一般的,lgwr写到current log buffer block并且包括current log buffer block,(writes up to and including the current log buffer block),current log buffer block通常仅是部分满。包含redo generation的index(containing the index into the log buffer for redo generation)接着移动到next log buffer block的开始,before LGWR releases the redo allocation latch prior to writing.在写出前,在LGWR释放redo allocation latch前。按照这种方式,在绝大多数redo writes的last log block剩余一些不可用空间。这些剩余的字节数累计在redo wastage统计中。
Redo wastage并不是问题。的确,it benefits performance by preserving LGWR's sequential I/O pattern.但是,high redo wastage表明lgwr过于活动,这可能是一个问题。在许多实例中,期望lgwr highly active,但是如果log buffer太小(more correctly, the _log_io_size),或commit rate太高,他可能overactive。如果LGWR overactive, it will place unnecessary load on the redo latches, and on the I/O subsystem.
来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/9599/viewspace-472959/,如需转载,请注明出处,否则将追究法律责任。
转载于:http://blog.itpub.net/9599/viewspace-472959/