[olug] article i wrote ;-)

[Brian Roberson]

http://olug.org/article.php?sid=8&mode=&order=0

and for those that wish not to open a web browser:

      lvm,journaled file systems, Linux is getting better and better...
      Posted on Tuesday, October 24 @ 23:49:57 CDT
       This really is not news per say,
      I just thought I would share my recent experience with these new ( to
linux ) items....


      Recently, I found the need to use an lvm and wanted to experiment a
little with the journalled file systems (JFS) available to linux. As most of
you may know, I have always been a big advocate of the the RedHat
distribution, however I may soon be a convert... not necisarily to another
distro, but possibly back to my roots, basically rolling my own distribution
:-) In the mean time, I've been messing with Suse ( http://www.suse.com )
Why.. oh why you ask? these two simple reasons... it comes with the linux
lvm ( http://lvm.msede.com/lvm/ ) and reiser filesytem support
 http://www.namesys.com/ ) reiserfs is a journaled file system, Which I've
found to be alot faster than ext2, in quite a few situations. ( I'll leave
that up to you to research ) For those of you who do not know what a
journalled filesystem means, A quick and dirty explanation is that JFS not
only has a standard file allocation table (FAT) but it also contains what is
called a journal, a low-level logfile if you will, this is essentially a
blow-by-blow pre-execution log of every instruction that the disk is doing.
what does this do for you? faster crash recover. what will happen in the
event of a system crash.. hmm that's bad... lets say the power fails
instead... anyways, on remount of the disk that has the JFS, the log is
replayed, instead of fsck'ed.. if there are any error's physically on disk,
the log is able to either continue the write/execution, basically recover
from a mid-write power off situation. lvm is the other beautiful thing I've
been playing around with. I am by no mean's a "virgin" to lvm/jfs systems,
as I've used other commercial *NIX lvm/jfs software quite extensively
 IBM/HP-UX, which is veritas, Solaris etc... ) What I have found with the
linux port is that the concept's are all the same ( create a volume
group(VG), assign physical disk to a VG and then create logical volumes(LV)
inside this new VG, create the filesytem on this new LV and you are set. I
bet you are asking.. what benefits does this give me? well, first and
formost, you get growable LV's , you can grow them as big as your VG is, you
can also span multiple physical media to creat one big LV, and not to
mention the ability to stripe ( similar to raid ) accross multiple disk's to
increase your I/O bandwidth ( reading from 2 seperate ide/scsi channels is
faster than 1 ) I bet you are asking, so I will continue... can you use ext2
on lvm? sure, however you've basically nullified almost every point of using
a lvm, the on-the-fly gorowing of vg,lv, and fs's.. reiser comes with a
handy "resize_reiserfs" command that allows you to grow the file system on
the fly, just as you would the vg, and lv.

      this is solely my viewpoint and should only be taken as that
      I will not be held responsible for lost data
      A full system backup will always save your ass....
      My perspective of the lvm/reiserfs sollution, on a scale of 1-10: 7
      (1 being extremely alpha code and 10 being rock solid )