RE: Microsoft thwarted again -- MIDRANGE-L

On Fri, 19 Nov 2004, Walden H. Leverich wrote:

I have NEVER used EDTRBDAP, not have any of my clients.  I only use
RCLSTG prior to OS upgrades.  I think I've seen it used twice at client
sites.  I have NEVER used STRSST for anything other than hardware
maintenance.


Good for you. I know large (multi-terabyte) SQLServer installs that
haven't had to rebuild their databases too.

Let's be honest. Corruption can occur on any machine (and not necessarily only due to cosmic rays). Thus providing fix tools is a Good Thing. However, I think it is *very* unlikely that MS SQLServer provides the same level of db integrity that the iSeries does. I would also bet that any (accurate) research data comparing the two system would back me up.

I've said for a long time that the dumbest thing MS ever did was name
their desktop and server OS the same thing (Windows) There is a _HUGE_
stability difference between W2KServer or W2K3Server running on
_ENTERPRISE CLASS_ hardware, and XP or W2K running on a desktop from
Dell that cost less than $1000. Solid servers are very possible on MS
operating systems, but be prepared to spend $$$$. Reliability isn't
cheap on the iSeries, not on Windows (nor on Linux).

Perhaps MS says it because they *are* the same? AFAIK, security patches released for W2K apply for all W2K installations, enterprise or not. It is true that there is a huge difference between serious hardware and cheap crap. But perhaps the difference you are seeing is simply the hardware, and the OS is still the same? Please note that I am not a MS person so I really don't know for sure - this is just the perception of an outsider.

Extreme reliability is never cheap. But reasonable reliability doesn't have to be expensive. Why does MS make reasonable reliabilty so difficult to achieve?

There are people out there who haven't IPL'd their machine in
years, much less rebuilt the operating system.

There are a few examples of MS machines that have uptime measured in years. But doing comparisons of uptimes reveals that by far the highest number of long uptime machines are running some kind of bsd (FreeBSD mostly) or linux. Simple laws af probability say that if two systems have equal reliability, the most numerous system should have the highest number of machines with long uptimes. MS is far more prevalent than FreeBSD and linux, yet both far outnumber MS machines in long uptime comparisons.

As there are windows boxes, and unix boxes, and probably DOS boxes --
heck you might find a CP/M box that hasn't been rebooted.

But the fact that most people have no clue they exist, much less
how to use them, says what needs to be said about iSeries reliability.


Don't get into a numbers fight Joe. There are more people that know how
to use the SQLServer recovery tools than there are people that have ever
HEARD of the iSeries. The numbers say nothing about stability, they
speak of install base.

Absolute numbers probably don't mean much, but percentages do. Comparing the percentage of MS admins who have to use such tools vs. iSeries admins who do would be meaningful. Dismissing numbers outright is not prudent.

But you simply can't with a straight face say that Windows
is anywhere near as reliable as OS/400.


Watch me! Spend the SAME $$$ on your windows box that you spend on your
iSeries and you'll get the same reliability -- yes, you can spend
millions on a "PC" server. Think LPAR is new? You could partition an
ES7000 server running W2K Datacenter years ago, with dynamic
reallocation of memory and processor on the fly. It wasn't cheap though.

No amount of money spent on hardware can fix the problems that lie in MS's software. You may be able to partition the box, but that doesn't magically make IIS secure. Part of reliability is how well a system copes with malicious input. On that front MS and IIS in particular have failed miserably.

Additionally, you're fighting the wrong fight. You're looking at the
uptime of a single server, and saying that it's up 99.999 percent of the
time (doubtful BTW) and comparing it to the uptime of a single windows
box. That's an invalid comparison, and shows a lack of understanding of
the role of clustering technology in running an enterprise-class windows
installation. You shouldn't rollout a single windows server if you want
that level of stability, you rollout a cluster. And when you do you'll
get closer to 100% uptime than a single iSeries ever could.

So you are suggesting I get a cluster of crappy machines instead of one good one? Joe's suggestion of two iSeries is much more reasonable if 100% uptime is a requirement. Then you have two good machines instead of ten crappy ones. Two good machines are no more work to maintain that one good one, but ten crappy ones are ten times worse than one crappy machine.

Really, MS Windows is so poor at reliability and security that billg announced the whole company would focus on nothing else. The results so far have not been impressive. Clustering is great tech and has a place, but often it is used to mask problems that should be fixed. Clusters should be used to enhance a platform, not hide its weaknesses.

James Rich

It's not the software that's free; it's you.
        - billyskank on Groklaw