titan wrote:Supposedly Ext3 is a better performing file system than ReiserFS.
Generally speaking, ext3 is the worst performing of the current Linux journaling filesystems (XFS, JFS, Reiser4) because it's an older design (ext2) with journaling bolted on. It doesn't support extents and it has static inode allocation. They could have made it better if they didn't have the design constraint of backward compatibility with ext2, but ext4 will fix some of these problems. That's not to say ext3 is terrible, but I've found that it certainly lags XFS and JFS in performance on most workloads I personally care about*. Now, I can't say as much about either ReiserFS or Reiser4 because I don't care for them and their performance is somewhat contentious (Namesys's benchmarks vs. others) as well as the quality of implementation. ReiserFS is in the kernel but kind of abandoned and hasn't moved forward architecturally, while Reiser4 is not in the kernel, less proven, and who knows what will happen now that Hans Reiser has been convicted of murder.
One complication in estimating ext3's performance is that ext3 offers several different journal modes: full data & metadata journaling, metadata only with strict ordering between data & metadata writes (ordered), and metadata only without strict ordering (writeback). With full data & metadata journaling, ext3 takes a pretty big hit, but it's actually safer than the other filesystems. XFS and JFS only do metadata journaling and (IIRC) don't guarantee ordering, so with ordered mode, ext3 is safer than XFS and JFS and not terribly slower, depending on the workload.
Personally, I've been using XFS for many years now and keep with it, but I'd go with ext3 unless you have a reason to switch. Space efficiency on all of the filesystems isn't going to be significantly different. ext2 could be faster than the journaling filesystems (with hashed directory indexes enabled, of course), but that's because it's doing less stuff and is less safe. I don't see why anyone would run without at least metadata journaling these days, though.
* I posted some
RAID5 benchmarks a while ago with bonnie++ sequential tests with large files comparing XFS vs. JFS vs. ext3 in ordered mode for reference. Obviously this is just one single workload, and probably unrepresentative of a laptop.