Slashdot Mirror
Survey shows NT admins looking at Linux
bpdlr writes "A recent survey carried out by the Microsoft Exchange Forum in the UK shows that 20% of NT/Exchange admins in UK enterprises see Linux as a potential alternative to NT, up from 0% in six months (it's a PDF file, see page 4). Compare that with only 29% sure they will upgrade to Win2000. "
Well, I have had some coffee and I am feeling a little more relaxed that I was last time I commented on NT ROI issues (where I was referred to as the "skinflint boss from hell", which has shown up in the company directory three times so far, thank you very much), so let's look at some basic issues with licencing and support.
... stretched out a little bit more.
Your point, concisely made, is that NT allows people to be more easily productive and thus the cost is justified. You define that as being able to change settings without talking to the admin or help desk (I assume that you carry this out to stuff like loading floppies without mounting and unmounting and so on). You then note the savings by comparing the licence costs to the retraining costs. This is where you lose me.
Now, I can certainly see the math; if the cost of the licences exceeds the costs of retraining, then it is not justified. If the licences are cheaper, then NT is a good idea. However, moving from this back to ease of use as a justification is not really justified. I see what you meant (I think), but let's look at the same concept
You make a few assumptions (correct me if I am off):
1. People are used to operating in a Windows 95 or NT 4.0 environment (as 3.51 and 3.11 are both a lot harder to mess around with). Fair enough -- where we use Windows, we use NT4.0, except on some laptops with 95.
2. People need in the course of normal business (and as part of a valid business reason) to make changes in their environments that would require actions too complex for normal users to accomplish without resorting to the well-known 4357 call. Windows allows them to do this without the call.
3. By allowing the changes to be made quicker, more work can be done, making the company more money.
And then the math, which seems right (your math is not in dispute).
I have to argue that you are making some wrong connections and that some of ypur assumptions are not correct. Here is why I think this:
1. Assumption 1 (familiarity with 95-type desktop) is seen to be causative in increasing the speed with which users are able to solve "mission critical" problems (perhaps not to the enterprise, but to the users themselves). This is not correct for several reasons:
a) unless the user is a hardware tech or doing odd graphics things, changing settings is not something that needs to be done by the user. Pick a corporate standard and stick with it. If the user objects to things like the fonts, let them change it via a menu (I am not a perl hacker, but even I could make a menu-based system to allow this in one afternoon, as I have done so before). If the user wants backgrounds with psychedelic flying lizards and feels strongly enough to hold up work until he or she gets said lizards, then the user a) has too little work or b)needs to find another job.
b)And if you are talking about settings in Word and Excel (I have found that many NT people lump applications in with the OS), the same applies. Is the person doing word processing or are they making art? If they are making art here, they will get Frame and 512MB of RAM and so on and so forth. Art is their job and we want to give them the right tools. If they want sixteen different fonts in five colors with Outlook and Word, well, there is that entire concept of a value-added vs. a non-value-added activity. This sort of thing is non-critical and is almost always a problem that the user has produced themselves. They should be punished, not given tech support (no, not with a beating -- we get the departments to pay per call and their managers are now paying attention).
2. Assumption 2 (that users need to make changes) is simply wrong, as I have touched on above. Where is user error and user eccentricity and users avoiding work something that we want to go out of our way to encourage by "helping them" do more of the same? I would argue that if the users are screwing with their settings, they need to stop. That is a management problem, not something that inevitably happens -- it happens if you let it happen. If people are messing around with their system, they are not working. The entire point that you are making (that elimination of problems speeds users to a point where they can work) suggests not that we solve avoidable errors faster but that we avoid the damned errors in the first place. Like by telling people that blowing up their machines will become a discipliary issue. Like by using NT instead of 95 and shutting down as much of it as possible (or CDE). The way that you deal with hitting yourself in the head with a hammer is not by taking some asprin so that it won't hurt as much; it is by stopping hitting yourself in the head with the hammer! Why is this so hard to understand.
3. Finally, Assumption 3 (that the avoidance of time loss pays for the licences) only works if you assume that the other two points are correct, which they are not. If you eliminate the obvious solutions to the problems, the tortured logic makes sense, but that is a little like buying a tank so that you can get to work safely because you insist on driving to work every morning on the wrong side of the road. There are easier solutions.
From my experience, which goes back a long ways (pushing 30 years at this point), user support is something that has become a huge issue only with Microsoft OSes and applications because they are a) so buggy and b) impossible to administrate centrally.
a) You cannot fix NT or Win3.x/9x because it is unfixable. It is so buggy and so badly developed that there is almost no way to permanently fix problems. That is why the industry average number for help desk staff:user on Microsoft OSes is 1:15. I am not a huge fan of either Macs or OS/2, but the normal number there have been 1:100 for years now. Our admins are normally in change of 400+ workstations and still find time to play Doom (or Quake). he help desk costs are horrible and they are unique to Microsoft operating systems. NT is only a little less problematic. And this is without going into the multiple revisions of Office that we have to support, many of which do not work well or similarly in very major ways. Even if everyone is familiar with the environment and the apps, it doesn't always help. Contrast this with vi and ISPF, with shell scripts and COBOL, with WordPerfect (for want of a better cross-platform exaple outside of Notes). You will get the calls anyway.
b) I'm not arguing for terminals -- look at what you have to do with 300 PCs -- you have to do either the same thing 300 times or do it from a distance. I didn't dislike UNIX because the admins could do most of the work without leaving the room and I didn't have to worry about getting the place depopulated because they were off site in case we had a disaster. Of course, the IBM folks never left the machine room at all. With the PCs, we had to send people to every damned machine. This was hugely expensive and continues to be a problem. One reason why Linux looks attractive to us is the remote administration tools from UNIX (and Tivoli, when they get their act together; we do not do CA, period). You cannot do this with NT. I know. I was hopeful when SMS came out. Nothing has or will happen. Microsoft products will always cost more.
So, basically, I would rather save money by avoiding NT. As I have stated before, it costs too damned much in licences and support. We have users on NCD Xterms that have been happily using SOlaris or AIX for years, some are using Linux now, and the only hurdle was moving to CDE. Some of these people have called us two or three times in 10 years. The mainframe people never call (that is a little misleading -- we have people here who have hacked MVS source code, so they aren't too likely to need any kind of help and occasionally get calls from friends at compeditors and help them with their MVS issues) and they seem to be reasonably receptive to those IBM NCs, so we will get them on a Xterm yet (only twenty years)(those plasma screens last a long damned time). The OS/2 and Mac people almost never call (some of the OS/2 people have never called, even once). The people on Microsoft OSes call all the time. The lesson is top stick with something that works, not to live with the creeping horrors that ooze out of Redmond on a regular basis. I realize that I sound like a LISP programmer saying bad things about UNIX and how Apollo did (whatever) many years before, but I am not that young! I can do the math, though, and mine has a lot more variables in it than your does, apparently. I didlike the expression "thinking outside the box" (when I was in business school, we had to think, we didn't have all of these buzz-words to save us time), but perhaps you should.
Customer A:
Has approximately 100 workstations (Windows 95/98) that will be accessing a central mail server for interoffice and internet email. Despite FreeBSD recommendation, client decides on two NT servers, one for web traffic, one for Microsoft Exchange. Reason for NT choice: "I want everything, from top to bottom, to be Microsoft. It's Microsoft. It's an integrated solution." Aftermath: Client's NT servers are rarely stable, as exchange buckles under heavy email load. Client spends astronomical fees on "NT Experts" to come in and "tune" IIS server to get the easiest of features working correctly. Conclusion: Client has overshot their budget. WAY overshot. Windows 95 workstations seem to destroy themselves left and right. And things still aren't working up to spec.
Customer B:
Similar to the above customer, but in a more email-intensive setting. Client has approximately 100 or so computers running Windows NT Workstation that will be accessing a central mail server for interoffice and internet email. Client accepts FreeBSD recommendation. Aftermath: Client's FreeBSD server is exceptional for mail, file, and print services. Sidenote: Windows NT Workstation is an ideal operating system in their type of environment, and is a strong performer on the desktop. Client may be accepting FreeBSD recommendation for HTTP services in the near future. Conclusion: Client is in WAY under budget, and a few months early to boot.
Based on my experiences, I can say that an ideal company setting would include a BSD Unix (or Linux if you prefer) on the server-side. I prefer BSD Unix because of its excellent networking code. Additionally, FreeBSD in particular is a breeze to install. That's why I carry my FreeBSD CD set with me wherever I go. On the client-side, I prefer Windows NT Workstation for the masses, and Linux or BSD for the techies, at their option. Windows NT Server may be good for nominal load, but I (personally) wouldn't trust it in a mission critical environment.
29% are planning to move to NT within 12 months.
20% are considering Linux as a possible OS alternative in the future.
Sounds to me like the questions are not equal (I don't think they were meant to be). Comparing these statistics is erroneous, as planning is alot different than considering. And future is alot longer than 12 months, they could mean 5 years. Puffing out your chests by spouting figures like this just makes the whole community look worse. Take the following statement, which is close to the survey's: I am planning on wearing clothes tomorrow, but I have considered streaking in the future...
This is not the greatest sig in the world, this is just a tribute.