JFFS3 memory consumption

[Josh Boyer]

On Wed, 2005-01-26 at 09:57, Artem B. Bityuckiy wrote:
> On Wed, 26 Jan 2005, David Woodhouse wrote:
> 
> > On Wed, 2005-01-26 at 08:44 +0000, Artem B. Bityuckiy wrote:
> > > K1 is decreased by the Ferenc Havasi's "summary" patch well. Would be
> > > nice to decrease K2 either. Any Ideas how to do so?
> > 
> > For a start, we can increase the maximum node size. Limiting it to
> > PAGE_SIZE means we have a larger number of nodes than we'd otherwise
> > have. If we quadruple that, we'll probably end up with about a third as
> > many nodes. 
> I thought about to have several 4K chunks within one node. Thus we will 
> keep reading fast but will have few nodes number. Didi you mean this?

If you use larger "chunks" you'll get better overall compression.  E.g. 
a 64KiB node should compress better than a 4KiB node.  This is partially
how squashfs achieves better compression that cramfs.

So the end result is better compression and fewer nodes.

josh