xen-devel.lists.xenproject.org archive mirror
 help / color / mirror / Atom feed
From: Dan Magenheimer <dan.magenheimer@oracle.com>
To: v.tolstov@selfip.ru
Cc: xen-devel@lists.xensource.com, Daniel Kiper <dkiper@net-space.pl>
Subject: RE: failed to start centos 5 domU with maxmem=30000
Date: Thu, 12 Aug 2010 11:00:47 -0700 (PDT)	[thread overview]
Message-ID: <7aace8e6-6878-40cb-8ed3-e66cf0e90ce1@default> (raw)
In-Reply-To: <1281615845.5454.20.camel@vase.work>

> > > Why this is not provided in documentation or on web site?
> >
> > Hi Vasily --
> >
> > This function limits how far memory can be reduced when
> > ballooning a guest (including dom0).  It is only a heuristic
> > but is intended to take into account the various overheads
> > a guest Linux kernel requires to manage memory to avoid
> > out-of-memory conditions.
> >
> > But I think you are correct... the same (or similar)
> > function should be published as it also serves as a
> > guideline for the ratio between memory= and maxmem=
> > parameters when creating a guest:  If the ratio
> > of maxmem divided by memory is too high, the guest
> > will not even boot.
> 
> Is that possible to use memory=32 and maxmem=60000 ?

I think the answer is no.  I believe there is a kernel
data structure for each 4K page in physical memory
I don't remember the size of this data structure, but
assume it is 4 bytes.  That means that for 1GB of
physical memory (as specified by maxmem) this data
structure requires 1MB of physical memory just to track
the 1GB.  So for your maxmem=60000, memory=60 would
only be enough for this one kernel data structure (and
the kernel requires many other data structures to
be functional).

Daniel Kiper is developing virtual hotplug memory for Xen
guests.  This may be a good use case for it.
 
> > I am curious as to why you would specify memory= so
> > much smaller than maxmem=.  Are you trying to overcommit
> > memory for guests that are often idle but sometimes use
> > a very large amount of memory?
> 
> We want to provide ability to use small as possible memory if the guest
> is idle. And much as possible when the guest under heavy load.

This is a much harder problem than it seems.  You may want
to look at some of the presentations on Transcendent Memory
that explain why it is hard.  (Most can be found at
http://oss.oracle.com/projects/tmem )

Dan

  reply	other threads:[~2010-08-12 18:00 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-08-12 10:24 failed to start centos 5 domU with maxmem=30000 Vasiliy G Tolstov
2010-08-12 11:37 ` Pasi Kärkkäinen
2010-08-12 11:37   ` Vasiliy G Tolstov
2010-08-12 12:17     ` Dan Magenheimer
2010-08-12 12:24       ` Vasiliy G Tolstov
2010-08-12 18:00         ` Dan Magenheimer [this message]
2010-08-13  5:56           ` Vasiliy G Tolstov
2010-08-12 18:34         ` Jeremy Fitzhardinge
2010-08-13  5:54           ` Vasiliy G Tolstov
2010-08-13 15:19             ` Dan Magenheimer

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=7aace8e6-6878-40cb-8ed3-e66cf0e90ce1@default \
    --to=dan.magenheimer@oracle.com \
    --cc=dkiper@net-space.pl \
    --cc=v.tolstov@selfip.ru \
    --cc=xen-devel@lists.xensource.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).