Re: help:redo logging flooding a disk + redo io size

dallen_at_fns.com.au wrote:

> Please help :
>
> I have an OLTP app that does lots of inserts and updates where lots of
> columns are changed. When high txn rates are happening it seems like the disk
> is getting flooded by IO's. I have the database on a dac960pg raid
> controller. The logs are on their own disk. If I use performance monitor it
> shows that the disk is doing about 60 I/O per sec and that the av io size is
> about 8k.
>
> Q1) how can I better handle the io load of the logs
> Q2) is it possible to make the size of the IO's to the logs larger ?
>
> Thanks in advance
>
> -----------== Posted via Deja News, The Discussion Network ==----------
> http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own

Just my 0.02, but what's the DB_BLOCK_SIZE. I wonder if making it 8K would help. Also, the product of DB_BLOCK_SIZE and DB_FILE_MULTIBLOCK_READ_COUNT can be manipulated, based on the OS (64K on HP-UX) . Also, what sort of striping/multiplexing mechanism is being used for the redo logs? Is it 2 sets of files, on 3 disks or some other configuration? Could the redo logs be sized higher?

Also, you seem sure it's the redo logs. How about the rollback segments? Are they optimally sized?

HTH. Received on Thu Mar 11 1999 - 17:30:40 CST