|
原帖由 Yong Huang 于 2010-4-13 01:12 发表 ![]()
老熊 (and a few others),
I think the problem here is that sometimes his insert (conventional) is slower than truncate. If the buffer cache size was a factor affecting the truncate speed, it would equally affect insert. Don't you think?
Yong Huang
但是insert与truncate时buffer cache中的数据流是两回事了。insert时将block从磁盘上读入到buffer cache,而truncate时buffer cache中的block被刷新到磁盘。如果buffer cache足够大,那么在insert过程中,buffer cache中的block不会被写到磁盘。有时conventional insert比较慢,是否有block被刷出buffer cache,是否是日志切换时间导致的较慢?
这种比较,建议用10046事件做个trace可能更有助于分析。 |
|