Ron Johnson <[EMAIL PROTECTED]> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On Sunday 22 July 2001 16:32, Adrian wrote:
> > Hi,
> >
> > This is a fool question but I need some help:
> >
> > How build a ramdisk in LM 8? ...
> Surely, I won't be the only one to say this, but here goes:
> 1. Trust linux.  It does a good job of caching data.
> 2. The more RAM that you use for a RAM disk, the less RAM that
>    the PC will have for "regular" use, and so the more likely 
>    that it will have to use the swap files.

(Note that there was another answer talking about why ram disk
is faster than cache, I don't wish to go there right now, sorry.)

(I've been running something performance monitoring software which
is (sort of) from Sun (called SE Performance Toolkit - really
cool, even though it missed a thing or too), and it has been
instructive - I wish we had something like it for Linux!  I've
learned a little more about performance enhancements through
watching what it detects...)

Ok, to the point: 

1 - ram is fast, disk is slow. (For very LARGE values of slow!)
        Before worrying about ramdisk, make CERTAIN that you have
        plenty of RAM for cacehing to do its job.  If you don't have
        enough ram, adding a ramdisk will simply make programs start
        running from the disk, which is worse than 'simply' having
        data accesses hit the disk.  If you can have enough ram
        to hold the entire database AND the program AND the OS
        AND anything else that needs to run, then you're starting
        where you want ;-)  and can THEN consider other issues.

2 - after writing all that, i went back and re-read your initial
        query.  What exactly is the test supposed to measure?
        'copying a database to it' - are you trying to measure
        the database access speed?  Data transfer rates from the
        database?  Something else?  Or are you doing what I assumed -
        trying to make the database run faster?  Depending
        upon what you're trying to do will result in different
        answers...

        if you are just trying to measure transfer rate, send the
        data to /dev/null - that's pretty fast ;-)  (and does not
        require memory :-)  But beware - running the same command
        twice in a row will most likely result in MUCH different
        numbers, because of cacheing (assuming you are not memory
        constrained).  (I could tell a story here, but I won't ;-)

rc

(note - bcc copy sent to Adrian - hopefully the list won't see but one!


Rusty E. Carruth       Email:     [EMAIL PROTECTED] or [EMAIL PROTECTED]
Voice: (480) 345-3621  SnailMail: Schlumberger ATE                   ___
FAX:   (480) 345-8793             7855 S. River Parkway, Suite 116   \e/
Ham: N7IKQ @ 146.82+,pl 162.2     Tempe, AZ 85284-1825                V
ICBM: 33 20' 44"N   111 53' 47"W  http://tuxedo.org/~esr/ecsl/index.html

Reply via email to