Hi Kevin, I'm trying to demonstrate the difference of writing/spooling on direct IO vs cached filesystem(and to understand how exaclty the redo log spooling(huge amount of data, parallel execution) differ compared to the small data spooling (and I got the answer from Carel-Jan for that). It's obvious that as long as the filesystem/SAN controller's cache size and efficiency is able to hold/flush the amount of data before the cache gets full and the new data arrives, it will be beneficial to put the log archive destination on buffered file system (think about 6x10 to 32Mb online redo logs for small OLTP instance)
> ettest to compare different network configs, maybe
Why network config? SQLNet messages "from client" in this case is the OK for successful "write" operations(the sqlplus client is running on the database server host).
> Of course a SELECT * from a small table cached > in the OS pagecache is going to be exponentially > faster than physical disk reads.
It's about "writes", not about "reads".
Cheers Dimitre
-- -- Original Message -- -- From: "Kevin Closson" <kevinc@(protected)> To: <oracle-l@(protected)> Sent: Friday, June 24, 2005 5:34 PM Subject: RE: Q: Filesystem choice for log_archive_dest
> > >>Just trying to estimate the speed of writing from Oracle >>instance to different filesystems. >> >>1. select * from table with 108127 records, > > I'm afraid you've got me confused. You are selecting > all the rows from a little table to ascertain > disk write throughput? This would be a good > ettest to compare different network configs, maybe, > but is an apples to bicycles comparison with regard > to archive log spooling. > > Of course a SELECT * from a small table cached > in the OS pagecache is going to be exponentially > faster than physical disk reads. We need to get > beyond that comparison. Memory is indeed faster > than magnetic media > > > -- > http://www.freelists.org/webpage/oracle-l
