the elephant file system https://open.library.ubc.ca/media/download/pdf/831/1.0051213/2 cc @SRAZKVT this is rly neat it's a legit filesystem with built in version control semantics
Replies:
7
@hipsterelectron
i have this paper somewhere about capability based filesystems i would think you would like
@zardoz that paper was actually me time traveling from the future where i actually implemented all of this
@SRAZKVT really interesting how they're super harsh on any excess CPU
The cleaner's job is to examine the
segments and collect live data into full segments freeing up other segments for future use. The cleaner process can consume a large amount of disk bandwidth as well as CPU time. With the addition of a second process to implement policies the
potential for creating load on the system is quite high.
when arguing against the log-structured approach
Writing file systems is hard. The interface between kernel subsystems and a file system is undocumented and asynchronous. Applications expect many behaviours that can only be found by trial and error.
PREACH!!!!!!
8.4 Version History Merge and Branch
Fundamental to Elephant's naming scheme is the assumption the multiple versions of a file never co-exist. We have just seen, however, that disconnected operation can violate this assumption. The same is true for software revision control systems.
Academic environments exhibit very different usage patterns from research
usage patterns. In addition to the different usage pattern schools frequently employ
disk quota schemes to ensure resource availability for all students.
