Bcachefs Going Through Period Of More Performance Optimizations

[phoronix]

Phoronix: Bcachefs Going Through Period Of More Performance Optimizations

Bcachefs was sent out for another round of review at the end of October. While it doesn't look like this file-system born out of Linux's block cache code will be mainlined in the immediate near future, it's still on a nice trajectory...

http://www.phoronix.com/scan.php?page=news_item&px=Bcachefs-More-Performance-2020

[fanbelt]

With another year or two I hope bcashefs is ready to mainline. Once that happens we could see a bunch of users and even companies switch over from ZFS/Btrfs and invest more resources into development.

I'm also curious how the expiration of IBM's patent on ARC in 2022 will effect how filesystems and maybe even the Linux kernel deal with caches. Apparently ARC isn't all what it's cracked up to be since newer and seemingly patent-free methods like W-TinyLfu outperform it using less resources.

[cynic]

With another year or two I hope bcashefs is ready to mainline. Once that happens we could see a bunch of users and even companies switch over from ZFS/Btrfs and invest more resources into development.

usually, in order to be declared "safe" a filesystem requires a lot of time, often measured in decade.
I'm not following bcache too much so I might well be wrong here, but I don't see it gaining mainstream adoption in 1 or 2 years.

also, declaring it "safe to use" [TM] too early might cost it its reputation and mark it forever as "unrealiable", as it happened to btrfs.

[lyamc]

I think that in regards to BTRFS, it was looking to be a replacement for ZFS. The killer feature of ZFS is data safety/integrity/redundancy. BTRFS still can't do RAID5/6 properly.

[cynic]

I think that in regards to BTRFS, it was looking to be a replacement for ZFS. The killer feature of ZFS is data safety/integrity/redundancy. BTRFS still can't do RAID5/6 properly.

btrfs can do safety/integrity/redundancy and automatic data healing if ther's enough redundancy.
it can also do safe RAID56 if you use metadata raid1c3 for metadata.

[Zan Lynx]

New filesystems look great until they go into wide use. Then the flaws show up.

I look forward to the first report of complete bcachefs data loss for someone who claimed btrfs was trash. I will then laugh like a loon.

[lyamc]

btrfs can do safety/integrity/redundancy and automatic data healing if ther's enough redundancy.

That's what I thought too which I why I set up 16TB (4x4TB) in BTRFS RAID10 with one backup 6TB drive.

Guess who needed that backup drive after a few months? Me.

it can also do safe RAID56 if you use metadata raid1c3 for metadata.

Look, if you want to do RAID5/6 with btrfs, be my guest. I hope it works well for you. I'm just not willing to take that risk.

New filesystems look great until they go into wide use. Then the flaws show up.

I look forward to the first report of complete bcachefs data loss for someone who claimed btrfs was trash. I will then laugh like a loon.

I've actually been using bcachefs for over 3 years now. I ran into one issue where the drives didn't mount properly and fsck didn't seem to do anything. One update later and it was all good.

It's not like bcachefs is brand new, written from scratch. It's based on bcache which has already been in the kernel since 2013.

It's probably about as stable of a new filesystem as you can get.

[cynic]

That's what I thought too which I why I set up 16TB (4x4TB) in BTRFS RAID10 with one backup 6TB drive.

Guess who needed that backup drive after a few months? Me.



Look, if you want to do RAID5/6 with btrfs, be my guest. I hope it works well for you. I'm just not willing to take that risk.



I've actually been using bcachefs for over a year now. I ran into one issue where the drives didn't mount properly. One update later and it was all good.

so, entirely based on your anedoctical experience, you are saying that a work-in-progress fs is more robust than a mature and widely used fs?
ok.

[lyamc]

so, entirely based on your anedoctical experience, you are saying that a work-in-progress fs is more robust than a mature a widely used fs?
ok.

bcachefs is a modification to the mature bcache which has been part of the kernel since 2013.
Wanna know why bcachefs is so stable? Kent makes sure the features don't break anything before he turns them on.

Compression right now works, but he purposefully made sure that it doesn't allow you to save any additional space because he wants to make sure it's rock solid.

[cynic]

Wanna know why bcachefs is so stable? Kent makes sure the features don't break anything before he turns them on.

wow! that's very impressive! gonna put bcachefs on all my system right now!
I'm sure all the other FS developers don't care about user's data and turns on random features just to destroy them all!