OpenVZ Forum: Devel » [RFC][PATCH 0/7] Resource controllers based on process containers

Home » Mailing lists » Devel » [RFC][PATCH 0/7] Resource controllers based on process containers

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: Re: [RFC][PATCH 2/7] RSS controller core [message #11103 is a reply to message #11085]

Tue, 13 March 2007 14:59

Herbert Poetzl
Messages: 239
Registered: February 2006

Senior Member

On Tue, Mar 13, 2007 at 03:48:34AM -0800, Andrew Morton wrote:
> > On Tue, 13 Mar 2007 13:19:53 +0300 Kirill Korotaev <dev@sw.ru> wrote:
> > Andrew Morton wrote:
> > >>>> - shared mappings of 'shared' files (binaries
> > >>>> and libraries) to allow for reduced memory
> > >>>> footprint when N identical guests are running
> > >>>
> > >>>So, it sounds like this can be phrased as a requirement like:
> > >>>
> > >>> "Guests must be able to share pages."
> > >>>
> > >>>Can you give us an idea why this is so?
> > >>
> > >>sure, one reason for this is that guests tend to
> > >>be similar (or almost identical) which results
> > >>in quite a lot of 'shared' libraries and executables
> > >>which would otherwise get cached for each guest and
> > >>would also be mapped for each guest separately
> > >
> > > nooooooo. What you're saying there amounts to text replication.
> > > There is no proposal here to create duplicated copies of pagecache
> > > pages: the VM just doesn't support that (Nick has soe protopatches
> > > which do this as a possible NUMA optimisation).
> > >
> > > So these mmapped pages will contiue to be shared across all
> > > guests. The problem boils down to "which guest(s) get charged for
> > > each shared page".
> > >
> > > A simple and obvious and easy-to-implement answer is "the guest
> > > which paged it in". I think we should firstly explain why that is
> > > insufficient.

> > I guess by "paged it in" you essentially mean
> > "mapped the page into address space for the *first* time"?
>
> Not really - I mean "first allocated the page". ie: major fault(),
> read(), write(), etc.
>
> > i.e. how many times the same page mapped into 2 address spaces
> > in the same container should be accounted for?
> >
> > We believe ONE. It is better due to:
> > - it allows better estimate how much RAM container uses.
> > - if one container mapped a single page 10,000 times,
> > it doesn't mean it is worse than a container which mapped only 200
> > pages and that it should be killed in case of OOM.
>
> I'm not sure that we need to account for pages at all, nor care about
> rss.
>
> If we use a physical zone-based containment scheme: fake-numa,
> variable-sized zones, etc then it all becomes moot.

sounds good to me, just not sure it provides what we
need, but I'm sure I'll figure that with your help ...

> You set up a container which has 1.5GB of physial memory then toss
> processes into it. As that process set increases in size it will
> toss out stray pages which shouldn't be there, then it will start
> reclaiming and swapping out its own pages and eventually it'll get an
> oom-killing.

okay, let me ask a few naive questions about this scheme:

how does this work for a _file_ which is shared between
two guests (e.g. an executable like bash, hardlinked
between guests) when both guests are in a different
zone-based container?

+ assumed that the file is read in the first time,
will it be accounted to the first guest doing so?

+ assumed it is accessed in the second guest, will
it cause any additional cache/mapping besides the
dentry stuff?

+ will container A be able to 'toss out' pages
'shared' with container B (assumed sharing is
possible :)

+ when the container A tosses out the pages for this
executable, will guest B still be able to use them?

+ when the pages are tossed out, will they require
the system to read them in again, or will they
stay available ala swap cache?

> No RSS acounting or page acounting in sight, because we already *have*
> that stuff, at the physical level, in the zone.

I'm fine with that ...

> Overcommitment can be performed by allowing different containers to
> share the same zone set, or by dynamically increasing or decreasing
> the size of a physical container.

here the question is, can a guest have several of
those 'virtual zones' assigned, so that there is a
container specific and a shared zone for example?

> This all works today with fake-numa and cpusets, no kernel changes
> needed.

sounds good!

> It could be made to work fairly simply with a multi-zone approach, or
> with resizeable zones.
>
> I'd be interested in knowing what you think the shortcomings of
> this are likely to be,.

will do so once I have a better understanding how this
approach will work ...

TIA,
Herbert

Report message to a moderator

[Message index]

		[RFC][PATCH 0/7] Resource controllers based on process containers By: xemul on Tue, 06 March 2007 14:42
		[RFC][PATCH 1/7] Resource counters By: xemul on Tue, 06 March 2007 14:47
		Re: [RFC][PATCH 1/7] Resource counters By: Balbir Singh on Wed, 07 March 2007 04:03
		Re: [RFC][PATCH 1/7] Resource counters By: xemul on Wed, 07 March 2007 07:17
		Re: [RFC][PATCH 1/7] Resource counters By: Herbert Poetzl on Fri, 09 March 2007 16:37
		Re: [RFC][PATCH 1/7] Resource counters By: xemul on Sun, 11 March 2007 09:01
		Re: [RFC][PATCH 1/7] Resource counters By: ebiederm on Sun, 11 March 2007 19:00
		Re: [RFC][PATCH 1/7] Resource counters By: Herbert Poetzl on Mon, 12 March 2007 01:16
		Re: [RFC][PATCH 1/7] Resource counters By: ebiederm on Tue, 13 March 2007 09:09
		Re: [RFC][PATCH 1/7] Resource counters By: xemul on Tue, 13 March 2007 09:27
		Re: [RFC][PATCH 1/7] Resource counters By: Herbert Poetzl on Tue, 13 March 2007 15:21
		Re: [RFC][PATCH 1/7] Resource counters By: xemul on Tue, 13 March 2007 15:41
		Re: [RFC][PATCH 1/7] Resource counters By: Srivatsa Vaddagiri on Tue, 13 March 2007 16:07
		Re: [RFC][PATCH 1/7] Resource counters By: xemul on Wed, 14 March 2007 07:12
		Re: [RFC][PATCH 1/7] Resource counters By: ebiederm on Thu, 15 March 2007 16:51
		Re: Re: [RFC][PATCH 1/7] Resource counters By: dev on Tue, 13 March 2007 09:36
		Re: [RFC][PATCH 1/7] Resource counters By: Herbert Poetzl on Tue, 13 March 2007 16:32
		[RFC][PATCH 2/7] RSS controller core By: xemul on Tue, 06 March 2007 14:53
		Re: [RFC][PATCH 2/7] RSS controller core By: Andrew Morton on Tue, 06 March 2007 22:00
		Re: [RFC][PATCH 2/7] RSS controller core By: dev on Sun, 11 March 2007 12:13
		Re: [RFC][PATCH 2/7] RSS controller core By: Andrew Morton on Sun, 11 March 2007 12:51
		Re: [RFC][PATCH 2/7] RSS controller core By: Balbir Singh on Sun, 11 March 2007 15:51
		Re: [RFC][PATCH 2/7] RSS controller core By: ebiederm on Sun, 11 March 2007 19:34
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Mon, 12 March 2007 01:00
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Mon, 12 March 2007 18:42
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Mon, 12 March 2007 22:41
		Re: [RFC][PATCH 2/7] RSS controller core By: xemul on Mon, 12 March 2007 09:02
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Mon, 12 March 2007 21:11
		Re: [RFC][PATCH 2/7] RSS controller core By: xemul on Tue, 13 March 2007 07:17
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Tue, 13 March 2007 15:05
		Re: [RFC][PATCH 2/7] RSS controller core By: xemul on Tue, 13 March 2007 15:32
		Re: [RFC][PATCH 2/7] RSS controller core By: dev on Tue, 13 March 2007 15:10
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: Kirill Korotaev on Mon, 12 March 2007 09:10
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: ebiederm on Tue, 13 March 2007 09:26
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: dev on Tue, 13 March 2007 15:30
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: dev on Tue, 13 March 2007 10:06
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: Andrew Morton on Tue, 13 March 2007 10:49
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Tue, 13 March 2007 14:59
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Tue, 13 March 2007 17:05
		Re: [RFC][PATCH 2/7] RSS controller core By: mel on Wed, 14 March 2007 15:38
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: Paul Menage on Sun, 18 March 2007 22:44
		Re: Re: [RFC][PATCH 2/7] RSS controller core By: ebiederm on Mon, 19 March 2007 17:41
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Fri, 09 March 2007 16:48
		Re: [RFC][PATCH 2/7] RSS controller core By: xemul on Sun, 11 March 2007 09:08
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Sun, 11 March 2007 14:32
		Re: [RFC][PATCH 2/7] RSS controller core By: xemul on Sun, 11 March 2007 15:04
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Mon, 12 March 2007 00:41
		Re: [RFC][PATCH 2/7] RSS controller core By: xemul on Mon, 12 March 2007 08:31
		Re: [RFC][PATCH 2/7] RSS controller core By: Balbir Singh on Mon, 12 March 2007 09:55
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Mon, 12 March 2007 23:43
		Re: [RFC][PATCH 2/7] RSS controller core By: Balbir Singh on Tue, 13 March 2007 01:57
		Re: [RFC][PATCH 2/7] RSS controller core By: Srivatsa Vaddagiri on Tue, 13 March 2007 02:24
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Tue, 13 March 2007 16:06
		Re: [RFC][PATCH 2/7] RSS controller core By: Balbir Singh on Wed, 07 March 2007 05:37
		Re: [RFC][PATCH 2/7] RSS controller core By: xemul on Wed, 07 March 2007 07:25
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Mon, 12 March 2007 23:02
		Re: [RFC][PATCH 2/7] RSS controller core By: ebiederm on Sun, 18 March 2007 16:58
		Re: [RFC][PATCH 2/7] RSS controller core By: akpm on Tue, 13 March 2007 06:04
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Tue, 13 March 2007 17:26
		Re: [RFC][PATCH 2/7] RSS controller core By: Alan Cox on Tue, 13 March 2007 19:09
		Re: [RFC][PATCH 2/7] RSS controller core By: ebiederm on Fri, 16 March 2007 00:55
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Fri, 16 March 2007 16:31
		Re: [RFC][PATCH 2/7] RSS controller core By: ebiederm on Fri, 16 March 2007 18:54
		Re: [RFC][PATCH 2/7] RSS controller core By: mel on Wed, 14 March 2007 16:47
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Tue, 13 March 2007 20:28
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Tue, 13 March 2007 15:11
		Re: [RFC][PATCH 2/7] RSS controller core By: dev on Tue, 13 March 2007 15:54
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Wed, 14 March 2007 20:42
		Re: [RFC][PATCH 2/7] RSS controller core By: mel on Tue, 20 March 2007 18:57
		Re: [RFC][PATCH 2/7] RSS controller core By: Dave Hansen on Fri, 16 March 2007 19:46
		Re: [RFC][PATCH 2/7] RSS controller core By: ebiederm on Sun, 18 March 2007 17:42
		Re: [RFC][PATCH 2/7] RSS controller core By: Herbert Poetzl on Mon, 19 March 2007 15:48
		controlling mmap()'d vs read/write() pages By: Dave Hansen on Tue, 20 March 2007 16:15
		Re: controlling mmap()'d vs read/write() pages By: ebiederm on Tue, 20 March 2007 21:19
		Re: controlling mmap()'d vs read/write() pages By: Herbert Poetzl on Fri, 23 March 2007 00:51
		Re: controlling mmap()'d vs read/write() pages By: Nick Piggin on Fri, 23 March 2007 05:57
		Re: controlling mmap()'d vs read/write() pages By: ebiederm on Fri, 23 March 2007 10:12
		Re: controlling mmap()'d vs read/write() pages By: Dave Hansen on Fri, 23 March 2007 16:41
		Re: controlling mmap()'d vs read/write() pages By: Herbert Poetzl on Fri, 23 March 2007 18:16
		Re: controlling mmap()'d vs read/write() pages By: Balbir Singh on Wed, 28 March 2007 09:18
		Re: controlling mmap()'d vs read/write() pages By: Nick Piggin on Fri, 23 March 2007 10:47
		Re: controlling mmap()'d vs read/write() pages By: ebiederm on Fri, 23 March 2007 12:21
		Re: controlling mmap()'d vs read/write() pages By: Nick Piggin on Wed, 28 March 2007 07:33
		[RFC][PATCH 3/7] Data structures changes for RSS accounting By: xemul on Tue, 06 March 2007 14:55
		Re: [RFC][PATCH 3/7] Data structures changes for RSS accounting By: ebiederm on Sun, 11 March 2007 19:13
		Re: [RFC][PATCH 3/7] Data structures changes for RSS accounting By: dev on Mon, 12 March 2007 16:16
		Re: [RFC][PATCH 3/7] Data structures changes for RSS accounting By: Dave Hansen on Mon, 12 March 2007 16:48
		Re: [RFC][PATCH 3/7] Data structures changes for RSS accounting By: xemul on Mon, 12 March 2007 17:19
		Re: [RFC][PATCH 3/7] Data structures changes for RSS accounting By: Balbir Singh on Mon, 12 March 2007 17:21
		Re: [RFC][PATCH 3/7] Data structures changes for RSS accounting By: Dave Hansen on Mon, 12 March 2007 17:27
		Re: [RFC][PATCH 3/7] Data structures changes for RSS accounting By: xemul on Tue, 13 March 2007 07:10
		[RFC][PATCH 4/7] RSS accounting hooks over the code By: xemul on Tue, 06 March 2007 14:57
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: xemul on Wed, 14 March 2007 15:43
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Cedric Le Goater on Wed, 14 March 2007 15:37
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: ebiederm on Sun, 11 March 2007 19:14
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: dev on Mon, 12 March 2007 16:23
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Dave Hansen on Mon, 12 March 2007 16:50
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: dev on Mon, 12 March 2007 17:07
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Herbert Poetzl on Mon, 12 March 2007 23:54
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: ebiederm on Tue, 13 March 2007 09:58
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Nick Piggin on Tue, 13 March 2007 10:25
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Ethan Solomita on Wed, 28 March 2007 20:15
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Dave Hansen on Mon, 12 March 2007 17:33
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: ebiederm on Tue, 13 March 2007 09:43
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: ebiederm on Tue, 13 March 2007 16:01
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Nick Piggin on Wed, 14 March 2007 03:51
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Balbir Singh on Wed, 14 March 2007 06:42
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Nick Piggin on Wed, 14 March 2007 06:57
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Balbir Singh on Wed, 14 March 2007 07:48
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Vaidyanathan Srinivas on Wed, 14 March 2007 13:25
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Nick Piggin on Wed, 14 March 2007 13:49
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: dev on Wed, 14 March 2007 16:16
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Nick Piggin on Thu, 15 March 2007 05:01
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Vaidyanathan Srinivas on Wed, 14 March 2007 14:43
		Re: [RFC][PATCH 4/7] RSS accounting hooks over the code By: Balbir Singh on Thu, 15 March 2007 05:44
		[RFC][PATCH 5/7] Per-container OOM killer and page reclamation By: xemul on Tue, 06 March 2007 15:01
		Re: [RFC][PATCH 5/7] Per-container OOM killer and page reclamation By: Balbir Singh on Fri, 09 March 2007 21:21
		Re: [RFC][PATCH 5/7] Per-container OOM killer and page reclamation By: xemul on Sun, 11 March 2007 08:39
		[RFC][PATCH 6/7] Account for the number of tasks within container By: xemul on Tue, 06 March 2007 15:02
		Re: [RFC][PATCH 6/7] Account for the number of tasks within container By: Paul Menage on Wed, 07 March 2007 02:00
		Re: [RFC][PATCH 6/7] Account for the number of tasks within container By: xemul on Wed, 07 March 2007 07:10
		Re: [RFC][PATCH 6/7] Account for the number of tasks within container By: Paul Menage on Thu, 08 March 2007 13:49
		Re: [RFC][PATCH 6/7] Account for the number of tasks within container By: xemul on Sun, 11 March 2007 08:34
		[RFC][PATCH 7/7] Account for the number of files opened within container By: xemul on Tue, 06 March 2007 15:05
		Re: [RFC][PATCH 0/7] Resource controllers based on process containers By: Paul Menage on Wed, 07 March 2007 02:02
		Re: [RFC][PATCH 0/7] Resource controllers based on process containers By: xemul on Wed, 07 March 2007 07:27
		Re: [RFC][PATCH 0/7] Resource controllers based on process containers By: Balbir Singh on Wed, 07 March 2007 06:52
		Re: [RFC][PATCH 0/7] Resource controllers based on process containers By: xemul on Wed, 07 March 2007 07:30
		Re: [RFC][PATCH 0/7] Resource controllers based on process containers By: dev on Wed, 07 March 2007 09:30

Previous Topic:	Re: [ckrm-tech] [PATCH 7/7] containers (V7): Container interface to nsproxy subsystem
Next Topic:	Linux-VServer example results for sharing vs. separate mappings ...

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sun Dec 08 01:06:32 GMT 2024

Total time taken to generate the page: 0.02825 seconds