Alternatives to consider for similar use cases (e.g. not requiring writing) are squashfs or cramfs. These are more transparent because they don't require applications to read the whole file from a single read(2) call, and they allow mmapping of individual pages.
Note that these are read-only filesystems. To mix with writability, the obvious suggestion would be a unionfs, but consider simpler alternatives: e.g. using essentially a writable filesystem but using an /opt scheme to allow individual pieces to be compressed. Mountpoints allow a per-directory choice between compression or writability. Symlinks can provide a simple form of compressing an individual file, perhaps comparable in convenience in practice to the scheme discussed in the parent article.
Purely read-only filesystems might offer performance advantages over filesystems that have to allow for writes. It isn't a straightforward "never needs more reads" (e.g. files are more likely to span hardware block boundaries in a filesystem that aims for compression), but it's different enough from a writable filesystem to be worth testing for a given use case.
Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds