An NSS Group White Paper

by Steve Broadhead

Table of Contents

Introduction - Client Operating Systems - The Current State of Play
Survey Details
Survey Results
Strategy Issues Analysed
Client OS: The NSS Groups� Own Experiences
Summary & Conclusions

The computer industry is continually referred to as being "dynamic".

And not without good reason. Barely an hour, let alone a day goes by without tens or hundreds of new products being launched, major upgrades made available, the latest batch of fixes and patches being posted on the Internet. While such dynamism is good for vendors, consultants, journalists, analysts and any other bodies which thrive on the effects of change, for the poor old end user it is increasingly a nightmare.

Take the PC client desktop operating system as a primary example. For the network manager in the '80's life was relatively straightforward. There was DOS, versions 1.0 through 5.0 - a steady progression which roughly fell in line with the changes in Intel's processor architecture. In the mid-80's Windows appeared in its original 1.0 incarnation. It just about worked. As did Digital Research's GEM interface which was ultimately acknowledged as a blatant copy of the Macintosh GUI. Neither were taken that seriously which left the network manager to support a combination of DOS-based PCs, a few Macs, and a mix of mainframe terminals. Slowly but surely, newer versions of Windows emerged and these were lightly evaluated by most companies but still very much out of curiosity rather than for genuine business reasons. For one thing they didn't really work and - possibly for this reason - third party applications support was relatively minimal.

Meantime UNIX-based GUI's also appeared but these were very much specialist and didn't impact on the mainstream PC client environment. Overall, it was time of relative manageability as far as upgrade policies were concerned, the latter very much dictated by which particular applications you chose to run. Microsoft had no particular domination of the PC applications market, unlike now, networking was dominated by Novell, the mainframe world was dominated by IBM and DEC (and ICL in England) and PC to host connectivity was a relatively simple, DOS-based application.

For the computer or network manager in charge of his or her end user base then, the key upgrade decision was whether or not to step up to the next version of DOS. An upgrade consisted of copying two or three floppy disks' worth of files onto the PC hard disk and keeping your fingers crossed that everything worked thereafter. So, far from Utopia then, but relatively simplistic all the same. And there was no 16-bit versus 32-bit (and soon versus 64-bit) where applications were concerned. Multi-tasking was not a feature of Microsoft's dominant DOS (though Digital Research and others did offer it in their DOS variations) and multi-threading was seriously futuristic stuff for your average desktop PC. IBM with OS/2 attempted to bring multi-tasking to the PC world but only IBM "True Blue" sites really took it in great numbers.

In the '90's the scene was set, then, for Microsoft to impose itself once and for all on the PC desktop operating system market and this it did by one simple action - producing a version of Windows which actually worked! From 3.0 onwards, Microsoft Windows began to sell in greater and greater numbers and - correspondingly third-party application support increased accordingly. Microsoft itself began to put far greater resource into developing Windows-based applications, which led to the Office suite which now dominates the market. As a result, for the first time network managers were under pressure to move from their pure DOS environment to a Windows/DOS hybrid. The introduction of Windows 3.1 then Windows for Workgroups 3.11 added the true peer to peer networking element that Windows previously lacked and this was the real beginning of Windows domination.

But hold on a minute. We're already talking about several generations of Windows (and DOS) and we're not even up to Windows 95 yet. The significance of the latter is that along with NT Workstation - and herein lies another decision headache for the network manager - it introduced 32-bit operations to the Windows desktop. Not only, then, was there the decision as to whether to upgrade or not for "keeping pace" purposes but suddenly there was a technology argument for doing so too. But the question remained - what were the real benefits in switching to a 32-bit desktop client operating system? And, it seems, many network managers didn't really know the answer.

At the server, certainly, 32-bit applications have been alive and well for many years with performance benefits clearly shown over their 16-bit predecessors, but at a users' PC the argument is less clear cut. If we take the case of PC to host connectivity especially, there is a big question market against the relevance of a 32-bit application. After all, green screen terminal emulators used to run under DOS in a few kilobytes of memory, and in many cases that is still how they are used today, regardless of the "bells and whistles" which come with almost every PC to host connectivity software package today.

Little wonder then, that Windows 98 sales have been very slow in the business environment to date. Ten years ago the upgrade decision might have revolved around: "do we move up to the next version of DOS immediately or wait one generation and do a double jump?" Now we have a situation where many networks are made up of everything from DOS 6.0 or 6.2x through Windows 3.1, 3.11 and 95, as well as NT Workstation 3.51 and 4.0. So just what client OS platform do you upgrade to and when? Perhaps an easier question to answer is: "who would be a network manager nowadays?"

The problem is that someone has to be that person and they have to help make the decision on client OS upgrade strategies. And their starting point is what combination of client operating systems their network currently consists of. Not only is this the starting point for the network manager planning a client OS migration strategy, but it is also key to the strategies of software application vendors. After all, there is no point in creating reams of 32-bit applications, each more packed with features and artificial intelligence than last, if the customer actually still wants simple, easy to manage 16-bit software which does the job they require of it, at a low price and nothing more.

Hence the purpose of this survey. Here we look at a broad cross section of end user companies in the UK and assess their current client operation system implementation and future requirements. In total, around two hundred companies were involved in the survey which covered every area of industry, commerce, local government and education. For Microsoft especially we imagine the results make for interesting reading! So read on...

The survey was conducted in the UK only.

Over 200 companies took part in the survey. These were drawn from several databases including software company, consultancy and service provider client lists.

The companies which took part, between them, covered a broad spectrum of user types. Representatives from the worlds of banking, commerce, manufacturing, public sector, education and retail were all included in the survey.

In the survey we included the following key questions:

Approximately what percentage of users within your organisation are using 16 or 32 Bit client software and how does this break down into the following:

- DOS
- Windows 3.1/ 3.11
- Windows 95
- Windows 98
- Windows NT
- Mac
- UNIX
- Others

Which would you describe as your primary strategic network/server operating system currently - NT, NetWare (please specify version) or UNIX (please specify version)?

Is this likely to change in the next 12 months?

Do you plan to upgrade your 16-bit client base in the near or longer term (please specify) future and, if so, to which operating system platform?

What would be the primary reason for making/not making this change?

How significant is the client OS platform in terms of:

- Network connectivity?
- Host (Mainframe/UNIX) connectivity?

Approximately how many users in your organisation have Internet access direct from their desktop?

Approximately how many users in your organisation have web browsers installed on their PCs and is this dependent on the client OS they are currently running?

How significant do you feel the client OS is, in relation to Y2K and EMU issues?

DOS

The use of DOS finally appears to have largely, though not totally, dried up. This doesn�t mean to say that there aren�t still thousands of plain old terminals still in use � there are. But as far as PCs are concerned it�s largely a case of Windows of some form having now, and generally long since, replaced DOS.

16-Bit Windows Clients

While many companies have moved on to full-blown, 32-bit clients in one form or another, there are still large numbers of 16-bit Windows clients around, in the form of Windows 3.1 or Windows 3.11, A.K.A Windows For Workgroups. This is a significant point to note for application software vendors as it is all too easy to assume complete 32-bit environments are now prevalent and ignore compatibility issues with 16-bit client populations. To do so means potentially alienating these customers who can simply move to another supplier which will support its older client base.

Not surprisingly, the 16-bit breakdown sees a huge predominance of Windows 3.x over DOS.

While 32-bit clients now clearly outstrip 16-bit clients in the UK, in terms of actual numbers, the 16-bit market still represents a huge opportunity for software vendors. When you consider that over 50% of companies in the survey still have 16-bit client operating systems in daily use, the argument is strengthened further. It also emphasises the reluctance of many companies to kill off their 16-bit client base, not least because of the costs involved. �

Notably, the primary reason giving for upgrading to a 32-bit client platform was not speed, ease of use, or any of the other features which Microsoft and others continually make noise about, but simply to keep compatibility with the applications world. This smacks of the application software vendors pushing the customers forward when they don�t necessarily want to change, otherwise known as the tail wagging the dog.

Unfortunately this kind of thing is so commonplace in the computer world that people become blas� about it. In truth, however, it is a serious issue and one which is especially common in the hardware and networking worlds. There is often no good reason for a company to upgrade to the next "latest and greatest" version of product "X" yet often they are effectively forced to do so. At some point this bubble, you feel, has to burst with the customer base saying: "hey, enough is enough � we�re happy as we are thank you." Maybe we�re seeing signs of it already with the slow uptake of Windows 98.

32-bit Windows Clients (95 & 98)

For many customers, the move to Windows 95 was THE most significant upgrade they�d ever made. Except that some haven�t even got that far yet and now we�ve got Windows 98 to consider. There is no doubt that Windows 95 offered many more useful features than its 3.11 predecessor, as well as a 32-bit engine, but in many cases those features were simply irrelevant to the needs of the user base. The result is that the picture regarding Windows 95 is somewhat fractured, with companies tending to commit either heavily or barely at all.

Within the survey results, figures varied from 0% of Windows 95 users to those in the high 90�s. Several had very low numbers of Windows 95 users however, which did come as something of a surprise, given the time now elapsed since its introduction. If those who were slow to take up Windows 95 were thought to be waiting for Windows 98, then maybe it is time to think again.

While it is still very early days for this client OS, uptake of those companies featured in the survey has been largely minimal with plenty of 0% and 1% figures. This, however, could easily be translated as simply not having had time to deploy the latest version of Windows, but the responses regarding future plans (see separate section) suggests otherwise.

32-bit Windows Clients (NT Workstation)

This is where the 32-bit story gets really interesting. Pushed by Microsoft as more of a high-end competitor to UNIX for heavy workstation usage, rather than for mainstream office use, customers seem to being seeing NT Workstation in something of a different light. Not only have the vast majority of our users already installed NT Workstation in at least part of their network, but it also looks to be playing a major part in their future.

In response to the question asking those looking to upgrade their 16-bit platforms to 32-bit in the near future, the answer was almost exclusively NT Workstation, with Windows 98 nowhere in sight. In a sense, this ties in with the relatively large number of respondents who have shown little commitment to 32-bit Windows platforms today. Perhaps they are simply looking to make what they perceive as a two-stage jump in one go, from Windows 3.1x to NT.

Other Client Types

The news of the Apple Macs� death may have been greatly exaggerated by the figures in the survey suggest it is very much only specialist users who have significant numbers of Macintosh clients now. Even then, NT Workstation is seen as a logical next step, even if it does mean changing the hardware totally to accommodate it. UNIX is still hanging around at the desktop in the minority, specialised way it has for some years but clearly NT Workstation is a very strong challenger here.

Network Operating Systems

The results from the survey follow the general trends to have emerged over the past 18 months. Novell�s NetWare is still around in quantity but is giving way to NT4, the latter being in the majority now among our surveyed users. However, there are still companies hanging onto NetWare � even the now two generations old v3.12 � and others which have UNIX as their chosen NOS platform, notably Sun users.

Internet Access

We posed the question: is the desktop client OS a significant factor in the number of web browser and Internet users you have deployed?

The answer actually varied from company to company. Some only had browsers installed on 32-bit clients while others were happy to install browsers across all OS platforms. Typically the number of users with Internet access was well short of the number with a browser installed � often 25% or less � which shows that the Internet is still not as prevalent as some would have you believe; at least not in a corporate environment.

Overall, however, the use of 16-bit clients did not appear to be a limiting factor where web browsers are concerned, even if this is contrary to many peoples expectations.

Significant Factors In Client OS Strategy

We posed the question: is the choice of client OS significant with respect to both network connectivity and host connectivity? The answer was very important where network connectivity is concerned and generally significant but not critical with respect to host connectivity.

Finally, we asked: is the client OS a significant factor with respect to the Year 2000 (Y2K) and European Monetary Union (EMU) issues?

In most cases (see over page), respondents felt both were very significant factors, though a few couldn�t see the relevance of EMU and some felt Y2K was not so important at the desktop. Generally, however, the pervasive feeling was that these issues would have an impact on companies� client OS strategies.

Obviously both of these issues will have a very significant impact in many areas of computing and networking but, equally clearly, the general direction of thought with respect to the client operating systems is somewhat confused at best. In 12 months time we�ll know the answer!

Working in a test labs environment, naturally we tend to stress the client OS more than most users which, in turn, throws up most of the problems anyone will encounter with them.

So how has each compared with the other? First, we need to make it clear that we still have all the Windows variations on a theme up and running from Windows 3.1 through to 98, as well as NT Workstation 4.0 (SP3). While the 16-bit versions of Windows have their limits, networking wise, they have equally proved to be reasonably stable and reliable over the years. The same cannot be said for Windows 95 - or 98 albeit with limited exposure to date - which has many excellent features, sure, but has also caused us a lot of problems, which equals time, which equals cost.

Best of the lot by far, from a resilience point of view, is NT Workstation 4.0, as least with Service Pack 3 in tandem. However, specification for specification, PC requirements are far greater with NT than with Windows, though the rewards are worth the extra benefits it brings. We were therefore not surprised by the attitude of the respondents in the survey towards favouring an NT upgrade policy over a Windows-based alternative.

It appears that Microsoft might be in for something of a surprise here! The first point of note, however, is that there are still considerable numbers of 16-bit clients around in the UK. Anyone who regularly visits end-users will hardly be surprised by this but it is still worth confirming through surveys such as this one. Indeed there are still sites around which consist of huge percentages of 16-bit clients. The biggest such example was a site with around 300 users of which 85% were still Windows 3.1 or 3.11 users.

So what are the reasons for this reluctance in some instances to move forward to a 32-bit desktop environment? The most obvious one is cost. Not only is there an incremental cost in upgrading to a newer version of Windows and the support costs involved in that upgrade, but there is a potentially huge hardware upgrade cost involved too. Windows 3.1x will run reasonably well on i386 and i486 machines with as little as 4MB RAM, but moving up to Windows 95 meant an i486 processor minimum and 8MB RAM absolute minimum though 16MB is the realistic lowest RAM count in practise. The latter plus Pentium or Pentium II (or equivalent) processing power is certainly true for Windows 98 and the absolute minimum for NT Workstation with 32MB RAM realistically being the minimum requirement here.�

However, most sites said they would be looking to upgrade their 16-bit clients in the short or mid-term future.

So why is this the case? Mostly it is in order to support the application set they now want to run. Microsoft and others continually push forward their application sets to work with ever faster and higher client specification and 32-bit is the absolute norm. What may surprise Microsoft, however, is that a huge majority of companies are looking to upgrade not to Windows 98 or even Windows 95 (as a tried and tested option) but to NT Workstation 4.0, or 5.0 when it finally appears.

The reasons for this appear to be two-fold. Firstly, in the move from 16-bit to 32-bit now, years after the first 32-bit client OS' arrived, there is a general feeling that it is worth making as big a leap as possible, including moving to new hardware platforms simultaneously.

A key point to make is that NT Workstation is perceived as being THE definitive 32-bit client OS, rather than the 95/98 incarnations of Windows. Given the multi-threading and service-oriented engine of NT, this is an accurate assessment and therefore would appear to make great sense. True, it makes greater demands on the hardware than Windows 95/98 but with current PC costs lower than ever, both literally and "bang per buck" basis, the extra cost of buying 64MB RAM rather than 32MB and a 3.2GB hard disk rather than a 2.1GB are negligible, even where the client count is high.

Where, then, does this leave Windows 98? Primarily as an operating system bundled with PCs for domestic use, would appear to be the answer. Given that the primary stated motive behind moving 16-bit clients to 32-bit was for application support, as outlined again in the chart above, NT has the backing of the business software community which therefore makes NT and not Windows the obvious 32-bit Client OS of choice for business use.

Send mail to webmaster with questions or�
comments about this web site.
Copyright � 1991-2003 The NSS Group Ltd.
All rights reserved.