The arrow of time

Ivan Voras' blog

SoftUpdates - the next generation (or SU+J)

Jeff Roberson is working on an addition to UFS softupdates which includes a tiny journal to keep track of things like free space and inode orphaning that were left to background fsck to deal with in the original implementation of softupdates. Basically, it's still the same old softupdates but without the need for (bg)fsck to be run for recovery!

The details are in the  linked articles. It appears it will combine the best aspects of both worlds - speed of softupdates and complete removal of the need for fsck (at least in the common case) - which is very very nice.

 

#1 Re: SoftUpdates - the next generation (or SU+J)

Added on 2009-12-10T00:20 by BSDfan

Awesome!

#2 Re: SoftUpdates - the next generation (or SU+J)

Added on 2009-12-10T15:16 by mato

awesome!!

 

i'm just said it didn't come a few years ago.  now zfs is going to be used more and more.  surely, zfs is not a silver bullet and it can't be used on memory constrained hw, though i don't expect huge ufs filesystems on such constrained systems.

#3 Re: SoftUpdates - the next generation (or SU+J)

Added on 2009-12-12T01:08 by Ivan Voras

UFS certainly has its place. To overuse a corny metaphore, file systems are like wine - those that age well are highly regarded. ZFS is of course a huge step forward, but I know for sure that that there are persons and organizations that will never use a file system that has not been known and tested for at least 10 years.

And this does not fare very well with ZFS, which still changes very very quickly. FreeBSD has imported version 14 when it was new - AFAIK this was at the very end of 2008. - today, the Sun's bleeding edge of ZFS is at version 22! This is way too fast.

Though some details have changed - like the increase in size of offset pointers (32-bit to 64-bit in UFS2) and the addition of SoftUpdates (which has started almost exactly 10 years ago), the basic logic and layout - the things that make UFS the UFS, have basically not changed in decades. Even this addition of a small, light-weight journal, will not change any of the basic logic and data structures.

All this is important because it means that its failure modes are still known and the information is organized in a well known way; it is relatively easy to create a simple program from scratch that will read data from a UFS file system directly off a drive. It is very hard to write such a program for ZFS since it's so much more complex.

#4 Re: SoftUpdates - the next generation (or SU+J)

Added on 2009-12-12T03:05 by Hussain

'but I know for sure that that there are persons and organizations that will never use a file system that has not been known and tested for at least 10 years.'

This is so true.  Even with petabytes of data in ZFS (20%+ of that in FreeBSD ZFS), there are some in my workplace that have that very same attitude on ZFS and UFS always wins out.

btw - small correction , FreeBSD 7-STABLE and 8-RELEASE contain ZFSv13.

 

Post your comment here!

Your name:
Comment title:
Text:
Type "xxx" here:

Comments are subject to moderation and will be deleted if deemed inappropriate. All content is © Ivan Voras. Comments are owned by their authors... who agree to basically surrender all rights by publishing them here :)