Re: What is a Block Device?

On Sun, 19 Dec 1999 12:57:57 -0200, "Andre Whittick Nasser" <awnasser_at_openlink.com.br> wrote:

>The term journaled refers to the fact that it keeps a log of recent
>operations on the filesystem so that is can redo the last changes and keep
>the filesystem consistent, roughly similar to the redo log in Oracle.

Actually, what JFS does is keep a "redo log" of changes in the i-node structures of UNIX file systems. It does not keep a redo log of written data files, nor does it need to. Its main function is to avoid the lengthy "fsck" operations once the system is rebooted after a crash. It just applies the i-node (or whatever other structure it uses) log rather than try to check/rebuild the lot like "fsck" does.

>
>I have read some articles, and it seem a trend, saying that becasue of the
>overwhelming improvements on HD technology,
>the difference in speed between raw/block devices and filesystem does not
>make up for the trouble of managing awkward LVM.
>And it makes sense to me !

Doesn't to me. It seems that since raw operations are direct to disk, they would be the ones taking advantage of higher speeds of later drives. JFS and Veritas and others just add overhead, regardless of the speed of the disk. So far, it seems to me to be a subterfuge for people who can't figure out how to work with a LVM. They used to be hard to understand, but the latest versions are easy as pie. So I stick with raw whenever I can and have never regretted it.

Actually, one of the best ways to speed up a write-intensive ORACLE instance is to put the redo logs in raw and the rest of the datafiles in filesystems. Since the raw write is much faster than anything in filesystems so far and redo logs are mostly written-to, that gives us the best of both worlds.

>
>Doug Cowles escreveu na mensagem <385C3C05.4860A348_at_nospambigfoot.com>...
>>How does that differ from a JFS?
>>
>>- Dc

FWIW, YMMV, etc,etc.

Cheers
Nuno Souto
nsouto_at_nsw.bigpond.net.au.nospam
http://www.users.bigpond.net.au/the_Den/index.html Received on Mon Dec 20 1999 - 04:04:51 CST