Re: Effective compression ratio with minimal performance overhead

From: Tim Gorman <tim.evdbt_at_gmail.com>
Date: Mon, 23 May 2022 05:42:09 -0700
Message-ID: <4e6cfc0d-68a3-4305-b980-70f6d239bec2_at_gmail.com>



Mladen,

That is unlikely, as HCC compression (and TDE, for that matter) is performed immediately prior to direct-path INSERTs, so that completed blocks (including compression and/or encryption) are written to the underlying storage layer.

It would be interesting to see if Exadata somehow changes that?

Thanks!

-Tim

On 5/23/2022 5:05 AM, Mladen Gogala wrote:
> On 5/23/22 07:43, Jonathan Lewis wrote:
>>
>> _at_Mladen Gogala <mailto:gogala.mladen_at_gmail.com>
>>
>> When did that change; I haven't been watching closely but the
>> compression used to be done at the compute node, and HC decompression
>> at storage cells, through software, except for the cases where the
>> storage cell sent the CU to the compute node either to balance the
>> CPU usage or because the extract from a smart scan was too large for
>> the message buffer.
>>
>> Regards
>> Jonathan Lewis
>
> Hi Jonathan,
>
> I think that I've read in some white paper that HCC is done on storage
> nodes, transparently to the compute nodes, as of X7. I'll look for the
> paper and let you know.
>
> Regards
>
> --
> Mladen Gogala
> Database Consultant
> Tel: (347) 321-1217
> https://dbwhisperer.wordpress.com

--
http://www.freelists.org/webpage/oracle-l
Received on Mon May 23 2022 - 14:42:09 CEST

Original text of this message