On Sat, Jan 25, 2020 at 02:52:05PM -0700, Chris Murphy wrote:
Question and (pre)proposal:
Can Fedora converge on a single swap-on-ZRAM implementation, and if
so, which one? Fedora Workstation WG wants to move to swap-on-ZRAM by
default in Fedora 33, and the working group needs to pick something
I think it should be zram-generator. It's the most lightweight, can be
included by default distro wide. Without a configuration file, it
doesn't run. Thus, each edition/spin, and even the install
environment, can have their own configuration file, to setup it up
however they want, or not set it up.
I agree with the general proposal of enabling zram by default in
Workstation. From what we have seen so far, it helps in some cases,
without making things worse. If if turns out that there are some
unforeseen downsides, we can trivially back out.
As to which implementation should be used: I think zram-generator is a
solid choice, but I can't say much more than that: I simply don't know
enough about the alternatives to make meaningful comparisons.
I also suspect it's the only one that could be upstreamed to
proper, and just included like many other generators.
It is "upstream" in the sense that it is under the same umbrella.
There are no plans to move the code to the main repo, because it's in
rust and currently combining meson which is used for systemd proper
with rust and cargo is not very convenient. And the release schedule
of systemd and zram-generator is fully independent, so there isn't any
particular reason to want to merge them.
(I have seen the RFEs you filed. Been too busy with devconf and travel
to handle them, but they're on the todo list.)