On Fri, Oct 29, 2010 at 01:17:23PM -0700, Leslie S Satenstein wrote: > In a few earlier postings I indicated the problems I had with EXT4 and a hard > drive that was about 75% full (320gig at 75%). Performance was such that > simple uploads or downloads crawled to a 10k / second pace, or even stalled > for long periods of time.
I use ext3, and have not had this problem. But then, I don't have any nearly full USB drives to have the problem on either. > > I read an article about btfrs and how it is designed as copy on write. That > is, write the new data, then change the B tree to point to the new sub-tree > of data. I thought that was the pronciple behind the reiser file system. I also heard that the reiser file system was hell to repair if something *did* manage to go wrong. I'm not sure of the details, but I hope btrfs is better. -- hendrik > > Ergo, if someone inadvertantly pulls out the drive's plug from the USB > physical port, we are almost guaranteed to not have drive corruption. There > may be some data loss, but no drive corruption. That to me is most important. There might be physically part-written blocks if the drive relies on power that's suddenly gone. I'm not sure how part-written blocks work ont when the time comes to rewrite them ... I gather it's the kind of thing hardware-level disk-checkers look for. > > I am not certain about storage efficiencies. Would 10gigs of data with EXT4 > use less diskspace than 10 gigs of data stored in btfrs format. (What's a > gig when terrabyte drives sell for around $60.00. No idea. But if I recall correctly, when I changed from an ext2 file system to a reiser file system total available storage went down slightly. It wasn't a big deal. -- hendrik _______________________________________________ mlug mailing list [email protected] https://listes.koumbit.net/cgi-bin/mailman/listinfo/mlug-listserv.mlug.ca
