Talk:Soft updates

This is the name as McKusick writes it. I thought the article should be at the right place and fixed all the links to it except someone's user page. - Taxman 22:44, Dec 14, 2004 (UTC)

Are softupdates more flash SSD friendly than journalling? It seems that they should reduce the number of writes (no log) and hence provide a better wear. —Preceding unsigned comment added by Cyril (talk • contribs) 20:40, 27 March 2008 (UTC)

For FFS-like file systems, yeah, I'd guess so. But overall, FFS and the like aren't very flash friendly, even with softupdates. Even though softupdates will delay and aggregate meta-data updates, the FFS meta-data is still scattered into smaller blocks than optimal for flash. IIUC, a log-structured FS would be way better since flash doesn't have the high seek time of HDDs. Adding some softupdates-like functionality on top of that would be even better (like not simply queueing writes, but only writing the differences needed between the in-core cache and the on-disk flushed state), but probably overkill. Most of the unnecessary writes are prolly restricted to /tmp, and you can deal with that one other ways. —Preceding unsigned comment added by 24.200.77.59 (talk) 21:57, 31 March 2008 (UTC)