This content has been marked as final. Show 6 replies
Nnnnnnnnnnoooooooooo such thing. But google for Cary Millsap paper "Thinking Clearly About Performance."
Edit: If it's benchmarks you're asking about, see http://kevinclosson.wordpress.com/2012/02/06/introducing-slob-the-silly-little-oracle-benchmark/
Edit 2: I didn't put the slob reference in at first because it is contention that causes many performance problems, and maximal i/o is not an indicator of that.
Edited by: jgarry on Mar 13, 2013 2:09 PM
Edited by: jgarry on Mar 13, 2013 2:10 PM
Read on I/O callibration. It might be of some help.
user622061 wrote:The "limit" has NOTHING to do with Oracle RDBMS & has everything to do with the disk sub-system.
All in the title.
How to find the xxxxbytes/second that generate big perfomance problem on the database ? ....
I'm searching for the "limit" in term of io to not exceed, otherwise the database suffer of performance problem.
The sub-components (the number & type of hardware interfaces between the CPU & disk drives,
the number of disk drives, the speed of the disk drives, how they are mapped & managed by any Volume Manager) influence the maximum & sustained data transfer rate to the database.
Yes, I've never said the opposite of what you said.
My aim is to mesure the "limit" IO FROM database that our queries can reach without explodind the server ...
I'm not speaking about the factors that influence this measure ....
Did you take a look at the link in jgarry's reply? That would seem to do what you want.
If you are looking at a single number, though, you're going to have problems. The rate of data you can read from any set of disks via random single-block reads is going to be very different than the rate of data you can read from that same set of disks doing multi-block reads (i.e. table scans) for example. And then there are issues with different types of writes as well.
I like your answer.
You are right.
In other words, the IO performance problem comes from CONTENTION not for reaching maximum reads per seconds. Are you agree ?
If so, what is the indicator from database ( metric, stat, event) we can measure to check if our server is suffer for IO Contention ?
I think to the IO events Like 'db file sequential" for every current session ?
If yes, what is the threashold of this event that causes performance problem ?