Bcachefs Still Being Developed As A Next-Gen Linux File-System

[GI_Jack]

Another filesystem?

Do we really need another filesystem when

1. AUFS won't be mainlined
2. UFS support is limited
3. There is no decent zfs implementation.

[pal666]

I don't know how easier it is to develop his filesystem vs btrfs

not dozens of times easier, obviously

Also, having lots of manpower isn't a guarantee of good results, especially in coding tricky parts.

well, you could judge by current status. what distros support bcachefs?

[pal666]

just sad as usual about the lack of joint efforts.. don't we all deserve a default native FS with every distro with built-in COW and stuff?

it is btrfs. ssd-only can't be default

[pal666]

Why do you guys want RAID5/6 in the filesystem? You can already do it on a block-level basis below the filesystem.

raid in fs is better, but i don't understand why someone wants 5/6. it is raid1 for poor people, slower and less robust. drive space is cheap.

[starshipeleven]

Since the FS is taking over the responsibility of RAID, it presumably wants to have and can act upon knowledge about the state of the hardware.

No, it is about keeping the less overhead as possible with checksumming, mostly, as with RAID5/6 you are checksumming stuff that spans multiple disks (in stripes or whatever) and not having full control over the RAID itself would mean another layer to pipe communication around, lowering performance.

It probably doesn't perform better either, it's going to be hard to beat dedicated hardware RAID. There is some great value to be had for it though.

the "performs better" was relative to using LVM+btrfs instead. Checksums add another layer of communication that should go back and forth, if the filesystem handles itself the RAID it's easier to have this happen, with LVM it's slower.

Hardware raid may or may not be a good choice, it has a cache and this cache will give it better performance for workloads that don't fill it too much, otherwise it's not gonna change a thing.
Meanwhile the cache is a source of uncertainty and you need backup batteries and stuff.

[starshipeleven]

1. AUFS won't be mainlined
2. UFS support is limited
3. There is no decent zfs implementation.

1. there is OverlayFS
2. UFS is for dead technology like optical drives, it's sufficient for reading them and making the most common ones (CD/DVD).
3. lolwhut?

[pal666]

3. There is no decent zfs implementation.

zfs is obsoleted by btrfs

[starshipeleven]

not dozens of times easier, obviously

Are you perhaps assuming that everyone does the same amount of work?
There is not enough info to extrapolate, stop this.

well, you could judge by current status. what distros support bcachefs?

all distros support bcache since 3.11 and is in production already, bcachefs is a few features on top of that.

meanwhile only morons would use btrfs raid5/6 in production.

[starshipeleven]

raid in fs is better, but i don't understand why someone wants 5/6. it is raid1 for poor people, slower and less robust. drive space is cheap.

This is retarded bullshit. You never made arrays with more than 2 drives probably. When you have 6+ drives it makes tons of sense.

RAID1 wastes half array space.

RAID5 wastes only one drive's worth of it.

6 drives, 2TB each

RAID1 --> 3 drives wasted, 6TB array
RAID5 --> 1 drive wasted, 10TB array

To have a RAID1 with 10TB of capacity I'd need to use 10 of those drives, which means I need also another controller card, and a bigger case, and a bigger PSU and more space so on.

[pal666]

There is not enough info to extrapolate

it is enough to compare implemented features

all distros support bcache since 3.11 and is in production already, bcachefs is a few features on top of that.

i.e. exactly zero. btw, btrfs is a few features on top of block layer

meanwhile only morons would use btrfs raid5/6 in production.

and btrfs magically becomes worse than non-existent raid5 bcachefs