> these two values will be recalculated with the formulas of noworkload statistics, right?

Yes. The formulars provided at page 119 of TOP should apply.

> which is what could happen (with workload statistics gathered), for example when you

> change from 10g to 11g recalculating the system statistics, right?

Yes.

HTH

Chris

in your case, where mreadtim < sreadtim, these two values will be recalculated

with the formulas of noworkload statistics, right?

which is what could happen (with workload statistics gathered), for example when you change from 10g to 11g recalculating the system statistics, right?

the cost calculation could be less accurate

Regards,

Donatello

Honestly, I never spent much time investigating such cases. So, my general explanation is that caching coupled with read-ahead/prefeching algorithms might easily lead to such situations. In fact, an intelligent controller/SAN/NAS should be able to spot when large sequencial reads are performed.

HTH

Chris

Why some time we get the mreadtime<sreadtim.

Thanks

]]>