Linux now 11% of IBM MIPS sold

Stephen D. Williams
Sun, 27 Jan 2002 17:30:12 -0500

And what OS, pray tell, is less of a 'toy' than Linux (in every sense, 
which is what you imply)??

There are many axii to measure an OS and on many of those, I would put 
Linux before BSD*, Solaris, HPUX, etc.  Each has its strengths, all 
excel in one or more areas.  Still, it's clear that Linux has a growing 
and assured future.

You want stability?  Probably HPUX, followed by Solaris possibly (used 
to crash on me plenty though).
You want secure OS capabilities?  Until recently I would have said HPUX, 
Solaris, but with HP's Secure Linux, NSA's secure releases, and the 
existance of at least one more commercial secure Linux offering, I would 
have to analyze it more.

You want the ultimate in stability?  Plan your overall application and 
system to expect failures, possibly multiple failures, and make sure 
that things failover in millisecond time.  You want clusters in other 
words, and not some lame 1% idea of a cluster, but end-to-end clustering 
methods that allow you to just walk up to any part in the system, yank 
it, replace/upgrade/fix, and pop it back in.

It is through the use of clustering features in addition to the existing 
stability and new support for hot-swap IO channels (Linux supports 
industrial PCI hot swap Ethernet, etc. now), hot-swap CPUs, cc-NUMA, 
etc. that Linux will overcome and outpace more scalable and stable 
platforms.  All of this is still accelerating as far as I can tell.

In most application situations, it's just not that important to have 
millisecond failover.  It costs extra no matter what hardware/software 
you start with.

Mainframes should have a cluster architecture internally but provide 
either SSI (Single System Image) or seamless multi-system views of 
computing resources.  IBM is doing the latter and possibly some of the 
former with their Linux Mainframes, which is just fine with me.

What's your app, what OS do you use, and why is it more stable than Linux?


Adam L. Beberg wrote:

>On Fri, 25 Jan 2002, Joseph S. Barrera III wrote:
>>You can agree or disagree with the numbers, but the point is that "the
>>whole purpose" proposed therein is TCO, not increased reliability at the
>>OS level.
>Right, but when your machines are crashing and your customers are freaking
>out, that's not exactly good for business. Since basicly all my work is with
>products that are more then $19.99/month, I can say they don't care much
>about TCO, they want _ZERO_ downtime. A 7-digit check is no big deal, but 7
>milliseconds of downtime means heads will roll... Linux is still a nice toy
>OS, and you get exactly what you pay for.
>- Adam L. "Duncan" Beberg