redo log block size and improving performance
Oracle 10.2.0.4:
I am tyring to get answer to this question but have not been able to get enough information. I am seeing high "log file sync" wait events when running under high load and looking at other resources as well as Metalink it looks like all that is related to IO. This affects online users, basically they send a transaction with 100K blob which we insert into DB. So if we have 80-100 transactions/sec going on we slow down like crzy. Now I am trying to see what's causing it. So I started to read about how to improve IO performance, came accross http://download.oracle.com/docs/cd/B19306_01/server.102/b14211/iodesign.htm . We are using SAN storage and all of our disks are stripped across, I think our stripe size is 64K. Our default DB_BLOCK_SIZE is 8k. redo logs are 2GB with 2 members. So now I am trying to understand the following:
1. What happens when a LOB of 100K is written to the redo log file? Does it do 13 IO operations because the redo block size is 8K?
2. Is the redo log size is 8K also?
3. Is there a way to reduce IO operations by changing block sizes?
4. Above doc talks about using ASYNC IO but is that also applicable when user does a commit. Is there something I can look at or should look at?
I need your advise in terms of what else I should look at?