Linked by Thom Holwerda on Thu 13th Jun 2013 14:35 UTC
Mac OS X The third and final WWDC product I want to talk about is - of course - OS X 10.9 Mavericks. While iOS 7 was clearly the focus of this year's WWDC, its venerable desktop counterpart certainly wasn't left behind. Apple announced OS X 10.9 Mavericks, the first OS X release not to carry the name of a big cat.
Permalink for comment 564653
To read all comments associated with this story, please click here.
RE[2]: Interested in compression
by malxau on Thu 13th Jun 2013 19:13 UTC in reply to "RE: Interested in compression"
Member since:

I don't. When NTFS crashes it can be far more subtle, let's say a cluster bitmap corruption (i.e. the bit that tells the filesystem what space is used and what is free). Ever seen that?

So firstly, I work on Windows filesystems (see bio.) I get to see a lot...

It looks, on the users' end, like files are randomly disappearing as they're being erased by new files without the os realizing it.

I don't think this is what happens. Imagine a cluster is in use by FileA but (for whatever reason) the bitmap doesn't record it as allocated. Now FileB gets allocated the same cluster. FileA is still there, because the system didn't detect the condition. It's just that FileA's contents end up the same as FileB's contents, because they're using the same block. And if only one cluster was in this condition, the party stops here, because the cluster is now marked in use so only these two files will share it.

Given how bad this is, NTFS works hard to avoid it, both at the design level and by making pessimistic assumptions in event of detected corruption. As above, I see a lot, and I don't see this often. the ext? filesystems and zfs, it is self-correcting as there are multiple copies of the superblock and inconsistencies are checked against them.

ZFS has checksums so it can detect corruption in its structures (assuming, of course, that the corruption was not generated by a bug), and use a good copy if one exists. ReFS, Microsoft's next generation file system, also checksums its metadata, and when used with Storage Spaces can also locate a good copy of data if one exists. Ext2/3 do not. Ext4 checksummed its journal and more recently metadata too; I don't know how it would be able to recover though, since AFAIK it doesn't keep redundant copies, so it has nothing to recover from.

Reply Parent Score: 5