NT vs UNIX workstations for engineering:
... However, I think the crash argument is overargued since occasional
crashes, although undesirable, are (1) a function of the applications
you are using [how well-written are they], (2) your NT workstation
configuration, and (3) not as productive as they seem..
That depends on your application. I don't want my system crashing
while I'm running a simulation, even if I can do a restart
from the last simulation state written to disk. My experience has
been that if a system crashes enough times, eventually, you're going
to start to have disk and OS problems. I had this happen with one
NT installation. I had to reinstall NT even though I had Emergency
Repair disks. No fun.
My perspective is from dealing with very large medical images
which require lots of image processing, graphical manipulation,
and various types of optimization.
I assume this is the simulated annealing you referred to. These
really run in less than a minute on a PC? My only (minor) exposure
to the use of simulated annealing has been in artificial neural net
(ANN) training. Those tend to be ghastly "pain in the rear" problems
that take substantial computational effort. I don't fiddle with
ANNs much now; I saw too much hype and misapplication. In fact ANNs
remind me of stocks in a way. I would see people defending the use
of ANNs for situations in which they were clearly not the best
or even an appropriate solution method. People would rationalize
and hype them alot -- much like stocks. They'd stick by their guns
no matter what the facts seemed to suggest. (Don't get me wrong;
there are of course many excellent applications for ANNs. I'm
getting off topic though.)
I've found that NT crashes can be very tied to the type of
workstation you are using.
I agree with this statement. But this can be interpretted two
ways. On one hand, it can be used to explain why NT is flakey on
some systems but better on others. On the other hand, I think
you can argue that this is one of the advantages of going with Sun.
You know what hardware you're going to be running on (assuming
you go with Sparc and not x86) and you know it'll be reliable.
think NT 4.0 is better than NT 3.51. Hopefully, NT 5.0 will
be even more bulletproof.
Someone correct me if I'm wrong, but I thought that NT4.0 was
actually less bulletproof than 3.51. As I understand it, certain
stability features (for lack of a better term) had to be relaxed
in order to accomodate the Win95 style shell (GUI). As for NT5.0,
I assume it will be an improvement, but won't Solaris have
also improved during the same time period?
More importantly though, I think you also have to consider the
richness of the environment. And there is a definite trend in
porting all kinds of applications to NT. NT has probably the best
software development environment available with multiple
vendors all provide excellent products.
I think this is the best argument to be made for WinNT. The fact
of the matter is that there is more software available for
NT than for Solaris. This is why I have an NT box on my desk along
with a Sun box. For word processing, spreadsheets, presentation
preparation, and such, there just isn't as much software available
for Solaris. I'd say this is the biggest road block to Solaris
gaining wider acceptance in the lower end market. Then again,
the low end market isn't where Sun makes their money right now.
One thing is for sure -- this battle will be raging for some time
to come (IMHO).
--Batskinner |
