Oracle Now Supports Btrfs RAID5/6 On Their Unbreakable Enterprise Kernel

[starshipeleven]

note that on the official btrfs wiki, raid 5/6 is still considered unstable, and write holes still happen as checksums aren't leveraged by current RAID parity code. (the whole reason to use RAID5/6 in BTRFS instead of classical LVM + MD stack).

WTF?
The write hole just means that the filesystem isn't power-failure-safe as there is a short amount of time the writes aren't completed fully.
This is not a major issue for a server because they are all protected by an UPS for other good reasons already.

[starshipeleven]

Lets face it, ZFS failed the moment Oracle bought SUN and Apple dropped their plan of adopting ZFS as the filesystem for OS X due to Oracle refusing to licence it. Apple begun developing APFS in 2009 for reference so it and BTFS took around the same time to get first deployed (7/8 years).

With the main difference that APFS isn't even remotely comparable to btrfs or ZFS, lacking most of their features.

[sdack]

Lets face it, ZFS failed the moment Oracle bought SUN and Apple dropped their plan of adopting ZFS as the filesystem for OS X due to Oracle refusing to licence it. Apple begun developing APFS in 2009 for reference so it and BTFS took around the same time to get first deployed (7/8 years).

Agreed. Considering how Oracle sued Google for 9 billion dollars over copying the Java API can one assume anything is going to turn into shit as soon as Oracle gets their hands on it. Luckily has the last court decided in favour of Google that it's still only a case of fair use, but I'm sure the greed in Oracle is strong and they will keep trying to fight them further. Too bad really for ZFS.

[arokh]

With the main difference that APFS isn't even remotely comparable to btrfs or ZFS, lacking most of their features.

It has copy-on-write (metadata), snapshots, checksumming (metadata), encryption and logical volume management. How is that not comparable?

[dragon321]

It has copy-on-write (metadata), snapshots, checksumming (metadata), encryption and logical volume management. How is that not comparable?

Compression, integrated RAID, subvolumes, deduplication, online defragmentation. Still comparable?

APFS is comparable to NTFS, which has same features. Well, NTFS hasn't got checksumming, but APFS hasn't got compression.

[arokh]

NTFS literally has none of the features I mentioned, which APFS has in common with BTRFS. NTFS is from 1993, if you think it's comparable to APFS then you must be high.

[starshipeleven]

It has copy-on-write (metadata), snapshots, checksumming (metadata), encryption and logical volume management. How is that not comparable?

a filesystem that has CoW and checksumming only on metadata cannot be compared with one that is CoW on everything (btrfs/zfs) AND can software-raid on its own.

I still don't see the point in CoW and checksumming only metadata. Isn't a journal enough?

NTFS literally has none of the features I mentioned,

CoW and checksumming of metadata are comparable to NTFS's journaling and redundant MFT (metadata), unless someone can explain me why NTFS's system is so much worse.
NTFS has encryption since a long time ago. https://en.wikipedia.org/wiki/Encrypting_File_System
NTFS has online fsck (in Win8 and Server 2012 and later). https://technet.microsoft.com/en-us/...(v=ws.11).aspx
NTFS supports deduplication (although it is enabled only on server versions, you need to hack the Home/Pro versions to enable it afaik) http://gestaltit.com/syndicated/step...tfs-windows-8/

NTFS is from 1993, if you think it's comparable to APFS then you must be high.

NTFS added features well after that.

[Spazturtle]

a filesystem that has CoW and checksumming only on metadata cannot be compared with one that is CoW on everything (btrfs/zfs) AND can software-raid on its own.

I still don't see the point in CoW and checksumming only metadata. Isn't a journal enough?

CoW and checksumming of metadata are comparable to NTFS's journaling and redundant MFT (metadata), unless someone can explain me why NTFS's system is so much worse.
NTFS has encryption since a long time ago. https://en.wikipedia.org/wiki/Encrypting_File_System
NTFS has online fsck (in Win8 and Server 2012 and later). https://technet.microsoft.com/en-us/...(v=ws.11).aspx
NTFS supports deduplication (although it is enabled only on server versions, you need to hack the Home/Pro versions to enable it afaik) http://gestaltit.com/syndicated/step...tfs-windows-8/

NTFS added features well after that.

If a NTFS system crashes whilst writing a change to a file that file is lost even though the metadata survives, APFS writes changes to files as deltas so only the changes are lost. You can also set a volume to use full copy-on-write for files as well as metadata. Also copy-on-write is better then journaling because journaling writes to the journal and then the file, so it has to write the data twice, copy-on-write only has to write it once.

Also Apple's docs explain why redundant metadata is useless:

Does Apple File System support redundant metadata?

With modern Flash/SSD storage, writing two blocks of data to different locations does not guarantee that the blocks will be written to separate locations. The Flash translation layer typically groups writes together into the same NAND block. Therefore it affords no extra protection to write a second copy at the same time the first copy is written.

NTFS's encryption is nowhere near as good as it is a single key, with APFS you can use "You can choose one of the following encryption models for each volume in a container: no encryption, single-key encryption, or multi-key encryption with per-file keys for file data and a separate key for sensitive metadata."

"NTFS supports deduplication" No it doesn't per your article "Microsoft’s deduplication is layered onto NTFS in Windows 8," So it is effectively a program that scans you files and links duplicates, you can do that on any file system.

[starshipeleven]

If a NTFS system crashes whilst writing a change to a file that file is lost even though the metadata survives, APFS writes changes to files as deltas so only the changes are lost.

Wait, I though that this trick was working only for "cloned files" (i.e. files that are a copy of another file, which in APFS and CoW filesystem are basically both pointers to the same data on disk).

You can also set a volume to use full copy-on-write for files as well as metadata. Also copy-on-write is better then journaling because journaling writes to the journal and then the file, so it has to write the data twice, copy-on-write only has to write it once.

Yeah I know, I was talking of metadata only. If you want to protect data too there is only CoW or you have garbage performance.

Is the "set the volume as full CoW" a preview feature they want to enable by default in the future or not?

NTFS's encryption is nowhere near as good

Yeah, I know it's not particularly good in general too, but it does have encryption.

"NTFS supports deduplication" No it doesn't per your article "Microsoft’s deduplication is layered onto NTFS in Windows 8," So it is effectively a program that scans you files and links duplicates, you can do that on any file system.

That's a bit more involved than that, it is block-level deduplication, not just hard-linking files.

https://msdn.microsoft.com/en-us/library/hh769303(v=vs.85).aspx

http://jeffwouters.nl/index.php/2012/01/disk-deduplication-in-windows-8-explained-from-a-to-z/

[Zucca]

oracle doesn't need zfs, it has superior btrfs
this article makes me laugh at zfs idiots

This will raise all those trolls from their graves. Don't feed them any brains.