Vista - Virtual PC - RAM

When you setup a guest OS, you need to allocate RAM to it.

Let's say you have a host with 3 GB of RAM with two guests, each of
which you allocate 1 GB of RAM.

If both guests are in use, the host and two guests will get 1 GB of
RAM each.

Is there a way to set it so that if only 1 guest is in use, it can
draw on the spare RAM that is allocated to the other guest that is not
used.

On Thu, 21 May 2009 19:34:16 -0700 (PDT), kellyapproved@xxxxxx
wrote:
RAM is assigned statically. If you give your VM 1GB that is what it
will get, no more, no less. If there is not enough available, the VM
won't start.

--
Cheers,
Steve Jain, Virtual Machine MVP
http://vpc.essjae.com/

kellyapproved@xxxxxx wrote:
If one VM could monitor the memory usage of another VM to see what
blocks were unallocated, you just threw out the isolation that
virtualization is supposed to afford for security.

<kellyapproved@xxxxxx> wrote in message
news:254ab120-ea96-4a9c-87f4-e8456db46402@xxxxxx
No.


--
Mark Rae
ASP.NET MVP
http://www.markrae.net

kellyapproved@xxxxxx wrote:Unfortunately not. Remember that to the operating system inside the
Guest PC, the virtual PC is a real PC and the amount of memory in a real
PC typically cannot change without a reboot.

Currently I don't think the following facilities exist in MS VPC, but
they would be nice to have for your (common) problem:

1. Allow the virtual RAM given to a virtual PC to come from the Host PCs
virtual memory, not physical memory (in other words allow it to be
swapped out to the regular pagefile.sys of the host OS, just like
ordinary application memory consumption). This would allow running more
and larger virtual PCs on the same hardware, at the cost of some of
those virtual PCs running slower.

2. Adapt the amount of virtual RAM given to a virtual PC at virtual PC
reboot to depend on available real RAM at that time.

3. In the Virtual PC extensions for various operating systems, include
code that releases unused RAM pages back to the Host PC, thus making the
memory requirement smaller when the Guest OS is idle.

4. Emulate the advanced support for physically adding and removing
hardware memory without a reboot as found in expensive high end servers.
Many recent operating systems such as Server 2008 (and thus in
principle Vista too) contain support for handling hot memory
plugging/unplugging and could thus be made to cooperate with a variable
RAM size Virtual PC.
For Microsoft's own operating systems they could make a license
exception allowing the hot plug support to be used in Virtual PC (or
equivalent) Guest environments, even in SKUs that are not otherwise
licensed to include support for such high end servers.


--
Jakob Bøhm, M.Sc.Eng. * jb@xxxxxx * direct tel:+45-45-90-25-33
Netop Solutions A/S * Bregnerodvej 127 * DK-3460 Birkerod * DENMARK
http://www.netop.com * tel:+45-45-90-25-25 * fax:+45-45-90-25-26
Information in this mail is hasty, not binding and may not be right.
Information in this posting may not be the official position of Netop
Solutions A/S, only the personal opinions of the author.

In message <gv5748$usq$1@xxxxxx-september.org> VanguardLH
<V@xxxxxx> was claimed to have wrote:
With all due respect, that isn't true at all. VMWare allows RAM to be
over-allocated, which would be the first step to implementing this
request.

To avoid the performance implications of over-utilizing RAM, a second
step would be to have the integration components/additions/tools report
back the amount of available RAM within each VM to the host, and the
host would reply back telling the additions to allocate or release RAM
within each guest, tricking the guest into not using certain amounts of
RAM to make that RAM available to the host (and therefore, for
allocation to other VMs)

Guests already communicate with hosts in a number of ways, adding a
counters for available RAM, available clean pages, cache sizes and an
API to the guest components wouldn't cause any sort of security issue if
implemented properly.

Dave Warren wrote:
http://www.vmware.com/pdf/esx3_memory.pdf

It does talk about overcommitting the memory (i.e., having multiple
guests with a reserve specified for each whose total exceeds the real
memory for the host). It also mentions reallocation -- and that's where
they also mention clearing the reallocated memory to provide the
security of which I mention between those guests.

Allocating memory doesn't wipe it. Deallocating memory doesn't wipe it.
Whose going to do the wiping? In this case, the VMM has to do the
wiping and that means overhead. Their ESX product does that. Does
their free VMWare Server? I didn't see mention in VMWare Server's
manual that its VMM will reallocate and wipe memory between guests
(http://www.vmware.com/pdf/vmserver2.pdf). I'm still reading but
reserve, overcommittment, and swapping are not reallocating wiped memory
from a pool to share amongst guests. Technologies employed in a
rebranded Connectrix variant that targets consumer-grade platforms
hardly equates with VMware's enterprise solutions for server-quality
hosts.

If anything like VMware's ESX memory sharing shows up in a Microsoft
product, I'd expect it to show up in their Hyper-V product, not in
something like VirtualPC. That's not to say Hyper-V has wiped memory
reallocation for sharing a memory pool across guests. I'm saying it's
no big surprise that VPC doesn't have it. Can reallocation between
guests be done? Yes. WILL it be done in VirtualPC? No, not without
such a massive overhaul that it isn't VirtualPC anymore, becomes an
enterprise-level product, and probably won't be free so we wouldn't be
discussing it here.

VanguardLH wrote:Wiping memory before allocating it to another process (in a single OS)
or another virtual machine (in a VMM) is a standard and trivial security
requirement for physical memory management code. All the NT-derived
Windows versions (including consumer-grade XP Home) do it for processes.
There is also a really good chance that at least the Microsoft "cleaned
up" versions of the old Connectix VPC co debase do it already for the
initial RAM allocation made for a newly started VPC.
I think you are greatly confused here. Just because VMware Inc may be
offering this feature in their enterprise product does not make it an
enterprise feature. On the contrary, the ability to run virtual
machines whose total virtual RAM exceed you physical RAM is very much a
desktop feature too. A really common case would be running a virtual
Vista and a virtual XP on a Win7 laptop whose physical RAM size is just
big enough to run one of the 3, and cannot be easily/cheaply upgraded.

While the VMware ESX micro operating system (it runs on the bare metal
with no normal Host OS) may use impressive names such as
"overcommitting" memory; But when running on top of a full featured
Windows OS, this all translates into simply allowing the virtual RAM to
be allocated from virtual memory rather than physical memory, and then
*optionally* adding some extra code to reduce the amount of swapping by
forcing the Guest OS-es to leave spare memory unused rather than using
it to "cache" disk I/O.


--
Jakob Bøhm, M.Sc.Eng. * jb@xxxxxx * direct tel:+45-45-90-25-33
Netop Solutions A/S * Bregnerodvej 127 * DK-3460 Birkerod * DENMARK
http://www.netop.com * tel:+45-45-90-25-25 * fax:+45-45-90-25-26
Information in this mail is hasty, not binding and may not be right.
Information in this posting may not be the official position of Netop
Solutions A/S, only the personal opinions of the author.

In message <ev5QfGS3JHA.4880@xxxxxx> Jakob Bohm
<jb@xxxxxx> was claimed to have wrote:
VMWare Workstation also allows overcommiting of memory, so this isn't an
enterprise-only feature.