As you are probably aware, the big battle in the world of Network Operating Systems has been between Novell's Netware, and Microsoft's Windows NT Server.
Without a doubt, Novell has been the industry leader for many years, having been shipping products to interconnect small computers since the early 1980s. My first experience with Netware was in the summer of 1985, with Netware 86 (yes, 86. Not 286, not 386...it would turn a standard 8088 PC with 640k of RAM into a file server), and by this time, they had an established reputation already. To keep things in perspective, Apple had just released a year earlier a curious machine called Macintosh, 8088-based computers were standard issue, and the 6 MHz 80286 was the cutting edge, really cool, super fast system. 10M and 20M hard disks were a cool upgrade from floppy only systems, but far from universal. And, Novell was tying them together, and amazingly, there is much in common with modern Novell systems to that old Netware 86 system I was playing with in 1985.
Microsoft is a comparative newcomer to the "Network Operating Systems War", although they do have a history with networking that goes back before Windows NT. Before NT, Microsoft introduced Windows for Workgroups, basically MS Windows with built-in peer-to-peer networking. Before that, Microsoft had Lan Manager, a joint development with IBM, which ran on the OS/2 platform (back when OS/2 was a joint MS/IBM project). Before that, I believe, but this is just from memory, they had a hand in the IBM PC-Network, a package which, inspite of very little market penetration, set a number of standards in the industry, including NETBIOS, which still plague us today.
The general industry perception is that Novell is a "has-been". Microsoft Windows NT is where everyone is going.
I often get people asking me if they should switch to NT, and I ask them why they think they should. The answer: "Well, isn't everyone else?" The reply: 1) No, they aren't. 2) even if they were, how does that mandate that you should?
No, not everyone is switching to NT. Netware still holds a very significant (though, admittedly shrinking) market share. There are more computers networked by Netware than by NT, and that number is growing (although, the market share is shrinking). Many companies which had invested huge amounts of money into NT are discovering it just doesn't work as advertised, and some are switching back to Netware, and if you consider the amount of egg on the face of an IS director who would say "Oops, we goofed", that is a pretty remarkable statement.
Even if everyone else in the world is switching to NT, should you?
Do you really want to be the last person on Netware? What if Novell
goes belly-up? Let's look at some of the arguments:
Further, even if you believe this, Novell now offers "Pure IP" on their
"new" Netware 5 package. Novell likes to point out that Microsoft's
implementation of TCP/IP is actually just a TCP/IP package wrapped around
a NetBUI packet, whereas when Novell embraced TCP/IP, they did it for real.
Now, if I were to set up a web server, I would probably choose NT over Netware. I'd choose Linux over either. I'd choose FreeBSD over Linux, and I'd choose OpenBSD over all of them. I can't imagine a real life situation where I would consider trying to run my business off the same file server I was trying to use as a web server. This concept of "interoperabilty with the internet" is about as bogus as it gets.
If you really believe you want your file server on the Web, look at
Netware 5. I'm not sold on NW5, by the way, primarily just because
it DOES interoperate with the Internet.
While the NT Server installation process is fairly simple, NT is famously picky over hardware. Netware is not. I've run Netware 3.x and 4.x very successfully on very modest computers: as small as old 486 systems (years ago, I set up many 386 Netware 3.x systems). I often use old computers too slow to effectively run Windows 9x applications, but still reliable machines. Literally, junk servers (at least one was really about to hit the trash can!), installed for the price of memory and a hard disk (cheap IDE). When you have a problem getting NT to install, you will go nuts trying to resolve it -- there is a limit of what you can do to resolve the issue. I've found very few 386 or 486 (or Pentiums, or etc.) computers I could not get Netware to run on. The Netware 4.x installation process is actually pretty smooth now, although I really appreciate the flexibility of the old Netware 3.x process. It wasn't hard, but you had to understand it. Netware 4.x will let you get a server up and running without understanding the process, just like NT. I find that a little scary, but the good news it is fairly easy to clean it up when you are finished.
For small networks, the relatively few administration programs of Netware make it much easier to control. For large networks, the system administrator rapidly learns the Netware administration programs, while lacking the cute cartoons of the NT administrative functions, are lean, fast, and effective.
Microsoft likes to cite issues of multi-server network administration problems on Netware. This argument is entirely bogus. While it is true that administering multiple servers under Netware 3 was a clumsy process (each user had to be administered separately on each server), this was not true of Netware v4.x. Further, one will find that a single Netware server can support many more users than a single NT server. If you have five servers and 20 users. NT is easier to administer than Netware 3. If you have five servers and 200 users, you will probably find Netware 3 easier and faster to administer. If you have one server and 50 users, you will find Netware 3 easier and faster to administer. If you have multiple servers, you will find Netware 4 easier to administer, period.
Virtually anyone who has really compared them considers Novell's Netware Directory Services (NDS) a vastly superior administrative model than NT's model. Microsoft's "Active Directory" is supposed to address that, but when you consider Novell will have a seven year lead on Microsoft, don't expect great results from "Active Directory"
Most of the Netware networks which are administrative disasters are
such not because of the platform, but because the administrators made bad
design decisions, and cobbled a "Quick Fix" to a problem which should have
been addressed properly, and the "Quick Fix" became permanent. You
can not blame the network operating system for improper management, and
if anything, NT certainly encourages a "sloppy" system with minimal planning,
and is typically being installed by people with virtually no real-world
experience in business computer systems. The real solution for a
system which is badly designed is to clean it up or redesign it, not to
Netware 5, I found, disappointing, as it requires a very fast processor
just to get out of its own way. Really not a problem for most people
who were very fond of putting excessively large processors on Netware 3
and 4 servers, but you won't be doing "junk" servers with Netware 5 for
a number of years yet.
However, a few interesting facts:
You can also very effectively mix NT and Netware on one network, using
Netware as your file and print services, and NT as your database server.
After all, you will end up using a separate server box for your database
server anyway under NT, why not keep the Netware doing what it does better
than just about anything else?
Even if you knew that Novell was about to go bankrupt within a month, does that mean you should install NT in your new network? Again, hardly.
If you are to put in a network this month, look at the options. If you put in Netware, you will have a good product for several years, basically until products start shipping which don't connect to Netware. This will be a while. If you put in NT today, within a year or so after the release of Windows 2000 (formerly, Windows NT v5.0), you will have to plan on replacing it, and replacing (Microsoft calls it "Upgrading") is a major task. You see, every version of NT so far has been a major departure from the previous versions, and to properly implement it, you will have to redesign your network. NT was designed without a deep, coherent plan, but rather a whole bunch of features that sound impressive. The shortcomings of each version so far have been so horrible MS has had to radically change the product every release. Also, after-market support and technical support for old versions of NT have historically vanished quickly after the release of a new version, there is no reason to believe that this will stop happening. Novell still supports Netware 3.12, an operating system that is seven years old now. It was a very well designed OS, and in fact, its basic design is the foundation of Netware 4.x, and in fact, to a somewhat lesser extent, Netware 5.
If you choose Netware, and IF Novell goes bankrupt, you will have to
eventually redesign your network for NT (or, something else). If
you go with NT, you will DEFINITELY need to redesign your network.
Your choice: Novell, with a possible eventual redesign, or NT, with
a definite redesign on Bill Gates' schedule, not yours.
NT's extras are something of a joke. Most "experienced" (and that word definitely belongs in quotes. It is very hard to be "experienced" with a product that is totally recreated every two or three years) NT installers will use a separate NT server (and thus, another license) for each and every NT application. From the people I have talked to, it sounds like in a large office, the NT server fan-out rate is something like 1 server for every 20 to 30 users. That is pretty bad. And expensive.
Netware 4.x implements much superior functionality in Netware Directory Services (NDS), a distributed database of network security. This database is distributed among the servers that need it. It is implemented efficiently enough you rarely, if ever, need to have a "dedicated" machine controlling it. The database is very sophisticated, permitting very logical setup of system security and administration. It automatically updates and replicates information and changes. In the event of a failure of a server, the database is still intact through its replicas. The concept of "Primary" or "Backup" is not part of NDS -- all the information is everywhere it could be useful, and it is served from wherever it is convenient to be served from, not from one central place. When a failed server is restored to service, the database re-replicates to the repaired server.
NT's network structure is very much like the very old 3Com 3Share+3 system, a network which was an interesting competitor to Netware 2.x in the mid to late 1980s. Good reason, as 3Com was one of the partners who helped design LAN Manager, and Microsoft bought the technology when 3Com decided to get out of the software business. Also interesting, while 3Share+3 was a powerful network structure in theory, the OS was too clumsy to exploit that power. Netware 2 outperformed 3Share very significantly.
The only version of Netware which Microsoft's Domain system can be compared
with is Netware 3 and before. Netware v4 and later is in a class
completely away from NT. Anyone who crows about the NT Domain Controller
is really lacking a grasp of what not only CAN be done, but also what HAS
been done by Microsoft's competitors.
I often get funny looks from other network support people when I tell them I don't do Windows NT networks. "But why?" they always ask.
I tell them I've never seen an NT network I would be proud to stick my name on. They crash too often, cost too much, and so on, basically, I rattle off a short version of this document. When I get done, they still look puzzled: "Oh, of course, but you make so much more money supporting NT!"
Well, if I were in the business just for the money, I'd be pushing NT. No doubt. That's not my goal in life, however.
I've seen a Netware server run for over 900 days without interruption (A multi-day power failure finally took it down). This system wasn't some kind of super-server, either. It was a cheap clone low-end Pentium class machine with a 1G IDE hard disk, and 16M of RAM running Netware v4.10, with (I believe) no patches installed. I rather think you would find it impossible to find any version of NT in any office which has run 900 days without being UPGRADED, much less running that long without unexpected downtime.
Another client who is switching to NT, observed that he had forgotten almost everything he knew about managing and maintaining a Netware server -- there was just almost nothing he had to do. The Netware server basically ran itself, whereas the NT servers were constantly requiring tweaks, adjustments, and of course, the crash recovery.
NT does NOT have a lower support cost than Netware.
Now, this is not to say that a poorly designed, poorly implemented Netware
network can't be more difficult to maintain than a well designed NT network,
but this reflects the skills of the installer and administrator, not the
merits of the product itself. I have seen some terrifyingly poorly
designed Netware networks, installed by people I wouldn't trust to install
Windows 95 on a workstation.
Microsoft's plan for the future is to get your money.
Microsoft spends a lot of time telling you what features it will add -- someday -- to NT. This is basically irrelevant. The question is, what is in the box you buy today? There is no apparent plan to NT at all, other than SELL IT!
Novell spent serious time developing Netware. The Novell IPX/SPX protocol which is over 15 years old, is still a very solid protocol today. The primary criticism levied against it is "It isn't TCP/IP". When you look at how networking has changed from the early and mid 1980s to now, it is truly remarkable how well this protocol has stood up. The primary limitation, the fact that IPX/SPX is a rather "chatty" protocol was resolved years ago with the introduction of Novell's Packet Burst technology. In this same period of time, Microsoft has gone from NetBIOS to NetBEUI (both non-routable protocols) to TCP/IP encapsulation of NetBEUI.
Netware's basic structure -- what could be considered a "microkernel" (or perhaps, more accurately, a mini-kernel) operating system -- goes back to the introduction of Netware 3.0 in the late 1980s. The kernel OS is loaded, and the various services and drivers are loaded on top of the basic OS. These drivers can be unloaded or reloaded on the fly very easily and quickly. This model was well enough planned that it has served well for Netware 3.0 through Netware 5.0 -- almost a ten year span.
Novell should not be criticized for having planned their product out
carefully in the first place, rather than stuffing something out to market,
and re-engineering it later. Curiously, Microsoft's Sell-First-Engineer-Later
approach works well -- they get free publicity for resolving design flaws,
everyone who complains about installing yet another NT Service Pack makes
the rest of the world think EVERYONE is going NT, whereas Novell gets ridicule
for not revising a product which doesn't have the design flaws in the first
Novell Netware is amazingly hardware independent. You can take
a disk system out of a broken Novell server, and plug it into a very different
computer and have it back up and running in a matter of minutes.
Novell doesn't really care about the details of your main board, your video
board, etc. It needs to know about your network card(s) and your
disk controller card(s). If you have to change these, the steps to
load the new drivers are very straight-forward and fast to implement.
The same can most certainly not be said for Windows NT, where you might
have to spend hours or days fighting with drivers to get a broken NT system
back up and running. You must basically consider an NT system "bound"
to its main board -- pulling the drives out and sticking them in another
system may or may not be fully successful. The ONLY times I have
ever had a Netware 3 or later system down for more than 24 hours were systems
where the owner opted for warranty repairs over pay money for locally.
I have several clients, who for a minor additional investment, have opted
to go for a full-onsite-redundancy system, where repair times can be 15
minutes or LESS. I like this feature and ability a lot. I know
of no other network operating system where anyone can restore a broken
system to full functionality as fast as I can restore a Netware 3 or 4
Let's start with a very brief comment about Linux vs. Windows NT. NT is, in addition to a file and print system, also a general purpose OS. In this regard, Linux is very similar -- a general purpose OS, which also does Networking. Some may consider Linux a newcomer, but this is quite unfair -- Linux and Windows NT actually had their early development taking place at much the same time. Whereas NT's publicity is fed by the marketing might of Microsoft, the recent "news" of Linux is fed only by the fanatical support of its users. Put another way, you were told NT was good. People TRIED Linux and found it good, so good, they formed a huge volunteer marketing department. Further, the networking of Linux is based on Unix networking, and Unix is the platform that networking was basically "invented" on. The people who develop Linux had a lot of experience to build upon.
Linux's native networking system is called NFS. While PCs can support NFS, it isn't native, and it isn't as pretty as Windows users are used to. Some programmers got together and developed a package called "Samba", an implementation of the Microsoft networking protocols (SMB) on the Unix environment. Like Linux, Samba is free. And, from what I have heard, Samba is a remarkable product, promising to make almost free network servers.
Well, the price is right for Linux systems. The capability is basically there, too. I've got some concerns, however, on support. In some ways, this seems strange to say, as there are probably more people who know Linux than there are that know Netware, and probably more people that know Linux than really know Windows NT. The problem I see, is that most of these people are young, fanatical lovers of Linux, but people with very little real-life experience in the world of business, people who don't understand the difference between "working" technically and working as real people use a system. Most of these, well, kids, have lived their life around computers, and really don't know how the computer-ignorant work with computers.
Much like NT, I've also got my concerns about repair time. I can repair almost any Netware problem, from a bad main board to a bad hard disk to a bad software problem in a matter of between minutes and a very few hours. This is critical for business. NT is a disaster to repair. Linux appears to be somewhere in between. Part of the Linux problem is the product is changing enough that it is hard to know all the details of any given installation. Still, I have seen Unix systems down for as long as a week over obscure problems, and this scares me a bit.
At the moment, I would rather support a Netware server over a Linux server. But, I'd also rather support a Linux server over an NT server.
Now, another area I haven't touched on is application servers.
Here, Linux (and other Unix systems) stand miles ahead of NT, and Netware
isn't even attempting to enter this market. These are the "traditional"
uses of Unix and other mid-size systems, where the program resides AND
RUNS on the "application server", and what is at the desktop is nothing
other than a terminal. This kind of application has been dying out
lately, and quite unfairly. From a maintenance standpoint, they are
far superior to anything running on a PC on the desk. The problem
is, they aren't as "pretty" as modern Windows applications, and therefor,
they fell out of favor. This is quite wrong, and I think IS departments
are slowly starting to realize that "pretty" does not equal "productive",
and that Windows applications are a maintenance nightmare. Microsoft
is trying to get into this business with their "Thin Client" solutions
and "Windows Terminal Server", but so far, the results are less than perfect.
This is an area where Unix excels, and for the applications where the users
will permit, should be implemented. Linux is very capable of doing
this kind of application.
to Computer Opinion Page
Back to Home
(c)opyright 1999, Nick Holland