In comp.os.linux.advocacy, Tim Smith
<reply_in_group@xxxxxxxxxxxxxxxx>
wrote
on Fri, 06 Jul 2007 08:54:57 -0700
<reply_in_group-A5B637.08545706072007@xxxxxxxxxxxxxxxxxx>:
> In article <5g02m4-5c2.ln1@xxxxxxxxxx>, "[H]omer" <spam@xxxxxxx> wrote:
>> Verily I say unto thee, that Roy Schestowitz spake thusly:
>>
>> > Speed up your Linux box with a thumb drive
>> >
>> > ,----[ Quote ]
>> > | Linux only: Like ReadyBoost for Vista, SwapBoost uses the extra
>> > | space on a spare USB drive to speed up your OS.
>> > `----
>> >
>> > http://lifehacker.com/software/featured-linux-download/speed-up-your-linux-b
>> > ox-with-a-thumb-drive-274911.php
>>
>> Erm ... they mean?:
>>
>> swap_file="/media/IPOD/swapfile"
>> dd if=/dev/zero of=${swap_file} bs=1024 count=716800
>> mkswap -v1 ${swap_file}
>> swapon ${swap_file}
>>
>> Yes, very innovative :)
>
> If you want to duplicated ReadyBoost, you'll need a couple more steps:
>
> Rewrite Linux VM and cache system
> Recompile kernel and reboot
>
> Flash drives have low latency but usually also have low transfer rates
> compared to internal drives. Just adding them to your swap pool is not
> all that useful. To use them effectively, you want to just cache on
> them things that benefit greatly from low latency but don't require high
> transfer rate.
>
> If you are going to be an anti-Microsoft Advocate, you should at least
> spend a few minutes learning about Microsoft's software.
>
The Linux swapspace system already has prioritization;
there's a -p option on /sbin/swapon that allows this if
one wants to specify it explicitly.
If one does not specify -p, or puts the entries in
/etc/fstab instead, the first one has highest priority.
Pages are written to the highest swap space (partition or
file) that isn't full.
I'd have to research things such as read-ahead, but Linux
doesn't really distinguish one fault-page from another, as
far as I know. If one wants to precache small executables
in the flash drive so that one can later pull out the drive
and plug it into another system, that would require a fair
amount of reengineering of the Linux kernel (and would
include things outside of the VM space, as the standard
method of loading an executable just reads the header bit,
and the other pages are faulted in later as the program
counter needs them; since the PC jumps around like a flea
on a hot griddle the pages may not be read in any semblance
of order).
What exactly is desired here?
--
#191, ewill3@xxxxxxxxxxxxx
Useless C++ Programming Idea #40490127:
for(;;) ;
--
Posted via a free Usenet account from http://www.teranews.com
|
|
