took 0.0 seconds ( 0.0 objects/sec).
2006/06/14 16:39:34| Beginning Validation Procedure
2006/06/14 16:39:34| Completed Validation Procedure
2006/06/14 16:39:34| Validated 0 Entries
2006/06/14 16:39:34| store_swap_size = 0k
2006/06/14 16:39:35| storeLateRelease: released 0 objects
FATAL: xcalloc: Unable to allocate 1 blocks of 4108 bytes!

Squid Cache (Version 2.5.STABLE14): Terminated abnormally.
CPU Usage: 214.289 seconds = 73.790 user + 140.499 sys
Maximum Resident Size: 1577032 KB
Page faults with physical i/o: 0