RE: ideal CPU/Memory relation

From: Mark W. Farnham <mwf_at_rsiz.com>
Date: Fri, 19 Aug 2022 18:50:33 -0400
Message-ID: <16a301d8b41e$17484360$45d8ca20$_at_rsiz.com>


'for OLTP, all the queries are always known and completely tuned'.

LOL. Good one Frits!

That is only possibly true for an unconfigurable OTS suite of OLTP applications, which, being unsellable, has never been offered on the open market.

I have never seen an OLTP suite that did not contain a substantial datawarehouse component of reports, and the shape and size of those reports is completely dependent on the scales of transactions and things like number of inventory items.

An Oracle salesman famously said "Of course E-biz will handle Burlington Coat Factory's volume. Oracle runs it. We do a half billion in sales a year and you do a half billion in sales a year. Why wouldn't it work?"

Hint: Burlington Coat's average transaction was more like a six pack of crew socks and a windbreaker than a software license sale.

Getting Coat's volume of transactions to work circa 1989-1990 was a wonderful playground for me. On the topic, a Sequent engineer pointed out that while they *could* put 24 CPU boards in, 22 maxed out the memory bus, so we would be better off using the slot for an additional disk controller. Kevin might remember who told us that (which we verified). It might have been the world's best computer salesman ever, Ed Rearick.

mwf

-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Frits Hoogland
Sent: Friday, August 19, 2022 6:25 AM
To: Stefan Koehler; Lothar Flatz
Cc: oracle-l_at_freelists.org
Subject: Re: ideal CPU/Memory relation

There is no single ratio that is optimal for all use cases of any database. But that is the gist of what everybody else is saying here.

Be aware that most benchmarks typically are done for a single use case, and thus allow more to be sized with a fixed ratio, whilst in real life there is no single use case for a database, unless it's very specific (like Martin Klier mentioned).

In fact, I had a huge twitter discussion with the CTO of an OLTP database, where I stated that no real life database strictly is OLTP and not all queries are known and tuned.
His persistent view was that 'for OLTP, all the queries are always known and completely tuned'.
Oh the irony, as many on this list have been working on getting the issues that come from this solved for many years, and many years to come.

Frits

> On 19 Aug 2022, at 12:17, Stefan Koehler <contact_at_soocs.de> wrote:
>
> Hello Lothar,
> there are such databases like SAP HANA that use such an approach - quoting
doc: "Using core-to-memory ratios which can be derived from the certified HANA listings. The memory requirement drives the number of cores required."
>
> ... but I guess you are using Oracle and never have seen such
> guidelines in the field until yet :-)
>
> Best Regards
> Stefan Koehler
>
> Independent Oracle performance consultant and researcher
> Website: http://www.soocs.de
> Twitter: _at_OracleSK
>
>> Lothar Flatz <l.flatz_at_bluewin.ch> hat am 19.08.2022 09:02 CEST geschrieben:

>> 
>> 
>> Hi,
>> 
>> had somebody ever heard of a ideal CPU/Memory relation for a database 
>> server?
>> A supplier of a customer stated such thing, I suppose they made it 
>> up.
>> Any comments?
>> 
>> Thanks
>> 
>> Lothar
>> --
>> http://www.freelists.org/webpage/oracle-l

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





--
http://www.freelists.org/webpage/oracle-l
Received on Sat Aug 20 2022 - 00:50:33 CEST

Original text of this message