Subject: Re: File Size Limitations
To: David Laight <david@l8s.co.uk>
From: Louis Guillaume <lguillaume@berklee.edu>
List: netbsd-docs
Date: 01/19/2005 01:00:41
David Laight wrote:
> On Mon, Jan 17, 2005 at 10:08:21PM -0500, Perry E. Metzger wrote:
> 
>>"Geert Hendrickx" <geert.hendrickx@ua.ac.be> writes:
>>
>>>I also find there is very limited documentation on the filesystems
>>>included in NetBSD.  About UFSv2 the docs only mention that "among other
>>>enhancements, it is better suited for filesystems larger than 1 TB".
>>>What does that mean, better suited?
>>
>>It uses 64 bit block numbers instead of 32 bit ones. Since blocks are
>>512 bits, well, you can see why 2TB becomes a hard limit.
> 
> 
> Nope, FFSv2 uses 64 bit fragment numbers, since fragments can be 64kB
> FFSv1 can support filesystems with 2^31 x 64kB which is rather larger
> than 2TB.
> 
> Under NetBSD the only differences are:
> 1) FFSv2 inodes are twice the size
> 2) FFSv2 netfs is faster - doesn't write all the inode sectors.
> 3) FFSv2 can be used with 512byte fragemnts on a 2^32 sector partition
> 
> 	David
> 


2^31 x 64K = 128 TB

Wow! That's big!

Just curious - what are the limitations in practice rather than in 
theory? At least, what have people seen out there?

I've been trying to research which fs type to use for large partitions 
on NetBSD. The jist of what I've heard is that...

<  1TB - FFSv1
 >= 1TB - FFSv2
 >  (5?)TB - SAN/NAS

But mere curiosity leads me to wonder what happens if we put our trust 
in FFSv2 for giant filesystems, ignoring the advice to look at SAN/NAS 
solutions.

Where does FFSv2 start to break down in practice and why?

Does FFSv2 do journaling/logging?

How does it stack up against xfs, ext2, reiserfs, jfs?

Thanks - this is a great discussion!

Louis