Re: Resource Management on General-Purpose Systems

[olafBuddenhagen]

Hi,

On Thu, Jul 12, 2007 at 04:41:59PM +0200, Neal H. Walfield wrote:

> Given the increasing amount of resources available, Tanenbaum, in a
> private discussion, argued that paging of anonymous memory is no
> longer interesting.

Well, I made a similar observation: It seems that over the past few
years, the amount of RAM available in typical PCs has outgrown the
demand for anonymous memory of virtually all applications.

Nevertheless, I don't agree with his conclusion: While anonymous memory
might easily fit into the RAM, the amount of disk data that is processed
and needs to be cached efficiently for good performance, still by far
outgrows available RAM sizes in many scenarios, and very likely always
will. We still need the mechanisms and algorithms for efficient paging.
I doubt wiring all anonymous memory really helps anything.

Also, while it seems that the rate between available and demanded memory
has improved over the past few years (at least for PCs; not sure about
servers), thus making efficient disk paging less crucial, memory
management becomes more prominent in other places: The local memory each
graphics card has is being used in a very dynamic fashion nowadays,
increasing the importance of a good video memory manager. And there
seems to be a general trend towards such non-uniform memory
architectures: The Cell processor for example has a small local memory
for each of the synergetic processing units.

> Third, some devices have less, not more, resources.  OLPC is a
> particularly good example.  Linksys Routers, in which saving a few
> megabytes of flash are enough to motivate an entire rewrite of the
> system [1], and hand held PCs such as the Nokia 770 and N800 also fall
> into this category.

None of these devices has anywhere to page anonymous memory -- so this
hardly disproves Andy's claim...

-antrik-