Thanks Matthew / Michael for your responses on this.
On 9/14/2011 2:51 PM, Matthew Seaman wrote:
On 14/09/2011 18:27, Michael Sierchio wrote:
On Wed, Sep 14, 2011 at 6:55 AM, Matthew Seaman
<m.sea...@infracaninophile.co.uk> wrote:
... In these days of plentiful RAM, the new rule of thumb is "if you're
swapping, then you're doing it wrong."
I think your response follows the excellent pedagogical principle: "a
little inaccuracy saves a lot of explanation." But... disk is still
(by far) the cheapest commodity, and the opportunistic paging
algorithm manages VM very well. VM is not by any means obsolete, and
seeing paging behavior is not a sign of a misconfigured system.
Well, yes. I was certainly glossing over a lot of complexity -- but I
would maintain that I am fundamentally correct.
Having some pages swapped out is absolutely not a problem. True. In
fact, it's a positive benefit: swapping out memory pages that are
exceedingly rarely referenced makes more room in RAM for more actively
used pages.
On the other hand, having pages continually swapping in and out
definitely is a problem in terms of performance, given that disk IO
takes of the order of milliseconds, while reference to main RAM is of
the order of microseconds or less. Orders of magnitude faster.
Now, while disk may well be the much the cheapest storage medium
available, that's only part of the expense. In fact, up-front capital
expenditure on the kit (perhaps several thousand pounds/euros/dollars)
is outweighed by the operational expense (power, cooling, hardware
support etc.) over the life of the equipment, so spending a bit more
(capex) on components that run at lower power (opex) makes a lot of
sense. Even more, if the server is being used for eg. e-Commerce, then
the volume of the transactions and the data processed by the server
makes all the difference to your margin: the more you can do with the
same hardware - viz, the more efficiently and faster you can make the
hardware run - then the more profit you make. Buying more RAM is
peanuts on that scale.
Cheers,
Matthew
_______________________________________________
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
