NT vs UNIX --- is it as good?
Barry Merriman
Subject: NT vs UNIX---is it as good?
Summary:
Followup-To:
Distribution: world
Organization: UCLA Dept. of Math, UCLA Inst. for Fusion and Plasma Research
Keywords:
I'm from the UNIX world, but I'm considering switching to
NT (I don't think too much of UNIX + X...it was state of the
art years ago, but is too slowly evolving...).
Is NT really comparable to UNIX in functionality and
robustness, in a networked environment?
How difficult is NT system admin (since I would probably
do my own admin)
How well does NT fit in on a heterogeneous network
(say with other UNIX machines)?
Does the Windows GUI provide a good graphical adbstraction
for the os, or does it poke through alot and require dropping
down to the command line level and/or using a lot of hacky,
poorly coordinated graphical tools?
What workstations are NT available on (DEC...)? And
is the Windows software as low cost and plentifiul
for these platforms as it is for the INtel one---or
is INTEL a greatly preferred platform for NT?
Is there a cost/availability difference between
Windows 3.x and Windows NT software?
Is NT the way of the future for MS, or are they
goinf to circumvent/leapfrog it with
windows 3x -> win95 -> Cairo?
In Daytona (= NT 3.5?) available now? Has it been well
recieved?
Dan Sullivan
In a nutshell it is the functional equialent of Unix and has its robustness.
However it works in a different way. Nodes are always computers, never
terminals. Conceptually it has all the complexity and headaches of maintaining
a Unix network, but its tools for doing this are well thought out, GUI,
and fairly easy to use.
It seems to work quite nicely in a mixed network. This first mixed network
had SCO and Suns on it. Literally 10 minutes after hooking up the ethernet
I was rsh'ing and telnet'ing to the other machine. NFS is available from
third parties also.
With the release of 3.5 there is no performance penaltity in runing
regular (16-bit) application. In general there are few comodity NT
native applications. However from a developer/user point of view there
is not difference between Win95 and NT3.5. NT3.5 has added capabilities
and requires more resources. Except for very special programs anything
that runs on Win95 will run on NT3.5. The executables (and obj's and lib's)
are the same binaries. Win95 is Intel only and in general can run even the
most bizarre, touch the hardware DOS or Windows applications. NT provides
much more robustness, scalability, finer control over access, and real
services (like daemons in Unix), and an advanced graphics API (functional
equivalent of Display Postscript though quite different in use).
Pays your money, takes your
choice. If you are talking mission critical, robust enviornment you are
talking NT. If the robustness level current provided by 16-bit windows
is sufficient (and it is for many people now) then Win95 is the answer.
In addition mixing 16-Windows, Win95, NT, and Unix works well.
In my view NT provides all the capabilities and robustness of Unix, but does
so in an enviornment familiar to most users (Microsoft Windows).
If you go the NT route you want NT 3.5. I have been running a NT Server 3.5
and NT Workstation 3.5 on my network for three or four weeks now.
Before you decide against NT you should at least set up an NT Workstation 3.5
on your existing Unix network. I think you will be very pleased.
Dan
Sarir A. Khamsi
In article <1994Nov27.0...@midway.uchicago.edu> ba...@math.uchicago.edu (Barry Merriman) writes:
barry> I'm from the UNIX world, but I'm considering switching to NT (I
barry> don't think too much of UNIX + X...it was state of the art
barry> years ago, but is too slowly evolving...).
barry> Is NT really comparable to UNIX in functionality and
barry> robustness, in a networked environment?
Yes, in a nutshell. You get file and printer sharing, TCP/IP and its
utilities, NetBeauooeoeoe (sp?), IPX for Netware, SLIP, PPP and other "stuff".
barry> How difficult is NT system admin (since I would probably do my
barry> own admin)
Very, very easy. It's really point-and-click.
barry> How well does NT fit in on a heterogeneous network (say with
barry> other UNIX machines)?
You get TCP/IP as mentioned above, but you don't get NFS, SMTP or a
telnet demon. Why, who knows?!! You can get these from others, but in
my opion, Microsoft should have!!!!!!
barry> What workstations are NT available on (DEC...)? And is the
barry> Windows software as low cost and plentifiul for these platforms
barry> as it is for the INtel one---or is INTEL a greatly preferred
barry> platform for NT?
Intel x86, DEC's Alpha, MIPS and soon PowerPC. Also roumors of
UltaSPARC and PA-RISC, from what I've heard.
barry> Is there a cost/availability difference between Windows 3.x and
barry> Windows NT software?
Some NT software costs more and there is a ton more Windows 3.1
software out there, but more are starting to write NT code (Win32 API)
and some are writing Win32s (which runs on both Windows 3.1 and NT).
barry> Is NT the way of the future for MS, or are they goinf to
barry> circumvent/leapfrog it with windows 3x -> win95 -> Cairo?
It appears that Win32 is the future for MS, but I would not even start
to predict what they are going to do.
barry> In Daytona (= NT 3.5?) available now? Has it been well
barry> recieved?
Daytona = NT 3.5, is available and has been well received from what
I've seen. It's faster, does more and cost less (than NT 3.1).
Riri
--
====================
Sarir (Riri) Khamsi
kha...@ll.mit.edu
w:617-981-4011
h:617-861-7440
====================
Olav Tollefsen
kha...@ll.mit.edu (Sarir A. Khamsi) wrote:
> barry> What workstations are NT available on (DEC...)? And is the
> barry> Windows software as low cost and plentifiul for these platforms
> barry> as it is for the INtel one---or is INTEL a greatly preferred
> barry> platform for NT?
>
> Intel x86, DEC's Alpha, MIPS and soon PowerPC. Also roumors of
> UltaSPARC and PA-RISC, from what I've heard.
I think you should take Windows NT on SPARC and PA-RISC with a grain of salt.
Someone will gladly spread this roumor to prevent people from leaving their processor
architectures. Someone even claim that a port of Solaris to Alpha is in the pipeline to
prevent people moving to OSF/1 on Alpha. Get such statements in writing from the
involved companies before you make any decisions. Anyway, Windows NT is avialable
on the fastest microprocessor in the world *NOW*.
--
Olav Tollefsen
Digital Equipment Corporation, Norway
233 MHz Alpha AXP + Windows NT 3.5 = RAW SPEEED!!!
Guy Harris
>In article <941201163...@LL.MIT.EDU>
>kha...@ll.mit.edu (Sarir A. Khamsi) wrote:
>> Intel x86, DEC's Alpha, MIPS and soon PowerPC. Also roumors of
>> UltaSPARC and PA-RISC, from what I've heard.
>
>I think you should take Windows NT on SPARC and PA-RISC with a grain of salt.
>Someone will gladly spread this roumor
For UltraSPARC, it's not a rumor; Sun and Intergraph have announced that
Intergraph will be porting Windows NT to some UltraSPARC-based
machine(s). (I don't remember whether those are future Sun machines, or
future Intergraph machines.)
Now, just because they've said this is going to happen doesn't mean it
*will* happen, of course.
Ian Kemmish
>involved companies before you make any decisions. Anyway, Windows NT is avialable
>on the fastest microprocessor in the world *NOW*.
>--
>Olav Tollefsen
>Digital Equipment Corporation, Norway
>233 MHz Alpha AXP + Windows NT 3.5 = RAW SPEEED!!!
>
AND....
233MHZ Alpha AXP + OSF/1 == CIVILISED SPEED.
... where CIVILISED == RAW plus 20%
:-):-):-):-)
--
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Ian Kemmish 18 Durham Close, Biggleswade, Beds SG18 8HZ
i...@eeyore.dircon.co.uk Tel: +44 767 601 361
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Ian Kemmish
>In article <1994Nov27.0...@midway.uchicago.edu> ba...@math.uchicago.edu (Barry Merriman) writes:
>
>barry> I'm from the UNIX world, but I'm considering switching to NT (I
>barry> don't think too much of UNIX + X...it was state of the art
>barry> years ago, but is too slowly evolving...).
>
>barry> Is NT really comparable to UNIX in functionality and
>barry> robustness, in a networked environment?
They are some very different design decisions in key parts of
the kernel. Some of these decisions affect the performance
of different classes of apps. in umdocumented ways. This is
particularly true of what one might term `workstation class'
applciations, that seem `resource hungry' to people from the
PC world. If you just want to run PC-type applications, then
NT is probably as good a choice as UNIX+X, but if you wantt
to run large programs, then you're better off staying where
you are.
Think of NT as a way to turn a mips or an Alpha into a fast
PC, not as a way to turn a PC into a workstation....
Nik Simpson
they do say it will happen., Don't expect it to run on ANY existing SPARC hardware.
--
Nik Simpson n...@iquest.com
Henrik Nordborg
From: ia...@tdc.dircon.co.uk (Ian Kemmish)
Newsgroups: comp.os.ms-windows.nt.misc
Date: 4 Dec 1994 16:11:34 -0000
Organization: The Direct Connection
Think of NT as a way to turn a mips or an Alpha into a fast
PC, not as a way to turn a PC into a workstation....
--
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Ian Kemmish 18 Durham Close, Biggleswade, Beds SG18 8HZ
i...@eeyore.dircon.co.uk Tel: +44 767 601 361
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Would you mind giving some sources for your statements. Otherwise I must
regard them as lies, since my experience is that NT is at least as fast as
UNIX on most hardware.
I have not tried NT on any systems but Intel, but the results there are
striking. A 486DX/33 MHz is as fast as a Sparc 2. A Pentium outruns a Sparc 20
with ease. To compare compilers is even more fun. Compile with GNU and you
need a fast computer :-). It is amazing how much better code the Microsoft
compiler delivers. Not to mention that one gets a nice developer environment
with it for a reasonable price.
Unless one is very dependent on UNIX programs which are not (yet) available
for NT, I see no reason to buy UNIX anymore.
--
--------------------------------------------
Henrik Nordborg
Institute of Theoretical Physics
Swiss Federal Institute of Technology (ETH)
CH-8093 Zurich
phone: +41-1-633 25 73
fax: +41-1-633 11 15
hen...@itp.phys.ethz.ch
--------------------------------------------
jim frost
>Would you mind giving some sources for your statements. Otherwise I must
>regard them as lies, since my experience is that NT is at least as fast as
>UNIX on most hardware.
I concur, and for some application domains it's considerably faster.
One application I wrote, a threaded object request broker, performed
better on a 486DX2/50 under NT 3.1 than on a 4-CPU Sparc 10 under
Solaris 2.3+jumbo. It had both better throughput and far better
predictability. (This is probably not the case under Solaris 2.4 --
they fixed a lot of I/O performance problems, although I think there
are still problems with thread scheduling in common threaded
architectures.)
>Unless one is very dependent on UNIX programs which are not (yet) available
>for NT, I see no reason to buy UNIX anymore.
There are places where UNIX is a far superior choice. For instance,
NT is not an appropriate system if you want timesharing. At a minimum
it needs distributed GUI support, resource quotas would be very nice,
and the scheduler really needs reworking.
If you want client/server, though, NT is a pretty good choice.
jim frost
ji...@world.std.com
--
http://www.std.com/homepages/jimf
Guy Harris
>I have not tried NT on any systems but Intel, but the results there are
>striking. A 486DX/33 MHz is as fast as a Sparc 2. A Pentium outruns a Sparc 20
>with ease.
How fast is a 486DX/33 running NT relative to the same 486DX/33 running
SunOS 5.4? That'd be a better comparison of NT and UNIX (or, at least,
NT 3.<whatever you ran> and SunOS 5.4). (I'm assuming here that Solaris
2.4, the OS component of which is SunOS 5.4, is currently available for
PeeCee's. If not, said comparison would have to wait, as I think the
previously-available Solaris release for PeeCee's was 2.1, which had, I
think, its problems.)
(I.e., is this a comparison *solely* of OSes, or does it also compare
the processors? I think reasonably fast x86-family processors, at least
at present, and on integer stuff, are competitive with Sun's current
SPARC processors, except perhaps for the fastest of them.)
(NOTE: I'm not predicting that NT would lose the comparison; I'm merely
noting that the hardware platforms are different, and that it may be the
case that a reasonable PeeCee platform might well perform better than a
SPARCstation, and that the results you're seeing might be in part due to
the hardware.)
What were the things you were doing when you made this comparison?