Re: Extended RAC on SE

From: Mark Brinsmead <mark.brinsmead_at_gmail.com>
Date: Wed, 2 Sep 2015 17:04:09 -0600
Message-ID: <CAAaXtLA3RfV1gWf4pwkVPetLk8RrnQ0zdHQWpYqPiq_XTyCjxg_at_mail.gmail.com>



I think this is *almost* a no-brainer, isn't it?

One core with two hyperthreads offers more performance (throughput) that one core with hyperthreads disabled, but nowhere near as much as two full cores without hyperthreads.

If you have enough cores to bring you to (or near) the licensing limit, disable hyperthreads. If you are well below the limit (50% or less) enable hyperthreads.

The only place this would be tricky would be at about 75% of the limit. You would need to do some testing to decide whether it is better to run on 12 native cores, or to use 8 cores with hyperthreading enabled for a total of 16 hypethreads. Which one performs better will probably depend completely on your workload.


This limitation on CPU threads may be particularly nasty, though for users of other architectures, we you can have 4 or 8 (or more) hyperthreads per core. The arithmetic and the tradeoffs will not be nearly so simple there. Even so, you will probably get better performance from 16 native cores than from 16 hyperthreads on almost any architecture. Hyperthreads stall under conditions where cores do not.

On Wed, Sep 2, 2015 at 4:10 AM, Ls Cheng <exriscer_at_gmail.com> wrote:

> now choose... should I use threads or cores if we use Intel CPUs?
>
> I think someone did some comparison between hyperthreading and core a
> couple of years ago, I forgot who though!
>
>
>
> On Wed, Sep 2, 2015 at 11:53 AM, Tom Dale <tom.dale_at_fivium.co.uk> wrote:
>
>> Oracle has updated their licence doc with SE2 details
>>
>> http://www.oracle.com/us/corporate/pricing/databaselicensing-070584.pdf
>>
>> Oracle Database Standard Edition 2 may only be licensed on servers that
>> have a maximum capacity of 2 sockets. When used with Oracle Real
>> Application Clusters, Oracle Database Standard Edition 2 may only be
>> licensed on a maximum of 2 one-socket servers. In addition, notwithstanding
>> any provision in Your Oracle license agreement to the contrary, each Oracle
>> Database Standard Edition 2 database may use a maximum of 16 CPU threads at
>> any time. When used with Oracle Real Application Clusters, each Oracle
>> Database Standard Edition 2 database may use a maximum of 8 CPU threads per
>> instance at any time. The minimums when licensing by Named User Plus (NUP)
>> metric are 10 NUP licenses per server.
>>
>> This is a pretty big change :
>>
>> - a change to a maximum of 2 sockets, this will be a problem for anyone
>> on a 4 socket server, even if only 2 are populated.
>>
>> - standard edition RAC now requires 1 socket servers, and as Mark pointed
>> out these are very hard to find as data centre rack servers, you won't get
>> them from the big vendors, HP etc, its only supermicro I know of who supply
>> them.
>>
>> - Capped at 16 threads, but 16 threads on physical cores is actually
>> quite a lot of capacity, eg Intel E5-2667 v3 (3.2 GHz, Max Turbo 3.6 GHz, 8
>> core), considering you can't use multiple RMAN threads etc
>>
>> Andrea,
>> I have not seen a list of chips that qualify for what license,
>> Has anyone else?
>>
>> Tom
>>
>> On Wed, Sep 2, 2015 at 10:21 AM, Andrea Monti <ilsuonogiallo_at_gmail.com>
>> wrote:
>>
>>> Hi ,
>>> in the past I was told by Oracle people that "multi-chip modules" only
>>> refers to some unix-class processors (some IBM Power and SPARC processors).
>>> Do you have any other clues about that?
>>> Did anyone find some evidence to say that *any* x86 processor will not
>>> be considered a multi-chip module?
>>>
>>> Regards
>>> Andrea
>>>
>>>
>>> 2015-09-02 2:45 GMT+02:00 Mark Brinsmead <mark.brinsmead_at_gmail.com>:
>>>
>>>> Yeah. And that "multi-chip module" language is a serious issue too.
>>>>
>>>> Is your CPU a multichip module? How do you know? Its surprisingly
>>>> hard to find out, even if you *do* know the exact model of CPU.
>>>>
>>>> On Tue, Sep 1, 2015 at 1:30 AM, Tom Dale <tom.dale_at_fivium.co.uk> wrote:
>>>>
>>>>> I agree Mark,
>>>>>
>>>>> ​Finding single-socket servers is indeed tricky!​
>>>>>
>>>>> We have bought many servers from broadberry in the UK, they allow us
>>>>> to have SSD's, lsi controllers and 4hr hardware support, at a reasonable
>>>>> cost, they do some single socket servers
>>>>>
>>>>> eg
>>>>>
>>>>>
>>>>> http://www.broadberry.co.uk/superservers-supermicro-servers/as-1012a-m73rf
>>>>>
>>>>> Full spec :
>>>>>
>>>>> http://www.supermicro.com/products/system/1u/1018/sys-1018gr-t.cfm
>>>>>
>>>>> I just configured one
>>>>>
>>>>> E5-2699 v3 Intel 18 Core Xeon 2.3GHz 45Mb Cache 145 Watts
>>>>> 6x 800GB Intel SSD S3500 DataCentre SERIES 2.5IN SATA3 MLC
>>>>> LSI MegaRAID 9380-4i4e 12Gb/s SAS/SATA RAID Controller, 1Gb DDR4
>>>>> Cache, with Internal & External Ports
>>>>> 10GbE Dual-Port SFP+ (Direct Attached) Server Adapter - Intel
>>>>> E10G42BTDA
>>>>> 1st Year 24/7 Support - Up to 4 hours after submission of ticket, up
>>>>> to 5 incidents per year
>>>>>
>>>>> £9,618.81 Ex. VAT
>>>>>
>>>>>
>>>>> The E5-2699 v3 Intel 18 Core is a single chip processor, so my
>>>>> understanding has always been it only needs one license, but I am no
>>>>> licensing expert!
>>>>>
>>>>> From http://www.oracle.com/us/corporate/pricing/sig-070616.pdf
>>>>>
>>>>> When licensing Oracle programs with Standard Edition One or Standard
>>>>> Edition in the product name, a processor is counted equivalent to an
>>>>> occupied socket; however, in the case of multi-chip modules, each chip in
>>>>> the multi-chip module is counted as one occupied socket.
>>>>>
>>>>> But as Alfredo and Svetoslav have said a change in licensing might
>>>>> make this a pointless exercise.
>>>>>
>>>>> Oh well!
>>>>>
>>>>>
>>>>>
>>>>> On Wed, Aug 26, 2015 at 4:33 PM, Mark Brinsmead <
>>>>> mark.brinsmead_at_gmail.com> wrote:
>>>>>
>>>>>> This is true. However, finding single-socket servers can be
>>>>>> something of a challenge.
>>>>>>
>>>>>> The example cited seems to work -- unless that 18-core processor is
>>>>>> implemented as a multi-chip-module ;-) -- but the majority of servers these
>>>>>> days are equipped with at least two sockets. You'll also need to make
>>>>>> certain your single socket server has enough network interfaces.
>>>>>>
>>>>>> I don't know whether Oracle ever removed the language about
>>>>>> multi-chip modules from the OLSA. If they haven't though, it (still) makes
>>>>>> license management with SE very tricky.
>>>>>>
>>>>>>
>>>>>> On Wed, Aug 26, 2015 at 3:05 AM, Tom Dale <tom.dale_at_fivium.co.uk>
>>>>>> wrote:
>>>>>>
>>>>>>> As I read it you can 4 nodes,
>>>>>>>
>>>>>>>
>>>>>>> https://support.oracle.com/epmos/faces/DocumentDisplay?_afrLoop=358910255816631&id=220970.1&_adf.ctrl-state=znnb3zko4_57#A5750
>>>>>>>
>>>>>>> Oracle Database Standard Edition can only be licensed on servers
>>>>>>> that have a maximum capacity of 4 sockets. If licensing by Named User Plus,
>>>>>>> the minimum is 5 Named User Plus licenses. Oracle Database Standard
>>>>>>> Edition, when used with Oracle Real Application Clusters, may only be
>>>>>>> licensed on a single cluster of servers supporting up to a total maximum
>>>>>>> capacity of 4 sockets.
>>>>>>>
>>>>>>> *NOTE: This means that the server capacity must meet the restriction
>>>>>>> even if the sockets are empty, since they count towards capacity.*
>>>>>>>
>>>>>>> So if you have 4 single *socket* servers
>>>>>>>
>>>>>>> eg you could have a single 18 core intel 2600 v3 with 512gb of ram
>>>>>>> in -
>>>>>>>
>>>>>>> http://www.supermicro.com/products/system/1U/1018/SYS-1018R-WC0R.cfm
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On Wed, Aug 26, 2015 at 12:40 AM, Fernando Andrade <
>>>>>>> correo_at_fjandrade.com> wrote:
>>>>>>>
>>>>>>>> Thanks _at_Joe
>>>>>>>> The answer was in the referenced doc. 220970.1
>>>>>>>>
>>>>>>>> On Tue, Aug 25, 2015 at 6:33 PM, Sweetser, Joe <JSweetser_at_icat.com>
>>>>>>>> wrote:
>>>>>>>>
>>>>>>>>> Check out Doc ID 220970.1 on MOS/metalink (old school J ).
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Short answer is yes but there are a few restrictions.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Hth,
>>>>>>>>>
>>>>>>>>> -joe
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> *From:* oracle-l-bounce_at_freelists.org [mailto:
>>>>>>>>> oracle-l-bounce_at_freelists.org] *On Behalf Of *Fernando Andrade
>>>>>>>>> *Sent:* Tuesday, August 25, 2015 5:23 PM
>>>>>>>>> *To:* Oracle-L Group <oracle-l_at_freelists.org>
>>>>>>>>> *Subject:* Extended RAC on SE
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Hi I recieved an extrange request for a client.
>>>>>>>>>
>>>>>>>>> He wants a extended RAC on SE, I have found this reference
>>>>>>>>> googling:
>>>>>>>>>
>>>>>>>>> http://isu.ifmo.ru/docs/doc111/license.111/b28287/editions.htm
>>>>>>>>>
>>>>>>>>> "Extended or stretch clusters are not supported with Standard
>>>>>>>>> Edition and Oracle RAC. An Extended or stretch cluster is defined as "A
>>>>>>>>> cluster where all nodes are not located in the same room"
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> I havent found this paragraph in the documentation provided by
>>>>>>>>> Oracle nor in 11.1, 11.2 or 12.1.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Any one with more information? Thanks
>>>>>>>>>
>>>>>>>>> FJA
>>>>>>>>>
>>>>>>>>> --
>>>>>>>>>
>>>>>>>>> ---------------------------------------
>>>>>>>>> Fernando Jose Andrade
>>>>>>>>> http://www.fjandrade.com
>>>>>>>>> 0983032550
>>>>>>>>> twitter: _fjandrade_
>>>>>>>>> _at_Quito,Ecuador
>>>>>>>>> ---------------------------------------
>>>>>>>>> Confidentiality Note: This message contains information that may
>>>>>>>>> be confidential and/or privileged. If you are not the intended recipient,
>>>>>>>>> you should not use, copy, disclose, distribute or take any action based on
>>>>>>>>> this message. If you have received this message in error, please advise the
>>>>>>>>> sender immediately by reply email and delete this message. Although ICAT,
>>>>>>>>> Underwriters at Lloyd's, Syndicate 4242, scans e-mail and attachments for
>>>>>>>>> viruses, it does not guarantee that either are virus-free and accepts no
>>>>>>>>> liability for any damage sustained as a result of viruses. Thank you.
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>>
>>>>>>>> ---------------------------------------
>>>>>>>> Fernando Jose Andrade
>>>>>>>> http://www.fjandrade.com
>>>>>>>> 0983032550
>>>>>>>> twitter: _fjandrade_
>>>>>>>> _at_Quito,Ecuador
>>>>>>>> ---------------------------------------
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Thu Sep 03 2015 - 01:04:09 CEST

Original text of this message