Re: Query with same plan running longer

This looks like it might be helpful.
https://exadatadba.blog/2021/09/21/exadata-quarantines-not-your-typical-oracle-quarantine/

On Tue, 1 Mar 2022 at 13:07, Jonathan Lewis <jlewisoracle_at_gmail.com> wrote:

> Obvious guess number 2 would have been about storage indexes
> Obvious guess number 3 would have been about flash cache
> I don't think I would have got to passthru without the statistics in front
> of me.
>
> If you don't look at your own results and think about them, and don't show
> us any relevant information, we just have to keep guessing until you
> confirm a guess; and that's a huge waste of effort. You spotted the
> passthru - what did you discover about that. Didn't you notice the "blocks
> decrypted"?
>
> NAME Diff
> blocks decrypted 11,058,075
> cell physical IO interconnect bytes 90,600,000,000
> cell num bytes in passthru due to quarantine 90,600,000,000
> cell num smart IO sessions using passthru mode due to cellsrv 1
> cell physical IO interconnect bytes returned by smart scan 90,600,000,000
> cell num bytes in passthru during predicate offload 90,600,000,000
>
> All your extra I/O corresponds to passthru - and the number of blocks
> decrypted (which could consume a lot of CPU) by the database is a good
> match the interconnect bytes. And there's a message there about quarantine
> - which should be investigated. And all that passthru happens during smart
> scan and predicate offload.
>
>
>
>
>
>
>
>
>
> On Tue, 1 Mar 2022 at 11:02, Lok P <loknath.73_at_gmail.com> wrote:
>
>> Thank you Jonathan.
>> I have checked v$stataname but have not seen any statistics with the
>> name '%load balanc%', is that statistics you referring to having a
>> different name? And here table TAB1 is not compressed and is not
>> partitioned too.
>>
>>

--
http://www.freelists.org/webpage/oracle-l