Is loopback-based FS slower than normal FS?

Is loopback-based FS slower than normal FS?

Post by Arcadio Alivio Since » Wed, 10 Jul 1996 04:00:00



        Is a filesystem based on the new loopback (not to be confused
with the loop back device related to TCP/IP networking) device of Linux
v2.0.0 slower, faster, or just the same as a "real" (i.e. on a real
partition on the HD) filesystem?

        I ask this because I got an idea for users using the UMSDOS
filesystem.  I'm thinking people use UMSDOS 'cause they don't want to
repartition their drives.  The price they pay for using UMSDOS is a
slower filesystem, and wasted disk space due to the way FAT stores data.

        I'm thinking, that UMSDOS users can mount a ext2fs formatted
loop-back based filesystem as thier /usr filesystem.  Would this offer
any advantages?  Assuming that using a loopback device is just as fast as
using a real partition, there would be a speed advantage.  Also, storage
efficiency may be increased; just make the size of loopback file a
multiple of cluster size.  

        Comments please .....

--
===============================================================================
Arcadio Alivio Sincero, Jr.
Sophomore, Computer Science Major at the University of Maryland at College Park
Amateur competitive bodybuilder

"D.A.R.E. .... to keep cops off donuts."

 
 
 

Is loopback-based FS slower than normal FS?

Post by Andries Brouw » Thu, 11 Jul 1996 04:00:00



:       Is a filesystem based on the new loopback (not to be confused
: with the loop back device related to TCP/IP networking) device of Linux
: v2.0.0 slower, faster, or just the same as a "real" (i.e. on a real
: partition on the HD) filesystem?

:       I ask this because I got an idea for users using the UMSDOS
: filesystem.

I prefer to call it the `loop device', and reserve loopback for
networking. It is slower than an ordinary file system because
an additional copy is involved. (But of course someone should
improve loop.c so as to avoid this copy when no encryption is done.)

I have no idea how ordinary UMSDOS and ext2 over the loop device
would compare.

Note that the loop device is not very stable yet.

 
 
 

Is loopback-based FS slower than normal FS?

Post by Steve Beckm » Thu, 11 Jul 1996 04:00:00



wrote *well*):

Quote:>Note that the loop device is not very stable yet.

True, and frustrating!  It works well on my machine, but only without
encryption.  Both DES and xor encryption give me ioctl errors.  Any ideas?
I have read all the docs I can find; maybe I am missing something?
 
 
 

Is loopback-based FS slower than normal FS?

Post by Arcadio Alivio Since » Fri, 12 Jul 1996 04:00:00



: wrote *well*):
: >Note that the loop device is not very stable yet.
: True, and frustrating!  It works well on my machine, but only without
: encryption.  Both DES and xor encryption give me ioctl errors.  Any ideas?
: I have read all the docs I can find; maybe I am missing something?

        So how stable is the loop device without encryption?  As my
purpose here is to provide UMSDOS users EXT2FS without re-partitioning.

        How does one measure the speed of a filesystem, anyway?  Perhaps
I should just measure it myself, but I need a way to accurately measure
filesystem speed.  Or should I just wing it?

--
===============================================================================
Arcadio Alivio Sincero, Jr.
Sophomore, Computer Science Major at the University of Maryland at College Park
Amateur competitive bodybuilder

"D.A.R.E. .... to keep cops off donuts."

 
 
 

Is loopback-based FS slower than normal FS?

Post by Steve Beckm » Fri, 12 Jul 1996 04:00:00



Quote:>    So how stable is the loop device without encryption?  As my
>purpose here is to provide UMSDOS users EXT2FS without re-partitioning.

Seems to work just fine without encryption.  I've been using it for normal
filesystem stuff, and it fits in seamlessly.  I'm going to give it a torture
test, maybe build a kernel on it and even try it as a root, who knows?!
I'll let you know....