It’s all over the place, Hans Reiser has been found guilty for murder. Whether he did it or not, the situation is really quite awful. I’ve been a long-time reiserfs user and fan. Mostly for performance, but also because it is so tolerant to power failures. I leave most of my computers on all the time, or work from laptops. This means that when there’s power failure, my machines usually don’t have a clean shut down. Sometimes I set laptops to hibernate when they are low on power, but usually I just let them run flat. ReiserFS has been great for this the last 4 years or so I’ve been using it. It’s also great when deleting lots of big files, or even copying large amount of small files and accessing big directories.
I’ve been giving ext3 a shot again, and found that it can run pretty much just as good as reiserfs, by enabling writeback mode. From the mount manual page:
writeback
Data ordering is not preserved - data may be written into the main file system after its metadata has been committed to the journal. This is rumoured to be the highest-throughput option. It guarantees internal file system integrity, however it can allow old data to appear in files after a crash and journal recovery.
I’ve found this to be true, perfomance is good, and I’ve had a few bad shut downs with no ill effect. The man page does warn that you can lose any recent changes to files when a bad shutdown occurs. ReiserFS has the same symptoms with bad shut downs, I guess it probably does writeback data as well. You can enable this by adding the data=writeback option in your fstab, or by adding an appropriate kernal parameter if you’d like to test. I suggest you peak at the man pages for fstab and mount before making any changes. For me, this has at least brought ext3 to the level that I don’t care too much about reiserfs anymore. I would enjoy the built-in compression in ReiserFS4, but hopefully we will see that in future extfs implementations.
Browse Feeds
YouTube Channel
Technorati Profile
LinkedIn Profile
Facebook Profile
LastFM Audioscrobbler
Twitter feed
FSF Member
Mugshot Page
Identi.ca Feed
Whoisi Page
Digg Profile
Plaxo Pulse
May 1st, 2008 at 9:51 am
i) Just because Hans Reiser has been found guilty (and I’m guessing that this is just the start of the usual appeal process) doesn’t mean that the Reiser 3.x code has ‘gone bad’ in some way, and you should now automatically avoid it.
You might have some kind of moral or ethical concern about, in some way, encouraging or supporting someone found guilty of murder, but I can’t really see how that would work. Does me using it make it more likely that a murderer gets off? I can’t see it.
And anyway, it wasn’t a one man effort but a team effort.
ii) You could say that Reiser4 was now less likely to get into the kernel, but I’m not sure that’s true. Hans Reiser has been famously prickly and difficult to get on with, so if the team holds together, maybe the absence of Hans makes it more likely to progress, not less. (And there has been an early version in Xandros some time ago, so it must be up to some vaguely workable state already.)
iii) I will re-assses the situation when ext4 seems ‘ready’, but ’till then I’m sticking. I’m comfortable here.
iv) When talking about performance, one thing that makes it difficult to compare is that performance does depend on configuration. In particular ‘noatime’ can make a significant difference under some load profiles.
May 1st, 2008 at 10:11 am
Hi Mark. Yes, I agree with everything you say above. I didn’t mean to imply that reiserfs was now ‘bad’. I suppose I should’ve said something to that extent in the post. I still look forward to Reiser4, and will certainly try it out when it’s available.