[clug] partition tables in memory

[Francis Whittle]

Interestingly enough, using LVM2, nfs became faster on my home network
than beforehand using just "normal" DOS disklabels.

That even considering the intemediary step of using LVM1 for about two
weeks while I was still on kernel 2.4

Before you ask, we have a Solaris x86 box on our network for reasons
known only to my father who no longer even works with an UltraSPARC at
ERIN, which was his reasoning for installing it in the first place -- he
needed a Solaris box to play with.
I'll give it this -- It means that when I got to CS at ANU in 2003, I
knew exactly what I was doing.  Even if I did have to repeatedly replace
'stutcsh' with bash.
And that is the epic story of why we are using nfs at home.

On Mon, 2005-04-25 at 11:15 +1000, Kim Holburn wrote:
> Don't get me started ;-)  Slowness in volumes mounted from an nfs 
> server using LVM may be a contributing cause of the problems with this 
> system that makes it require more memory.  In any case, God didn't 
> invent LVM, it's an attempt to fix underlying problems with hardware 
> and maybe a good idea, I always thought it was a good idea, it's just 
> the implementation is lacking something.  Adding yet more layers of 
> indirection doesn't always help, especially in a system with hundreds 
> of users.  Also there is no reason to use LVM on system partitions in a 
> system disk, it would just invite problems.  I'll shut up now.