Recent Features
| |||||||||||||
ext4 and data lossext4 and data lossPosted Mar 12, 2009 8:46 UTC (Thu) by michaeljt (subscriber, #39183)Parent article: ext4 and data loss
I'm no filesystem expert, so (disclaimer...) the following suggestion may well just be plain stupid. Why not just have a journal (metadata and data lumped together) with fixed blocks for data which does not yet have its own blocks? Certainly performance would be worse than if there were no such journal, but probably still better than having to allocate blocks for short-lived files. The journal could be moved from time to time if block wear is an issue, although I presume that these days that is nearly always worked around at a lower level.
(Log in to post comments)
ext4 and data loss Posted Mar 12, 2009 11:03 UTC (Thu) by eru (subscriber, #2753) [Link] Why not just have a journal (metadata and data lumped together) with fixed blocks for data which does not yet have its own blocks?I believe this is more or less how a log-structured file system works: http://en.wikipedia.org/wiki/Log-structured_file_system. For some reason the idea is not in very common use.
ext4 and data loss Posted Mar 12, 2009 11:11 UTC (Thu) by michaeljt (subscriber, #39183) [Link]
Indeed. Except of course that I was thinking more of a very short-lived log for data which would not otherwise be written back, rather than a full-blown log-structured file system.
ext4 and data loss Posted Mar 13, 2009 8:54 UTC (Fri) by michaeljt (subscriber, #39183) [Link]
In fact, one could sum up the idea as a mini-log filesystem on top of the normal filesystem, acting as a cache for things that can't immediately be written out to the real filesystem. Actually, I could imagine something like that existing outside of the actual filesystem code, but handled separately by the kernel, with the log either a pre-allocated file on the filesystem or a set of pre-allocated blocks on a swap device.
ext4 and data loss Posted Mar 13, 2009 0:23 UTC (Fri) by nix (subscriber, #2304) [Link]
Log-structured filesystems aren't in use for media on which seeks are
non-zero cost because they lead to fragmentation hell in short order. You don't always want to access your files in the same order in which you wrote them; they should be clustered differently. LFSes make that distinctly nontrivial to do.
ext4 and data loss Posted Mar 13, 2009 10:18 UTC (Fri) by job (guest, #670) [Link]
Perhaps LFSes will make a comeback with the SSDs, then?
|
|||||||||||||