Linked by David Adams on Fri 12th Aug 2011 03:55 UTC, submitted by Debjit
Red Hat The Fedora developers have decided that Fedora 16 will not use Btrfs by default. The announcement was made by Josef Bacik, Senior Software Engineer at Red Hat. Instead the switch to Btrfs filesystem has now been postponed to Fedora 17.
Thread beginning with comment 484799
To read all comments associated with this story, please click here.
Comment by Luminair
by Luminair on Sat 13th Aug 2011 01:46 UTC
Luminair
Member since:
2007-03-30

"we are working very hard on trying to get everything more stable and it is a slow going process. With your help we will be in a better situation for F17."

That is a mature comment. Maybe next year they'll be ready for serious use. It can't be much longer before they catch ZFS. It's been a long fricking marathon.

Reply Score: 2

RE: Comment by Luminair
by Kebabbert on Sat 13th Aug 2011 10:34 in reply to "Comment by Luminair"
Kebabbert Member since:
2007-07-27

"we are working very hard on trying to get everything more stable and it is a slow going process. With your help we will be in a better situation for F17."

That is a mature comment. Maybe next year they'll be ready for serious use. It can't be much longer before they catch ZFS. It's been a long fricking marathon.

You know, it took ZFS several years AFTER it was released to iron out most bugs. It takes years to make a filesystem stable.

If a kernel crashes, then you loose a few hours of work. If a filesystem crashes, you loose years of work. It is much more important that a filesystem is bug free, than a kernel.

There are many sysadmins that wont let ZFS into their server halls, because it is too young. ZFS is almost 10 years old today. But ZFS had a good reputation with a strong team. Sun knew servers. They knew server halls. They knew Enterprise and had much experience. Sun only did enterprise. Still ZFS is not really trusted yet by sysadmins.

To really believe that BTRFS will be trusted by sysadmins is a bit strange. It will take at least 10 years after it has been released. BTRFS does not have a strong team of server Enterprise guys. They are mostly some hackers that never ever been in a Enterprise server hall, nor any experience. I doubt BTRFS will be mature even after 10 years.



Regarding BTRFS performance. Just scrap that discussion. If you have a fast filesystem, then it is not safe and might corrupt your data. There is a reason ZFS is slow: it calculates many checksums all the time. If there is a fast filesystem, it does not compute checksums, and hence your data might corrupt.

I would never trust anything fast. I heard one guy said that XFS did fsck his RAID in 15 seconds or so. Well, if you fsck a 6TB raid in 15 sec, it means you never read the actual data. To read 6TB data takes many hours. Needless to say, I dont trust XFS fsck. Because it is too fast, it is not thorough.

Fast and reliable works against each other. Choose one of them, but not both. There is a reason IBM Mainframes have slow cpus: they double check calculations I heard.

Reply Parent Score: 3

RE[2]: Comment by Luminair
by NorthWay on Sun 14th Aug 2011 14:40 in reply to "RE: Comment by Luminair"
NorthWay Member since:
2007-02-22

There is a reason IBM Mainframes have slow cpus: they double check calculations I heard.

Yes. And no (AFAIK).
They do run the same instruction on two different units and compare.
That has implications on how much exotics you can do to out-of-order and suchlike, plus the instruction set is old and archaic. Which results in less opportunities to become speed demons. They take a lot back in having an IO interface definition that can be co-processor accelerated.

Reply Parent Score: 1

RE[2]: Comment by Luminair
by korpenkraxar on Mon 15th Aug 2011 07:58 in reply to "RE: Comment by Luminair"
korpenkraxar Member since:
2005-09-10

ZFS is a fine FS that's for sure. As a Linux guy, I have not used myself though and have a question.

About the checksum, if something is not correctly stored on the disk (lets say I save an ODF file), how is the end user informed of the problem? Assuming a hardware problem, what will ZFS do to avoid writing to that sector again? I mean, what is the level of interactivity and robustness of the checksum system to make it useful to say desktop users?

Reply Parent Score: 2

RE: Comment by Luminair
by ggeldenhuys on Sun 14th Aug 2011 07:58 in reply to "Comment by Luminair"
ggeldenhuys Member since:
2006-11-13

It can't be much longer before they catch ZFS.


I haven't used ZFS, but do read a lot about it. The features it has, and the enterprise support it has doesn't come overnight. Btrfs has promise, but it is still a long way from ZFS.

I watched a YouTube video demoing some of ZFS features. The ease of setting up a Raid5, growing or shrinking the array by adding or removing hard drives (without total array rebuilds or loosing data), built-in compression etc was all just astonishing! Btrfs is not there yet.

It's a real pity ZFS licensing conflict with Linux.

Reply Parent Score: 2