Questions about SYSTEM Statistics

Hello

After reading about the subject, some points stay unclear to me.

If I say:
exec dbms_stats.gather_system_stats('Start');

then a particular table, namely sys.aux_stats$ is populated with values measured while some DB activity took place. Should I flush the shared pool in order to force recalculation of execution plans?

Are these values considered by the CBO readily or is another command needed?

If I wanted to store these values in a different table, how can I create this table? (Since it must be suitable for this purpose).

Which might be the reason why after starting system stats gathering in two different servers (both V. 9.2.0.4, one Sun, the other Linux), only in one of them new values appeared on the sys.aux_stats$ table, while on the other the table remained empty? How long should I wait in order to get these values from the table?

And how is it possible that the MBRC value is reported as -1?

Is there any source for adequate or typical values that I can SET directly, trying to improve the way the CBO calculates costs? (You could report the values that you find in your server instead).

How has been your experience in deliberately changing the values in aux_stats$ differently from the ones gathered by Oracle?

How can I garantee that these values survive a server restart? Does the aux_stats$ table exist only in memory?

When will these gathered sysstat values will be actually considered: After stopping the gathering using the same command as above but with the ('Stop') argument? Or only BETWEEN the Start and Stop commands?

Why is it necessary to gather system stats according to different activity levels of the server if no matter how much activity is taking place any single block access will take the same time? I mean that the hardware won't get slower, any multiblock read operation will last exactly the same time at any point in time... or not?

Thanks in advance
Bye
Rick Denoire   Received on Sun Jun 20 2004 - 09:45:35 CDT