Originally Posted by lloyd mcclendon
hmm .. can you elaborate on this a bit ... what is 'so cool' about AIX? & why is the hardware so special? I'm really not at all familiar with it. A group we partner with at work has some things that run on AIX ... and, well, lets nicely put it this way - it doesn't seem to go very well for them
POWER and AIX is what you get when you cross mainframe technology with UNIX. It is a very unique system, but sadly misunderstood by UNIX administrators from other UNIX flavours. If your partners AIX machines are not working well, I would say its due to lack of expertise of the people managing the server.
A lot of the advantages are in the POWER hardware itself. Most of the functionality you get from the intel hypervisors like vmware are actually built into the POWER hardware. For example we could carve up our POWER machines at the bus level 12 years ago. I believe you guys call this IO passthrough. Except our implementation is decades more matured as the mainframe has been doing this for years and it shows. In fact there are many things that we were doing 5-10 years ago in AIX that people are only doing now on Intel.
The first thing you will notice about AIX is that it defines physical and logical layers throughout the entire system. In AIX all hardware interfaces have physical and logical layers. In other words it doesn't just end with having a logical layer at the storage level in the form of LVM. I believe LVM was first implemented in AIX in 1989 and it predates HP LVM by 5 years and Linux by 12 years.
On a storage level you really can't beat AIX LVM implementation which is streamlined into the system better than any other system. You cannot run AIX without LVM, in other words you can't partition a disk and throw JFS2 directly onto it. The boot partition lives inside the root volume group and is just another LV on the root PV, we have been able to LVM mirror the boot/root disks, and/or fibre boot and multipath the boot lun since before I have been working with AIX (8+ years). We also don't need a separate LOG disk/parition for LVM mirroring.
JFS2 is a world class filesystem, it supports CIO, DIO, AIO. I've never seen a JFS2 filesystem corrupt that FSCK couldn't 100% fix and JFS2 only corrupts if you do something really stupid like pull out all paths to its virtual disks while the guest is still running. We could shrink JFS2 filesystems on the fly for the last 9 years now. We also don't need to manually manage LV sizes and resize the FS to the LV. You just simply tell the system what size you want the filesystem to be. I still chuckle when I look at linux LVM because it looks so primitive in comparison.
AIX commands are remarkably consistent across the entire system, using the following prefixes:
mk="make" (eg. mklv, mkvg, mkuser, mkgroup, mksysb, mkitab)
rm="remove" (eg. rmlv, rmvg, rmuser, rmgroup, rmdev, rmitab)
ls="list" (eg. lslv, lsvg, lsuser, lsgroup, lsitab)
ch="change" (eg. chlv, chvg, chuser, chgroup, chitab)
The output of the ls commands are usable to other commands. In other words I can take the output of lsuser and apply it to mkuser or chuser without formatting the output.Edited by CaptainBlame - 3/4/13 at 1:30am