Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

NW 6.0 and 6.5 with Hyperthreading

6 views
Skip to first unread message

Daniel C. Hunt

unread,
Jul 21, 2004, 9:08:01 PM7/21/04
to
I was catching up on forum reading tonight and noticed some posts by
Massimo Rosen and Barry Schnur a while ago to the effect that
Hypertheading in a NetWare environment is not good.

I did some searching on the web and found an article written by
Novell's own Dana Henricksen in Dell Power Solutions Magazine (I didn't
know they had one), August 2002 to the effect that there was a
significant improvement using Hyperthreading, at least under some
circumstances. See this URL:

http://www1.us.dell.com/content/topics/global.aspx/power/en/ps3q02_henr
iksen?c=us&l=en&s=esg

Dana is a pretty knowledgeable fellow and writes a fair amount of
NetWare code so I would give his opinion some weight.

Barry and Massimo, why don't you think Hyperthreading is any good for a
NetWare 6.x server with current patches?

Thanks,

Dan (who doesn't want to go back and turn hypertheading off on all of
those servers <grin>)

Hamish

unread,
Jul 21, 2004, 11:04:46 PM7/21/04
to
Daniel,

Take a look at:
http://www.novell.com/coolsolutions/netware/assets/improve_backup.pdf

Page 4, paragraph 2.

On a live system, doing testing I found upto a 66% decrease in file
throughput performance if the client worker thread was on a virtual CPU.

--
Hamish Speirs
Novell Support Forums Volunteer Sysop.

http://haitch.net

(Please, no email unless requested. Unsolicited support emails will
probably be ignored)

Daniel C. Hunt

unread,
Jul 22, 2004, 6:25:55 AM7/22/04
to
Thanks, Hamish. When I re-read Dana's report, it looks like the issue
might revolve around having at least two physical processors that have
hyperthreading turned on (there might be an advantage here) compared to
only one physical processor.

So, especially if you only has 1 physical processor in a machine, it is
recommended that you do NOT turn on hyperthreading? Is that the
correct conclusion?

In other words, a WS that is doing spreadsheet or WordProcessing work
where there is not a lot of IO might benefit from Hyperthreading but a
server which is doing 80% IO would not?

Dan

Hamish

unread,
Jul 22, 2004, 7:45:51 AM7/22/04
to
Daniel,

> So, especially if you only has 1 physical processor in a machine, it is
> recommended that you do NOT turn on hyperthreading? Is that the
> correct conclusion?

My recomendation is to turn hyperthreading off, irrespective of the
number of processors you have.

> In other words, a WS that is doing spreadsheet or WordProcessing work
> where there is not a lot of IO might benefit from Hyperthreading but a
> server which is doing 80% IO would not?

Correct - the server in fact will potentially perform worse with HT enabled.

geoffs....@otcnetworks.invalid.com

unread,
Jul 22, 2004, 3:21:50 PM7/22/04
to
For an explanation of why HT does or doesn't help and some conditions that
must be met for it to help, see my web page at
http://www.otcnetworks.com/servermb.htm#smp

Daniel C. Hunt

unread,
Jul 22, 2004, 6:50:56 PM7/22/04
to
Thanks Hamish and Geoffs. I appreciate the links to the various
articles.

Dan

Eric Anderson

unread,
Jul 26, 2004, 12:05:38 PM7/26/04
to
Daniel:

A suggestion. If you want to run hyper-threading, at least baseline
your server without having it turned on first. At least then you'll
know what impact it's having on your server.

As for myself, I've had mixed results with it - and the first time it
was problematic, I didn't initially realize that was the issue. A
baseline should help you quickly rule in or out the impact of HT.

Incidentally, you mentioned that you didn't want to go back to all those
servers and make the change. You only need to turn it off in the bios.
Don't change out the drivers.

Eric

0 new messages