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

Home -> Community -> Usenet -> c.d.o.server -> Re: Oracle Linux/Virtualization Software

Re: Oracle Linux/Virtualization Software

From: Robert Klemme <shortcutter_at_googlemail.com>
Date: Fri, 30 Nov 2007 18:34:41 +0100
Message-ID: <5ravtjF13vs1tU1@mid.individual.net>


On 30.11.2007 18:04, shakespeare wrote:

> <shortcutter_at_googlemail.com> schreef in bericht 
> news:25895977-106a-4a6e-914a-23e1a89cb456_at_l16g2000hsf.googlegroups.com...

>> On 30 Nov., 14:53, hpuxrac <johnbhur..._at_sbcglobal.net> wrote:
>>> On Nov 29, 9:01 pm, DA Morgan <damor..._at_psoug.org> wrote:
>>>
>>> snip
>>>
>>>> I'm not counseling blind acceptance. I'm not advising anyone to take
>>>> anything at face value. Isn't it Tom Kyte who has advised over and over
>>>> again the importance of not accepting claims made by self-anointed
>>>> experts?
>>> Well actually I think it starts back with the Greeks and Romans as far
>>> as indo-european history goes.
>>>
>>>
>>>
>>>> What is at issue here is the difference between scepticism (healthy)
>>>> and
>>>> cynicism (attitude). I am not saying Oracle is correct and I'm not
>>>> saying Oracle is incorrect. Because I haven't tested the claim in my
>>>> lab
>>>> and don't know of any reputable independent source that has.
>>>> What I object to is this cynical repudiation of a claim based solely
>>>> on the person's attitude of unblemished negativity. There isn't one
>>>> person here that has weighed in with a value judgment that any single
>>>> data point, even a poorly created one, that supports their statement.
>>> You know this blind advocacy of anything oracle is getting pretty old
>>> here in cdos from certain posters. As far as I know my original post
>>> in this thread concerning VM said this ...
>>>
>>> "Oracle claims their VM is 3 times as fast as anything else. Sounds
>>> pretty fishy to many of us."
>>>
>>> Maybe you should read a definition of "fishy"?
>>>
>>> All in all based on the hype from Larry and others in OOW 2007 without
>>> any benchmarks provided, tangible testing results ... fishy seems to
>>> be right on target to me. It might even be optimistic?
>> Things become a bit clearer when looking at the FAQ:
>>
>> "How is Oracle VM three times more efficient than existing x86 server
>> virtualization products?
>> Oracle ran many performance benchmarks comparing Oracle products
>> running with Oracle VM against the existing leading server
>> virtualization product and also with Oracle products on non-
>> virtualized operating systems on x86 and x86-64. Oracle consistently
>> saw much better resource
>> utilization with an average of three times less overhead using Oracle
>> VM, and also saw significant
>> scalability with virtual SMP. In many cases, the comparison with real
>> hardware was approximately
>> equal in performance."
>>
>> http://www.oracle.com/technologies/virtualization/docs/ovm-faq.pdf
>>
>> I interpret "three times more efficient" as meaning "the overhead of
>> other VM products vs. real hardware was cut down to a quarter". So in
>> cases of low overhead, the net gain is small and the claim sounds far
>> less dramatic.
>>
>> The crucial questions here are of course: what is X? What hardware
>> was it tested on? What applications were benchmarked?
>>
>> Kind regards
>>
>> robert
> 
> So to make a calculation (just as an example, no exact figures...)
> 
> If VMware has an overhead of 15 % (leaving 85% efficiency) Oracle claims 
> about 4% overhead (96% efficiency), giving an net  efficiency gain of about 
> 11%.... which is different from 'three times as fast'.....

Which they have not claimed as far as I can see. Although of course this is what they intended the reader to believe. Watch out for marketing speak! :-)

> I think 'overhead' is relative, depending on how many vm's are running on 1 > host.

I think first and foremost "overhead" is a relative term in itself. Assuming that the VMWares of the world are already pretty efficient by using hardware virtualization and their vast experience in the area there is not much to save by cutting this small overhead.

I wonder whether they compared it to VMWare ESX. If they did not then their average efficiency gain might be due to stripping down an underlying Linux (i.e. removing all services that are not needed etc.) - from what I know this is the same approach ESX uses, they claim to install on bare metal (which they do in a way) but under the hoods it seems to be a customized Linux with their virtualization software embedded.

But I'm wandering off. 'Nuff for today

        robert Received on Fri Nov 30 2007 - 11:34:41 CST

Original text of this message

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