Author: Jake Bauer <firstname.lastname@example.org>
Date: Sun, 18 Sep 2022 14:25:51 -0400
3 files changed, 50 insertions(+), 10 deletions(-)
diff --git a/content/garden/arboretum/knowledge/sysadmin/general-tips-and-tricks.gmi b/content/garden/arboretum/knowledge/sysadmin/general-tips-and-tricks.gmi
@@ -4,3 +4,7 @@
+## Benchmarking Scripts
diff --git a/content/garden/greenhouse/index.gmi b/content/garden/greenhouse/index.gmi
@@ -16,16 +16,17 @@ The Greenhouse is the place where new things are incorporated into the garden. I
These are things that I have reviewed and taken notes on but don't know yet what to do with.
+=> tactics-and-mindset-shifts-for-making-the-most-of-life.gmi AoM Podcast #825: Tactics and Mindset Shifts for Making the Most of Life
+=> linux-issues.gmi Issues with Linux
+=> freebsd-issues.gmi Issues with FreeBSD
+=> zfs-issues.gmi Issues with ZFS
+=> user-profile-systems-bad-assumptions.gmi Bad Assumptions Made By User/Profile Systems
A collection of notes and clippings from articles that don't yet fit anywhere else, but which I still find valuable.
+=> clippings.gmi Clippings
@@ -40,11 +41,8 @@ Items are added to the top of the list as I come across them.
@@ -55,9 +53,8 @@ Items are added to the top of the list as I come across them.
^ And I really should just go through the Wikipedia pages on major philosophies to get at least a cursory idea of all of them
diff --git a/content/garden/greenhouse/zfs-issues.gmi b/content/garden/greenhouse/zfs-issues.gmi
@@ -0,0 +1,39 @@
+# Issues with ZFS
+Note: lots of stuff here needs to be tested. A lot of this is based off of feel and experience, not hard, cold metrics.
+## It's too damn complicated.
+ZFS has far too many features, some of which are desirable like its data integrity guarantees and software RAID capabilities, but others aren't _really_ needed, like filesystem-level compression. Then again, software RAID could easily be in another layer entirely, and doesn't have to be integrated into the filesystem layer.
+Also on the BSDNow podcast, Allan Jude once said that once a certain ZFS feature is done, it will become extremely hard to change anything going forward, so they had to make sure they got that right. I don't remember which podcast episode or what feature (it might have been RAID-Z expansion), but that is not something _good_. In fact, that is _very very bad_. Software so complex that it strikes fear into the heart of developers at the thought of changing something is _bad software_.
+## It needs too much memory
+Performance of the filesystem is heavily dependent on having enough memory to store frequently accessed blocks. If you don't have enough memory, performance will really start to suffer. This is undesirable given that not every modern machine should need 16GB+ of RAM... or even 8GB. People who want the data integrity guarantees of ZFS also have to put up with the fact that less RAM will be available for their browser or video game, unless they want their filesystem performance to suffer.
+## The license sucks
+Like the GPL, the license is hard to understand for the layperson. Given that Torvalds didn't want to integrate ZFS for this reason, and only after a review of ZFS by Canonical's League of LawyersM did ZFS come to Ubuntu, this license is tricky for people who care about such things.
+Like the GPL, the license is hard to understand for the layperson. Given that Torvalds didn't want to integrate ZFS for this reason, and only after a review of ZFS by Canonicals League of Lawyers did ZFS come to Ubuntu, this license is tricky for people who care about such things.
+## It's inflexible
+Once you create an array, you better not want to upgrade it in the future unless you want to spend a whole bunch of money on new drives. Since you can't expand an existing vdev with more drives, if you want to take your 12-bay storage chassis filled with 6 drives and add another drive for just a little more storage, well, nope, you need to buy another 4-6 drives to make another vdev so you can mirror it with the first.
+Btrfs handles this much better... (and no, it won't randomly lose your data and hasn't done so for a _long_ time).
+## It might kill SSDs faster?
+Some people say something about "write amplification" where ZFS will cause more writes than other similar solutions (Linux md, btrfs, etc.) which is fine on hard disks (which is a place ZFS shines) but will prematurely wear the NAND flash of SSDs. I don't know how much of this is true though.
+## It's kind of a cult
+Anybody who has come across ZFS people and not been immediately enamoured by the filesystem will attest to the fact that there are ZFS "evangelists" and other people like that who pretend it's the One True Filesystem, ignore its flaws, and try to put down any other filesystem often using misinformation. It's similar to the Rust community in that respect. This is not healthy. You should not worship a filesystem. Do you hear how ridiculous that sounds? Worshipping a filesystem...