I have been thinking of how to speed up up MAKE_HANDLE_PATH and have put up
a test patch ($subject).This will first try with a buffer of 1K and only
fall back to the calculation of the patch doesn't fit
in that space.
I think it likely that 1K will be sufficient for most data sets. A better
way would be to keep track of the max path len across calls and use that
instead of a fixed 1K size . Otherwise there would be a performance dip for
those data sets which may have very deep directory structures.
Thoughts on the above?
Gluster-devel mailing list