[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: EXT2 - EXT3 - Reiserfs

Hash: SHA1

On Thursday 14 June 2001 13:05, Rick Sivernell babbled:
> Doug
>   Educate me please. What is the diff between ext2 & 3. I am using reiserfs
> now and do like it. But I do like to have the stuff going, if I can.

Be warned, I know nothing about reiserfs other than what I've read.

short version:
ext3 is a further development of the ext2 codebase. ext2 was designed in such 
a way that certain core functionality can be easily replaced. There are also 
numerous hooks for future enhancement to be included. It is 100% backwards 
and forwards compatible with ext2 and ext4 (should there ever be a need). It 
is possible (I do it myself) to boot off a rescue disk that supports only 
ext2 and have it mount your ext3 partitions. You simply don't get the 
journaling features. However, when you do then mount under a kernel with ext3 
support, it will scan out the disk, updated the journal if needed, and you 
are 100% consistent from then on. Being an evolution of ext2, all the new 
features being kicked around on linux-fs-devel for ext2 will be merged right 
into ext3 without a hitch (if they make sense). Note that ext3 is a metadate 
only journalling fs. this puts it in the lightweight category. However, it is 
more than sufficient more home use, and most servers (except file/app 
servers). What this means in effect is that every fs operation becomes an 
atomic transaction. the I/O operation is recorded in the journal immediately, 
then the disk executes the I/O when it's convenient. then and only then does 
the the transaction get removed from the journal. so, should your system 
crash, the journal can be scanned. if something is in the journal, but not on 
disk, the I/O can be completed f possible, or rolled back and the entry in 
the journal deleted. the journal is all that is scanned (hence it is DAMN 
quick), where under ext2, fsck must scan the whole disk to find 
inconsistencies. under ext3, the inconsistencies are in the journal..

reiserfs is a 100% new fs from the ground-up. It is b-tree based, and is a 
middle-weight journaling fs. I don't think it does actual data journaling at 
this point (I might be wrong), but it definately does more than just metadata 
journaling. reiserfs has a lot of different performance features due to its 
b-tree heritage. Note though, that it is not optimized for the "low-end" 
cases. B-tree's are a little bit overkill in that department. You abslutely 
cannot use a reiserfs unless you have a kernel that supports it. It is 
well-suited to mid-to-heavy I/O machines (file/app servers. storage area 
network devices, etc).

so, what is a heavyweight journaling fs? Linux doesn't have one yet. SGI's 
XFS, and IBM's JFS (assuming all their features come w/ them) will be. But, 
except for the big-ass machines you and I will probably only see in the 
datacenter at work, they are overkill.

For a more detailed (though thoroughly understandable) run-through of ext3 
(in regular English not highly technical speak or code), I *HIGHLY* suggest 
you run (dont walk) over to 

Stephen does an excellent job of explaining the whole philosophy behind ext3.
Also, please note that I've CC'd the ext3 list so that they might jump in to 
clarify anything I glossed over or misinterpreted.
- -- 
Douglas J. Hunley (Linux User #174778)
http://hunley.homeip.net/	http://linux.nf/	

Miss Wormwood: What state do you live in?  
Calvin: Denial.  
Miss Wormwood: I don't suppose I can argue with that...
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]