Possible availability improvement

Hi all,

On the wave of Neil's recent roadmap proposal I wanted to forward an
idea to bring it upon discussion.

When things in md-world break, there's usually some big and long strain
on the array members. My colleague's idea is to virtually split the
members into smaller chunks and keep some metadata (events, checksum?)
for each of them to reduce the extent of recovery needed when faced
with, say, a power failure.

Does that make sense?

Kind regards
-- 
Michał Sawicz [off-list ref]