[KLUG Advocacy] AIX's Funeral March

[Adam Williams]

>>>The article mentioned "Most analysts agree that Linux is growing fast but
>>>still hasn't matched the abilities of AIX or of other versions of Unix,
>>>including Hewlett-Packard's HP-UX and Sun Microsystems' Solaris."   Where is
>>>Linux falls short?  
>>Features -  
>>  AIX offers full ACL support - Linux does too, but only recently, and
>>almost nothing supports them beyond Samba.
>>  AIX has an audited and certified security subsystem - Linux... Hmmm.
>>  AIX provides *VERY* robust LVM technology - Linux does too, again,
>>only recently.  It is stable, but do I trust it as much as the
>>implementation on AIX?  No way.
>>  AIX supports LOTS of SCSI devices - Linux support of lots of devices
>>is a pain, devices are numbered based upon discovery, rather than
>>"learned" and remembered.  That is a kludge,  and one can screw oneself
>>pretty easily.
>>  AIX supports thousands, even hundreds of thousands of threads,  and
>>the schedular doesn't go "Ahhhhhhhhhhhhh!!!!" - Linux....
>>  AIX supports SNA - Linux..... someone tried once.
>>  ulimits on AIX actually do something - Not so much on Linux
>>  AIX can motor through huge amounts of IP traffic, all zero copy -
>>Linux does pretty well, but gets stuttery in extreme conditions.
>Guess how Linux is going to get many of the above features? IBM (and perhaps
>players who have an interest parellel to IBM, like large users) is going to

Exactly.  And IBM is playing so nice it is almost unbelievable.  If you
don't track all the LVM/filesystem stuff you may have missed it.  There
where three competing system to replace the LVM1 in 2.4.x.  The two real
contenders was LVM2 (sistina) and E{something, escapes me now} (IBM). 
It was a really important call.  IBM's system had alot more
manageability, but the kernel team decided to go with LVM2 for some
technical reasons including current VFS compatibility.  The IBM team
said 'OK,  we'll begin porting our management tools now'.  Thousands of
developer hours down the drain, and not so much as a whimper.  Everyone
on those lists was VERY impressed.

>provide them! The above list was a lot longer 2 years ago, and it will be a
>lot shorter in 18 months, perhaps empty in 36.

Yep.  The 2.6.x feature list almost sounds like a bullet point answer to
the above gripes.  Hopefully it goes much better than 2.4.x did, ick.

>>As for apps. I think it is pretty tough to think of apps that haven't
>>been ported: AcuCOBOL, visiFAX, SAP, etc...
>That's right, this is an even more radical shift in the last 18-36, and was
>again done by major players who saw the handwriting on the wall. Some of them
>have been really impressed with how easy it was to port stuff, too.
>>>The area that falls short for me to switch is the some
>>>of the high end applications that we use just is not written for Linux but
>>>that is not a short coming of the os just a short coming of the developers
>>>realizing how to utilize Linux to there advantage.
>>Right, but if they port forward to current version of AIX, porting to
>>Linux from there is like falling of a log.
>IMHO the main gap between Linux and the commercial UNIX systems is on 
>high-count MP servers, which, while critical for a lot of enterpirse-
>level apps, are not a high percentage of the deployed base. They do
>reprsent a kind of sumiit, from which bragging rights to handling anything
>is seen among corporate IT types as a Good Thing.

Right, but you still don't see Linux on 100,000+ user file servers.  So
we are a bit from the summit.  I think there are actually multiple
summits.  Single-uber-compute and number-cruncher-clusters Linux seems
higher up the mountain range than in the gotta-move-lotsa-blocks
category.  The I/O seems to be the stickler.

>>>>Well, it is official.  What we've known for a long time.  Someday in the
>>>>not terribly distant future yet another ugly child of true UNIX lineage will
>>>>be lowered into its grave.
>>>>http://news.zdnet.co.uk/story/0,,t269-s2129537,00.html
>I must say that this article high a very high equivocation content, which is 

Oh yea, they *always* do that.  Anyone whose working with them knows how
to sweep it away in analysis.

>typical in some ways, of a large organization starting on a long road of sorts.
>The politics of all this are pretty clear; the AIX guys want to hold the old 
>ship together as long as possible, while the Linux proponents want to point us 
>in the direction of ultimate victory, without seeming to be disloyal to the 
>corporate line, or what custumers can bear to absorb. The result is a lot of
>rephrasing and verbal dancing around the realities of things....
> ...which seem to be that we are at the start of a Long March, which often 
>begins in considerable trpidation, with great rewards, but also great risks.
>Those of us who have been forecasting the start of the march can at least take 
>heart that it has finally and publicly started. The above technical issues 

Thats me!  But it wasn't really that hard to call.

>will br handled, just as face-saving ways to move the AIX proponents 
>off-center will be found; time and technology will hanlde a lot of this.