Xen devel - Jun 2010 - user-land tool for memory paging and memory sharing

Is there any documentation on these features.

Thanks,

K. Y


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

+1... and in fact, my earlier question was never answered
so let me ask it more bluntly...

Do these features work reliably today on 4.0.x, e.g. well
enough for an enterprise environment rather than just
as a demo?  Or are there known features/corner cases
that are missing today (that the discussion on mem-event
is perhaps intended eventually to rectify)?
> -----Original Message-----
> From: Ky Srinivasan [mailto:ksrinivasan@novell.com]
> Sent: Tuesday, June 29, 2010 10:58 AM
> To: xen-devel@lists.xensource.com
> Subject: [Xen-devel] user-land tool for memory paging and memory
> sharing
> 
> Is there any documentation on these features.
> 
> Thanks,
> 
> K. Y
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

I''m not aware of any formal documentation on either of these things.

I would imagine that there are still some issues with both sharing and
paging that may not make it suitable for enterprise use. Of course, if
anybody would like to stress test them in "real-world" scenarios and
report any problems encountered (or patches), that would be greatly
appreciated. Certainly with the upcoming changes to the mem-event
interface, there are likely to be new bugs/issues introduced as well,
so it would seem reasonable to not want to rely too much on either
feature until that is a bit more solidified.


Patrick


On 29 June 2010 10:11, Dan Magenheimer <dan.magenheimer@oracle.com>
wrote:
> +1... and in fact, my earlier question was never answered
> so let me ask it more bluntly...
>
> Do these features work reliably today on 4.0.x, e.g. well
> enough for an enterprise environment rather than just
> as a demo?  Or are there known features/corner cases
> that are missing today (that the discussion on mem-event
> is perhaps intended eventually to rectify)?
>
>> -----Original Message-----
>> From: Ky Srinivasan [mailto:ksrinivasan@novell.com]
>> Sent: Tuesday, June 29, 2010 10:58 AM
>> To: xen-devel@lists.xensource.com
>> Subject: [Xen-devel] user-land tool for memory paging and memory
>> sharing
>>
>> Is there any documentation on these features.
>>
>> Thanks,
>>
>> K. Y
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xensource.com
>> http://lists.xensource.com/xen-devel
>
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

>>> On 6/29/2010 at  1:37 PM, in message
<AANLkTikTMxl8PncfXcOGKBphVvGW6_lNhdP1M_qPUsjX@mail.gmail.com>, Patrick
Colp
<pjcolp@cs.ubc.ca> wrote: 
> I''m not aware of any formal documentation on either of these
things.
> 
> I would imagine that there are still some issues with both sharing and
> paging that may not make it suitable for enterprise use. Of course, if
> anybody would like to stress test them in "real-world" scenarios
and
> report any problems encountered (or patches), that would be greatly
> appreciated. Certainly with the upcoming changes to the mem-event
> interface, there are likely to be new bugs/issues introduced as well,
> so it would seem reasonable to not want to rely too much on either
> feature until that is a bit more solidified.
Thanks Patric. I am interested in testing these features and I was looking for
some documentation to get me started here. Is there a known list of
problems/shortcomings in the code as it stands today. Is this feature only for
HVM guests? In your tests, what levels of memory over-commitment were you able
to achieve.

Regards,

K. Y
> 
> 
> On 29 June 2010 10:11, Dan Magenheimer <dan.magenheimer@oracle.com>
wrote:
>> +1... and in fact, my earlier question was never answered
>> so let me ask it more bluntly...
>>
>> Do these features work reliably today on 4.0.x, e.g. well
>> enough for an enterprise environment rather than just
>> as a demo?  Or are there known features/corner cases
>> that are missing today (that the discussion on mem-event
>> is perhaps intended eventually to rectify)?
>>
>>> -----Original Message-----
>>> From: Ky Srinivasan [mailto:ksrinivasan@novell.com]
>>> Sent: Tuesday, June 29, 2010 10:58 AM
>>> To: xen-devel@lists.xensource.com
>>> Subject: [Xen-devel] user-land tool for memory paging and memory
>>> sharing
>>>
>>> Is there any documentation on these features.
>>>
>>> Thanks,
>>>
>>> K. Y
>>>
>>>
>>> _______________________________________________
>>> Xen-devel mailing list
>>> Xen-devel@lists.xensource.com
>>> http://lists.xensource.com/xen-devel
>>
>>


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

I''ve never tried the sharing code, so I can''t speak for it.
For
swapping, it''s only for HVM guests, however as far as I know, you can
currently only page out one guest at a time (this is clearly something
that we''d like to fix, although this is likely to happen after the new
mem-event interface is established). In theory with swapping, you can
remove most of the guest''s allocated memory, however the more pages
that are swapped out, the slower it runs. Grzegorz should be able to
provide more information about sharing.


Patrick


On 29 June 2010 11:34, Ky Srinivasan <ksrinivasan@novell.com>
wrote:
>
>
>>>> On 6/29/2010 at  1:37 PM, in message
> <AANLkTikTMxl8PncfXcOGKBphVvGW6_lNhdP1M_qPUsjX@mail.gmail.com>,
Patrick Colp
> <pjcolp@cs.ubc.ca> wrote:
>> I''m not aware of any formal documentation on either of these
things.
>>
>> I would imagine that there are still some issues with both sharing and
>> paging that may not make it suitable for enterprise use. Of course, if
>> anybody would like to stress test them in "real-world"
scenarios and
>> report any problems encountered (or patches), that would be greatly
>> appreciated. Certainly with the upcoming changes to the mem-event
>> interface, there are likely to be new bugs/issues introduced as well,
>> so it would seem reasonable to not want to rely too much on either
>> feature until that is a bit more solidified.
>
> Thanks Patric. I am interested in testing these features and I was looking
for some documentation to get me started here. Is there a known list of
problems/shortcomings in the code as it stands today. Is this feature only for
HVM guests? In your tests, what levels of memory over-commitment were you able
to achieve.
>
> Regards,
>
> K. Y
>>
>>
>> On 29 June 2010 10:11, Dan Magenheimer
<dan.magenheimer@oracle.com> wrote:
>>> +1... and in fact, my earlier question was never answered
>>> so let me ask it more bluntly...
>>>
>>> Do these features work reliably today on 4.0.x, e.g. well
>>> enough for an enterprise environment rather than just
>>> as a demo?  Or are there known features/corner cases
>>> that are missing today (that the discussion on mem-event
>>> is perhaps intended eventually to rectify)?
>>>
>>>> -----Original Message-----
>>>> From: Ky Srinivasan [mailto:ksrinivasan@novell.com]
>>>> Sent: Tuesday, June 29, 2010 10:58 AM
>>>> To: xen-devel@lists.xensource.com
>>>> Subject: [Xen-devel] user-land tool for memory paging and
memory
>>>> sharing
>>>>
>>>> Is there any documentation on these features.
>>>>
>>>> Thanks,
>>>>
>>>> K. Y
>>>>
>>>>
>>>> _______________________________________________
>>>> Xen-devel mailing list
>>>> Xen-devel@lists.xensource.com
>>>> http://lists.xensource.com/xen-devel
>>>
>>>
>
>
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Tue, Jun 29, 2010 at 11:39:15AM -0700, Patrick Colp
wrote:
> I''ve never tried the sharing code, so I can''t speak for
it. For
> swapping, it''s only for HVM guests, however as far as I know, you
can
> currently only page out one guest at a time (this is clearly something
> that we''d like to fix, although this is likely to happen after the
new
> mem-event interface is established). In theory with swapping, you can
> remove most of the guest''s allocated memory, however the more
pages
> that are swapped out, the slower it runs. Grzegorz should be able to
> provide more information about sharing.
> 
Does this thread help?:
http://lists.xensource.com/archives/html/xen-devel/2010-04/msg00373.html

-- Pasi
> 
> Patrick
> 
> 
> On 29 June 2010 11:34, Ky Srinivasan <ksrinivasan@novell.com> wrote:
> >
> >
> >>>> On 6/29/2010 at  1:37 PM, in message
> > <AANLkTikTMxl8PncfXcOGKBphVvGW6_lNhdP1M_qPUsjX@mail.gmail.com>,
Patrick Colp
> > <pjcolp@cs.ubc.ca> wrote:
> >> I''m not aware of any formal documentation on either of
these things.
> >>
> >> I would imagine that there are still some issues with both sharing
and
> >> paging that may not make it suitable for enterprise use. Of
course, if
> >> anybody would like to stress test them in "real-world"
scenarios and
> >> report any problems encountered (or patches), that would be
greatly
> >> appreciated. Certainly with the upcoming changes to the mem-event
> >> interface, there are likely to be new bugs/issues introduced as
well,
> >> so it would seem reasonable to not want to rely too much on either
> >> feature until that is a bit more solidified.
> >
> > Thanks Patric. I am interested in testing these features and I was
looking for some documentation to get me started here. Is there a known list of
problems/shortcomings in the code as it stands today. Is this feature only for
HVM guests? In your tests, what levels of memory over-commitment were you able
to achieve.
> >
> > Regards,
> >
> > K. Y
> >>
> >>
> >> On 29 June 2010 10:11, Dan Magenheimer
<dan.magenheimer@oracle.com> wrote:
> >>> +1... and in fact, my earlier question was never answered
> >>> so let me ask it more bluntly...
> >>>
> >>> Do these features work reliably today on 4.0.x, e.g. well
> >>> enough for an enterprise environment rather than just
> >>> as a demo?  Or are there known features/corner cases
> >>> that are missing today (that the discussion on mem-event
> >>> is perhaps intended eventually to rectify)?
> >>>
> >>>> -----Original Message-----
> >>>> From: Ky Srinivasan [mailto:ksrinivasan@novell.com]
> >>>> Sent: Tuesday, June 29, 2010 10:58 AM
> >>>> To: xen-devel@lists.xensource.com
> >>>> Subject: [Xen-devel] user-land tool for memory paging and
memory
> >>>> sharing
> >>>>
> >>>> Is there any documentation on these features.
> >>>>
> >>>> Thanks,
> >>>>
> >>>> K. Y
> >>>>
> >>>>
> >>>> _______________________________________________
> >>>> Xen-devel mailing list
> >>>> Xen-devel@lists.xensource.com
> >>>> http://lists.xensource.com/xen-devel
> >>>
> >>>
> >
> >
> >
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel