[Lustre-devel] global epochs [an alternative proposal, long and dry].

Nikita Danilov Nikita.Danilov at Sun.COM
Wed Dec 24 08:16:23 PST 2008

Andreas Dilger writes:
 > Nikita,


 > I still need more time to re-read and digest what you have written,
 > but thanks in advance for taking the time to explain it clearly and
 > precisely.  This algorithm does seem to be related to the one originally
 > described in Peter's "Cluster Metadata Recovery" paper where the epoch
 > numbers are pushed and replied by every request, but is much better
 > described.


 > I think what would help me understand it a bit easier if it could be more
 > closely mapped onto a potential implementation, and the issues we may see
 > there.  For example, the issue with fsync possibly involving all? nodes
 > (including clients) is not obvious from your description.

I agree with Eric that we can discuss this in more detail in Beijing,
and just want to make one rather obvious remark: your and Alex's
concerns about fsync are indeed justified, because in the global epochs
model fsync is no different from sync, as no per-object dependencies are
tracked. On the other hand, for the target use case of pNFS, where every
operation is synchronous, this is probably less important.


More information about the lustre-devel mailing list