Could you provide an explanation of the cleanerd's algorithm and method in
cleaning the file system.  What all does it do during a pass, what kind of
complexity is the operation, does it generally follow the cleaner suggested
in the original SPRITE LFS paper (where the write cost on a cleaning
operation is 2/(1-u) where u is the utilization of segments), etc.  The
reason I ask is that in my research, we have done some work on page cache
algorithms to improve the cleaning in a log structured file system.  NILFS2
seems to work quite well in our preliminary tests with Flash SSDs (as one
would expect), but before we move forward to actual implementation of our
page cache ideas, it would be good to understand how the NILFS2 cleaner
works and how we might interact with it.  I have done some digging in the
code, but I think a relatively high level explanation of the strategy and
motivation behind the design choices would be very helpful in understanding
exactly what is going on and how our algorithms might interact with it.
Thanks in advance.

 

John Garrison

_______________________________________________
users mailing list
[email protected]
https://www.nilfs.org/mailman/listinfo/users

Reply via email to