Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Re: RE: Running/Licensing Hyperthreaded Intel Pentium4 Xeon CPU's on

Re: RE: Running/Licensing Hyperthreaded Intel Pentium4 Xeon CPU's on

From: chao_ping <chao_ping_at_vip.163.com>
Date: Sun, 24 Nov 2002 02:53:37 -0800
Message-ID: <F001.0050A42A.20021124025337@fatcity.com>


Orr, Steve,

		Sorry for my late feedback. I went back home and did not receive the mail.
		As i said, i have two node running 9i rac, with the same load from the middleware(load balancing is done with hardware alton and middle ware tuxedo). 
		The following is one day's data after node rac1 is configured to use logical cpu and node rac2 is left with this feature off.
 

[oracle_at_rac1 log]$ tail sar021123.log

22:20:00 all 11.43 0.00 0.19 88.38 
22:30:00 all 10.43 0.00 0.17 89.41 
22:40:00 all 10.81 0.00 0.43 88.76 
22:50:00 all 10.45 0.00 0.17 89.38 
23:00:00 all 10.55 0.00 0.17 89.28 
23:10:00 all 10.65 0.00 0.38 88.96 
23:20:00 all 11.58 0.00 0.18 88.24 
23:30:01 all 10.90 0.00 0.14 88.96 
23:40:00 all 8.48 0.00 0.30 91.21 

Average: all 6.62 0.00 0.15 93.23

[oracle_at_rac1 log]$ rsh rac2 "tail log/sar021123.log"

22:20:00 all 23.88 0.00 0.38 75.74 
22:30:00 all 19.91 0.00 0.33 79.76 
22:40:00 all 19.93 0.00 0.74 79.33 
22:50:01 all 19.68 0.00 0.30 80.02 
23:00:00 all 19.27 0.00 0.31 80.42 
23:10:00 all 21.58 0.00 0.75 77.67 
23:20:00 all 19.88 0.00 0.30 79.81 
23:30:00 all 20.12 0.00 0.27 79.61 
23:40:00 all 17.26 0.00 0.61 82.14 

Average: all 12.45 0.00 0.27 87.28
        As we can see, the node rac1 has average 93.2 idle cpu in one days' time, and the second node has 97.28 average idle cpu.
		So, from my test result, hyperthread does give out better performance(at least more cpu power,can accept more concurrent user), and 6.8% average cpu usage vs 12.8 average cpu usage.
		



Regards
zhu chao
Eachnet DBA
86-21-32174588-667
chao_ping_at_vip.163.com
www.happyit.net
www.cnoug.org(Chinese Oracle User Group)

>I was off yesterday and came in this morning anticipating lots of feedback
>on this but now I'm surprised by the lack of response from all you seasoned
>gurus. :-) I guess this is a very new thing and you aren't aware of it yet.
>
>
>Anyway, here's what I've garnered so far:
>
>There was a discussion on slashdot where one guy said Oracle was wanting to
>charge licenses on the virtual CPU's instead of the actual number of
>physical CPU's. Then someone else said he had checked with Oracle for a
>definitive answer BEFORE implementing it because it was a make or break deal
>for him. He said Oracle was only going to charge for actual physical CPU's
>and not the virtual hyperthreaded CPU's so he went ahead with his
>implementation and says he got a 30 performance boost. Not the same
>performance boost as would be expected from adding real CPU's but
>significant nonetheless. He said it was a 30 "free" boost because he didn't
>have to buy anything; all he had to do was reconfigure Linux. In another
>discussion on the SUSE/Oracle list it was stated that the performance boost
>is about 20 of what you would expect from adding a real processor and that
>Oracle once again said the license charge would be by virtual processor and
>not the actual number of CPU's in the server. It was also stated that the
>max number of processors for the standard edition is 4 as decreed by his
>Oracle highness so with hyperthreading you could only have 2 real CPU's.
>Finally, I only got 3 hits on Metalink where Oracle support drones said you
>didn't have to do anything Oracle-wise to adjust for doubling the number of
>O/S visible CPU's. They did not respond to init.ora tuning questions and
>they refered us to our Oracle sales drones for licensing questions as they
>were clueless in Seatle.
>
>As usual there is controversy regarding Oracle's license confusion. Is there
>anything in the licensing legalese which specifically mentions licensing on
>virtual or multithreaded CPU's? Regardless, seems to me we need to have a
>Boston Tea Party type rebellion and refuse to pay for licenses on virtual
>CPU's.
>
>
>With righteous indignation and virtuous rebellion,
>Steve Orr
>Bozeman, Montana
>
>
>
>-----Original Message-----
>Sent: Wednesday, November 20, 2002 2:23 AM
>To: Multiple recipients of list ORACLE-L
>on Linu
>
>
>Orr, Steve,
> I am using Xeon MP as well on my DELL 6650 , and i have two
>nodes. I will patch my database next evening and later turn on the logical
>cpu feature on one node and leave the other node as it is.
> I will compare the cpu load/application responce time change and
>give you the feedback.
> Good luck
>
>Regards
>zhu chao
>Eachnet DBA
>86-21-32174588-667
>chao_ping_at_vip.163.com
>www.happyit.net
>www.cnoug.org(Chinese Oracle User Group)
>
>======= 2002-11-18 14:25:00 ,you wrote£º=======
>
>>OK, I've got this new Dell Linux server with 4 "Hyperthreaded" Pentium4
>Xeon
>>CPU's. There are 4 physical CPU's but with hyperthreading the O/S sees 8
>>CPU's as is reported in top. I've installed Oracle on this machine and it
>>seems to run as if there really were 8 CPU's. Could that be true? Is there
>a
>>performance gain for Oracle? --As if I really did have 8 CPU's? How does
>>this work? For init.ora tuning parameters should I count the 4 physical
>>CPU's or the 8 virtual CPU's? Any reports on running Oracle/Linux on these
>>hyperthreaded Pentium4 Xeon CPU's?
>>
>>I'm afraid to ask Oracle about the licensing since they always answer with
>>the higher number. ;-)
>>
>>
>>Steve Orr
>>Bozeman, Montana
>>--
>>Please see the official ORACLE-L FAQ: http://www.orafaq.com
>>--
>>Author: Orr, Steve
>> INET: sorr_at_rightnow.com
>--
>Please see the official ORACLE-L FAQ: http://www.orafaq.com
>--
>Author: Orr, Steve
> INET: sorr_at_rightnow.com
>
>Fat City Network Services -- 858-538-5051 http://www.fatcity.com
>San Diego, California -- Mailing list and web hosting services
>---------------------------------------------------------------------
>To REMOVE yourself from this mailing list, send an E-Mail message
>to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
>the message BODY, include a line containing: UNSUB ORACLE-L
>(or the name of mailing list you want to be removed from). You may
>also send the HELP command for other information (like subscribing).

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.com
-- 
Author: chao_ping
  INET: chao_ping_at_vip.163.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).
Received on Sun Nov 24 2002 - 04:53:37 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US