Re: Oracle RAID

In article <41e1gu$t03_at_news.ios.com> nocera_at_haven.ios.com "Nocera" writes:

             This is because redo logs write synchronously ( O_SYNC)
: while other Oracle writes to the database via DBWR use ether async IO
: or list IO.

Are you absolutely certain about this. Does DBWR _never_ use synchronous writes ?

Database blocks must never get into the 'future' relative to the redo log, which is why you get can redo synch writes, and why it would appear safe (generally) for DBWR to be using ASYNC (what is list I/O btw ?).

However, the purpose of a checkpoint is to bring the database up to date, and ensure that the redo log up to that point is redundant. This could not be guaranteed if the DBWR writes for a checkpoint were ASYNC, surely.

Question: if you start SYNC writing, would the previously queued ASYNC writes be flushed before the SYNC writes started ? Is this per device, or system global ?

Any comments anyone ?

-- 
Jonathan Lewis