Re: Hyperthreading - Oracle license

From: Kevin Closson <ora_kclosson_at_yahoo.com>
Date: Fri, 11 Jan 2013 09:39:24 -0800 (PST)
Message-ID: <1357925964.60465.YahooMailNeo_at_web162801.mail.bf1.yahoo.com>

SMT cores should be thought of as another .3 of a core...and platform experts would do well to accept the new reality: 70 is the new 100

SMT is not a bad thing..not understanding it can be.

If I can get my head out of the foxhole without seeing tracers I actually have a blog post teed up on SMT 

The feature is called Simultaneous MultiThreading but since a thread is stalled until its peer thread stalls (all the while the OS charges each of them against their time slice over any given wall-clock period) I sort of poo-poo the use of the word simultaneous.

No war-horse here, but best to test.

---

 From: Rich Jesse <rjoralist2_at_society.servebeer.com> To: oracle-l_at_freelists.org
Sent: Thursday, January 10, 2013 2:30 PM Subject: RE: Hyperthreading - Oracle license  

Mark writes:

> Actually, a quick MOS search, and Oracle specifically recommends staying w/
> the "doubled" count for cpu_count.
>
> See Doc ID 289870.1

I saw that, too, and I'm calling "BS" on the lack of evidence, empirical or otherwise.  Unless it can be claimed that turning HT on gives one at or near a 100% gain in CPU power, it stands to reason that Oracle cannot consume at or near an additional 100%.  And I'm reasonably certain that no one is advocating HT as giving anywhere near another the performance of another core.

I understand that the CPU_COUNT isn't based on the raw CPU power.  Perhaps the suggestion was made *assuming* the reader would somehow magically consider recollecting system stats after turning on HT?

Call me paranoid (no, really, go ahead), but it just doesn't smell right...

Rich

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

Received on Fri Jan 11 2013 - 18:39:24 CET