On Wed, Sep 15, 2010 at 09:57:26AM -0400, Richard Pieri wrote:
> On Sep 15, 2010, at 1:44 AM, Derek Martin wrote:
> > 
> > ...where he was clearly talking about virtual memory, and then you
> 
> No, actually, it was not clear

Not to you, perhaps, but in the context of his comment it was the
only thing that made sense.  To randomly exclude the java virtual
machine from a discussion about memory management *in the kernel*
(which is wat Jerry's original comments were about) -- were it had
not been previously mentioned -- makes no sense at all.

> If an application is paging to disk then you have far, far worse
> problems to deal with than memory fragmentation.

Yes, you do... but as was pointed out by me, Jerry, and the article I
linked, this is the only scenario where performance degredation caused
by fragmentation is going to be noticable.  Fragmentation, in and of
itself, isn't going to cause a problem...  Your processes aren't going
to suddently access memory more slowly just because one page is mapped
to a completely different block of physical RAM than the next page is.
Memory access is still just electrons on wires, no matter how
fragmented it is.  It's only when the kernel needs to do disk I/O to
make RAM available that degredation will be noticable.  

[It's possible that a number-crunching application which is both
memory- and cpu-intensive will see a small measurable degredation as
the kernel has to work a little harder to find RAM for it to do its
job, even without paging to disk... but you're not likely to notice
this unless you were already looking very closely -- and these were
not the sort of apps you were talking about running on your servers.]

-- 
Derek D. Martin    http://www.pizzashack.org/   GPG Key ID: 0xDFBEAD02
-=-=-=-=-
This message is posted from an invalid address.  Replying to it will result in
undeliverable mail due to spam prevention.  Sorry for the inconvenience.