Re: Raw devices de-support

From: Mladen Gogala <gogala.mladen_at_gmail.com>
Date: Sat, 19 Nov 2011 01:34:05 +0000 (UTC)
Message-ID: <pan.2011.11.19.01.34.05_at_gmail.com>



On Fri, 18 Nov 2011 15:54:25 -0800, John Hurley wrote:

> # I am not at all convinced that Oracle is doing a good thing by
> releasing their own distribution. Maybe they should have joined the BSD
> camp?
>
> Ahh ... sorta/kinda like Occupy BSD?

Not quite. My understanding of the "Occupy" movements doesn't have much to do with the technical merits. I'm not really a fan of that movement without a message, but I don't think that's comparable with this situation at all.

My point was twofold:
1) The decision to de-support raw devices was Linux driven, it wasn't

   some marketing whim by Oracle Corp. Linus Torvalds is more    probable culprit than Larry Ellison.
2) Red Hat 6.x and derivatives leave much to be desired and have very few

   advantages over RH 5.x

>
> All my systems are now up at OL 5.7 and doing quite well thanks. Don't
> have any active plans for 6.x yet ...

The only real advantage of OL 6.x is the possibility of I/O accounting. Commands like "iotop" and "perf" are not available in kernels before 2.6.20. That's another brilliant characteristic of Linux: it was an enterprise level OS without I/O accounting. BTW, Ext4 is also a piece of work: extent based file system without defragmenter. It is not possible to do something like this for Ext4 file system:

[root_at_medo mgogala]# xfs_fsr -v /home /home start inode=0
ino=2925526
extents before:4 after:1 DONE ino=2925526 ino=2925527
extents before:4 after:1 DONE ino=2925527

ino=2947655
ino=2947655 already fully defragmented.
ino=2947667

extents before:2 after:1 DONE ino=2947667 ino=272245176
extents before:2 after:1 DONE ino=272245176
ino=809065277
ino=809065277 already fully defragmented.
ino=809065296

extents before:2 after:1 DONE ino=809065296 [root_at_medo mgogala]#

When the machine is used for virtual machines and databases, defragmentation becomes critical because fragmentation will turn your sequential I/O read into random I/O operations, thus destroying the I/O performance and your virtual machine performance. XFS is, of course, a separate option on RHEL and has to be paid separately.

-- 
http://mgogala.byethost5.com
Received on Fri Nov 18 2011 - 19:34:05 CST

Original text of this message