LXer: Kernel developers squabble over Ext3 and Ext4

LXer · 03-27-2009, 01:20 PM

Published at LXer:

A number of senior kernel developers, including Linus Torvalds, Ted Ts'o, Alan Cox and Ingo Molnar, have been squabbling over the sense or otherwise of journaling and delayed allocation in Ext3 and Ext4. The trigger for the discussion was a response from Jesper Krogh to Torvalds' announcement of kernel version 2.6.29, in which he described massive delays in writing out the file system cache on Ext3 file systems despite fast RAID arrays on computers with lots of RAM.

Read More...

H_TeXMeX_H · 03-27-2009, 02:57 PM

Sorry, but Linus is 100% right in what he's saying ... as usual. Basically there's a delay time between writing metadata and real data within which if the computer crashes the data will suffer certain corruption. This delay time should be as small as possible to assure data integrity. Unless you want to trade greater performance for less data integrity.