4 GB RAM-based NAS

[Marc de Vries]

You are correct. Marcy does not understand the difference between physical and
virtual memory.

How old are you? You are behaving like a little child.

Should I continue the discussion but start calling you Erica? I won't
lower myself to your standards.

Come back when you have grown up. In the meantime enjoy the killfilter

*** PLONK ***

 

[Shailesh Humbad]

I see the source of the confusion now. When the documentation talks
abou the limit of "2 GB of application space and the 2 GB of kernel
mode space", they are referring to the usage of the physical memory by
the operating system's paging algorithms (in the 4GB-max versions of
Windows), NOT the process's ability to allocate virtual memory. Each
and every process can still allocate up to 4GB of virtual memory,
throughout its 32-bit virtual address space. However, those
application virtual memory pages will be mapped to at most 2 GB of
physical memory at any one time. I actually thought you were wrong
about the Photoshop thrashing at first Marc, but looks like you're
right. It's actually a very interesting limitation of Windows. The
Microsoft documentation has to be parsed carefully to be understood
correctly.

Shailesh

 

[Alexander Grigoriev]

So now you agree that instead of:

you should have written:

What will happen is that windows pro/windows server will use 4GB for
all applications in total. You will have 4GB of unused memory. The 4GB
ram that is used by windows will have to be shared among those 4
photoshop sessions that want 2GB each.

 

[Alexander Grigoriev]

No. "space" means "virtual address space". In Windows, the physical memory
is not pre-allocated for kernel-only or for apps-only. Whatever available of
total 4 GB will be given to whoever needs it. There is no limitation where a
physical page can be mapped - to kernel space (upper 2 GB of virtual
address) or to lower 2 GB of the current user process' virtual space.
Moreover, during I/O, a page can be simultaneously mapped to the user mode
space and to the kernel space, to make it temporarily available to a driver
while an user process can switch to another. It is also possible for a
kernel driver to allocate a memory block and to map it into a current
process' user mode space, so the application can use it, (although this
technique is discouraged).

 

[Marc de Vries]

So now you agree that instead of:


you should have written:

What will happen is that windows pro/windows server will use 4GB for
all applications in total. You will have 4GB of unused memory. The 4GB
ram that is used by windows will have to be shared among those 4
photoshop sessions that want 2GB each.

NO NO NO!!!

What you write contradicts EVERY document from Microsoft!

See the clear explanation from Microsoft that I quoted in the rest of
my reply which you have conveniently not quoted.

The part where I asked you to react to the sentences from Microsoft.

Again Microsoft is very clear about this question:

"Typically, a process running under Windows 2000 or Windows Server
2003 can access up to 2 GB of memory address space (assuming the /3GB
switch was not used) with some of the memory being physical memory and
some being virtual memory. The more programs (and, therefore, more
processes) that run, the more memory you commit up to the full 2 GB of
address space."

What do you not understand in that sentence?
Microsoft clearly states that your applications in total get 2GB
instead of your claim of 4GB.


btw: I'm still waiting for your reply to the questions I had in my
last reply.

Marc

 

[Marc de Vries]

No. "space" means "virtual address space". In Windows, the physical memory
is not pre-allocated for kernel-only or for apps-only. Whatever available of
total 4 GB will be given to whoever needs it. There is no limitation where a
physical page can be mapped - to kernel space (upper 2 GB of virtual
address) or to lower 2 GB of the current user process' virtual space.

I have already supplied several documents from Microsoft that clearly
show that your statement in not correct.

Instead of once again showing that you are wrong, I challange you to
supply documentation from Microsoft that supports your claim above.

And btw you also might try to explain the need for a /3GB switch,
because that wouldn't be needed if windows would work the way you
claim.

 

[Rita Ä Berkowitz]

btw: I'm still waiting for your reply to the questions I had in my

last reply.

I really hope that you aren't seriously expecting to get one from them, let
alone an honest one? You have proven that you are right; there is nothing
to be gained by getting into a pissing contest with these characters. If
you have been on this group for any length of time you will quickly figure
out who the resident idiots are and the many alias they use. This is why I
find it prudent to only offer SCSI solutions and Corncob therapy to them.



Rita

 

[Folkert Rienstra]

RAID 0+1 controllers that outperform the striping/mirroring capabilities
built into the Windows server versions are not cheap--

the cheap ones just use soft mirroring with a boot rom that can bring the
system up far enough for the drivers to load.

As if there is anything else.

 

[Folkert Rienstra]

No, it doesn't work that way.

Each process gets 4GB of VIRTUAL address space. The system then
translates the virtual 32bit addresses to PHYSICAL 32bit addresses
used by the hardware.

With all IA32 processors you are limited to 4GB memory that you can
actually use with the hardware.

That's like saying that a x86 can only use 980kB

 

[Folkert Rienstra]

You think I (as an MS MVP for Windows DDK
http://mvp.support.microsoft.com/default.aspx?scid=fh;en-us;mvpaward#Windows - DDK)
don't know these ABC?

Careful, careful. I acually get the distinct impression that NOT know-
ing anything is a carved in stone condition for becoming a MS MVP.
With the proverbial exception they all toppost and are incapable of
setting up their newsclient properly.

Virtual memory is good, but you may still want to avoid swap thrashing.
This is why you need enough physical memory.

GB,

Well, there you go.

[snip]

 

[Folkert Rienstra]

Right. It used to be called "memory banking".

Or "memory mapping", depending of what (computer) language you speak.

Big packages like Oracle can make effective use if it,

It should actually be the OS that is doing that.

but these apps, the people that need them will
be migrating to 64 bit platforms very quickly.

[snip]

 

[Eric Gisin]

It's obvious you have psychotic delusions. There is no Microsoft document
supporting your view. Alexander is a well recognized expert in the NT kernel,
you are a nobody.

All long time members of this group will remember Ron Reaugh, who had to leave
the group because of this mental disorder. He often referred to documents
supporting his delusions, but obviously did not.

 

[Eric Gisin]

Do you realize you are supporting someone with mental illness, just like Ron?

 

[Eric Gisin]

Thank god you have left the group. Everyone recognizes what a troll you are.
As I said, you don't know what virtual memory is.

 

[Arno Wagner]

All long time members of this group will remember Ron Reaugh, who
had to leave the group because of this mental disorder. He often
referred to documents supporting his delusions, but obviously did
not.

Huh? I thought he just changed his name to Rod Speed? (Might be that
my historic knowledhge is lacking...)

As to documents: There are sound enginering practices. Even MS cannot
go astray too far. Practice show that things stop working when these
practices are ignored. Still a lot of people have this delusion that a
written/printed word is truth. It isn't so. In fact reading a
technical document requires understanding of the subject matter and
thinking about whether the statements in the document can be
true. Even documents describing standards are wrong sometimes. In
addition there might be significant differences in what people claim
they do (in writing) and what they actually do. The informed state of
the reader is non-optional, even witth technical documents.

Arno

 

[Eric Gisin]

There is nothing difficult to understand in the MS KB article. It simply
describes the difference between a 2/3GB process virtual address space. You
have to be pretty ****ed up to think otherwise.