Re: Re: ENBD and swap

> No.  Not with current kernel architecture (AFAIK).  But as I said, so
> long as you run the connection it is likely that you will ALWAYS have a
> 1.5KB buffer already allocated, and if you keep the transaction
> blocksize to 1KB and never send one request before another is finished,
> and never fragment packets, and never have to reassemble or retry after
> collisions, you won't ever force the kernel to need more buffer space.
> 
> So why not just do that?

I was investigating about using ENBD in combination with LTSP. LTSP is
supposed to work on all kinds of old equipment (and is a software
project, not a single use-case, so nobody has control over the equipment
- the point is to keep LTSP as stable as possible in all circumstances).
So one must accomodate 10 MB hubs (no full-duplex), there will never be
a dedicated network link, etc.

LTSP currently works fine by using the swap-over-NFS patch. However this
patch is not present in the 2.6 kernel, and keeps LTSP stuck with 2.4
for now.

(It is for the same reasons that one does swapping over the network at
all, which is crazy under normal circumstances - 32 MB (for instance) is
too little memory for X in some situations (lots of images in a
browser...), but even if it performs poorly it is better than a crash,
and isn't used in the majority of the cases - a little sluggishness when
loading X heavily is just as expected by the users).

Still, I don't know if these are the issues they are experiencing - I'm
not the one doing the work, I merely saw discussion about NBD on the
LTSP mailing list, and, having success with replacing NBD with ENBD in
the past, I wanted to see if doing so in this case would solve any known
issues.

// Dag Sverre