[net.arch] Real Reasons For Large Main Memories

waynekn@tekig5.UUCP (Wayne Knapp) (09/09/86)

    Just because you have a large memory doesn't mean you have to use all
of it for running progams.  How about:

    1. Move your file system in to memory.  ( Big bennifits.)
       ( Just back up the memory image every so often )

    2. Increasing your working memory space as the work load increases.
       ( Hey it is stupid design to make the operation system worry about
         9999 Gbytes when you only need 2 Mbytes. )

    3. Huge neat programs and experiments. Virtual memory is a drag!  I want
       progams to fly.  Give me 1000 Gbytes of linear space for my program.
       I have options to check things to do, can't wait for slow IO. 

    4. Storing full length movies in Ram, and view by page flipping.

    5. Hiding stuff from system cops.

    6. Wow! I could use some really big lookup tables to speed up 
       calculations.

    7. Recursion anyone?  How about recursive compilers, hmm makes me wonder.

    * many more things I haven't even dreamed of yet!

    Okay so page tables are a little bigger, big deal you have room to spare.
It is your own problem if you think you have to maintain tables for the whole
memory.  Programs are larger, but who cares.  Bring on the big memories, I
for one can make use of them!

                        Wayne Knapp       

P.S.  The only problem I see is ram testing.  Maybe a second processor  
would be useful.  There should be at least two processors in every pot anyway.

henry@utzoo.UUCP (Henry Spencer) (09/11/86)

>   1. Move your file system in to memory.  ( Big bennifits.)

A potentially useful approach, but remember that most file systems contain
a modest amount of heavily-used stuff and a lot of seldom-used stuff.  It
may be better to use that memory for a big, intelligently-managed buffer
cache, which will give 90% of the speedup while using 10% of the memory.
(Or, alternatively, will give 90% of the speedup for a filesystem ten times
as large, using the same amount of memory.)

>   7. Recursion anyone?  How about recursive compilers, hmm makes me wonder.

The PDP11 cc, running in a distinctly small address space, is a recursive
design.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,decvax,pyramid}!utzoo!henry