Re: Fastest way to count exact number of rows in a very large table
Date: Mon, 05 Oct 2020 19:06:20 -0400
Message-ID: <48c1135c1faabf5909e14fc3e3401317d5c67c7c.camel_at_gmail.com>
Hi Andy, Thank you very much for sharing your experience. I have a
problem with a report which does count(distinct) but the customer is
still on 12.2. My employer is testing 19.8 . The customer has a
moderately sized DB of around 30 TB and they will upgrade from the full
rack Exadata X4 to the full rack Exadata X8-2 in the beginning of 2021.
The customer will also acquire in-memory license so the company that I
work for is looking how to utilize that. Being a system architect, I am
involved in the development. Thanks again for sharing your knowledge.
BTW, did you know that when upgrading to the newer model of Exadata,
Oracle Consulting will migrate the data for you?Regards
On Mon, 2020-10-05 at 22:33 +0100, Andy Sayer wrote:
> Mladen, RE: APPROX_FOR_COUNT_DISTINCT, it reduces the CPU impact for
> count distinct but you don't have that requirement for a standard
> count - there's no need to decide whether you've seen the value
> before. It is incredib ly fast (and highly accurate) when compared to
> count distinct but a standard count(*) would be quicker (and here the
> main time taken would probably be from IO).
>
> Thanks,
> Andy
-- Mladen Gogala Database Consultant Tel: (347) 321-1217 -- http://www.freelists.org/webpage/oracle-lReceived on Tue Oct 06 2020 - 01:06:20 CEST