Slashdot Mirror
A Strategic Comparison of Windows Vs. Unix
Ramsed writes "On LinuxWorld Paul Murphy wrote an article comparing Unix and Windows for a 500-student system and a 5,000-user manufacturing company. Summary: Most of the Windows versus Unix debate has been cast in terms of which is technically better or which is cheaper, but the real question is, 'Under what circumstances is it smarter to pick one technology rather than the other?'"
Obviously there are gross simplifications in the article but assuming that parents are going to buy BSD/Linux based PCs is ludicrous. Not to mention places like Dell have dropped installing Linux.
That means you would usually buy a complete PC with Windows then have to slick the drive and install Linux. And somehow I just don't see parents going with Linux. The *only* way this happens is if the school forces you to buy a prebuilt package(s) from them.
Sorry. That assumption is way too far gone to be overlooked.
Fsck cluebie moderators. I'll say what I want, offtopic or not. And fsck having to qualify every bloody statement just
While I won't attempt to make the estimates myself, I will suggest a few things to take into consideration
- Learning curves. In the school and corporate environments, people don't want to waste time learning unix or linux. They don't work the same as Windows, which is the standard desktop practically everywhere. A normal situation would be that only some of the I.T. staff and power users know unix. If you can teach the blonde bimbo that blows your boss and makes memos in MS Powerpoint to send via Outlook the advantages of being able to compile your own kernel, I'll shut up about that, but it's not realistic to assume that people can easily learn a new OS. After all, most of them don't even understand how to use Windows correctly.
- Interaction with others outside your office. Since Windows is the standard in the corporate world, you have to be able to communicate effectively with Windows. Samba is not easy for the average user to use like network neighborhood is. OpenOffice isn't able to work with MS Office as well as people tell you. It can read some old versions of word documents, but it doesn't work with Office XP. Microsoft will most likely make a conversion tool for Windows users who are using Office 2k or older, but not for unix. Unfortunately, until you have everyone agree to use unix it will never be a good office tool for people that communicate with those outside your office.
- Support costs. Corporate support is a very important thing. Anyone that works with big companies to maintain their server hardware and software knows that if you have a critical problem and you're paying $200k a year in support, they will have a patch out for you by COB the next day. (Perhaps that was a slight exaggeration, but they are still very quick to solve problems.) The problem is that Windows support is generally cheaper than Unix support. I wouldn't even consider linux in an office environment though, because those that support it are not the same group that developers the software.
There are others that I could mention but those are the main three things that seemed to be left out. It's hard us normal people to quantify the amount of money those things cost but most corporations have a team of people dedicated to that sort of stuff. I think that for how greedy most corporations are, if they honestly thought they could save money by not using Windows, they would switch in a heartbeat. However, after careful and detailed evaluation, much better than the one in this article, they decide to stick with Windows or migrate their stuff to it. They have to be saving money with Microsoft somehow, and I think those three categories are some of the major ways they justify it.Mas vale cholo, que mal acompañado.
Having working in a campus environment for a good part of the last 4 years I can say that not everyone wants to learn something new, let alone spend the time to familiarize themselves with software packages they are unfamiliar with. Only students of Comp. Sci/Comp. Eng. are for the most part willing to do this, and even some of them are not.
While the article states that there would be the need for only a single *nix support position, and four Windows support positions, we must think of this: How many additional postitions would have to be created to train students (even rudimentary training) for an infrastructure they are not accustomed to? I would guess at least 10, but depending on the size of said campus, it could grow to an exorbant amount, overshadowing the cost of the initial startup costs.
The campus I am at now is a great example (Northern Illinois) and especially the labs I work in (art/music). There are plenty of Mac's here for people to use, but unless they are die-hard Mac-heads or it is required to use them for a class, 99% of the students stay away from them for the sole reason that it is unfamiliar territory. This made the campus cut down to a single Mac support position for the entire campus (which has over 200 macs), solely because of peoples inability to accept things that are different.
Look at the makeup of the world's computer market, 90+% Windows. People fear change and are afraid to learn. Even in academia.
Later
Josh
This article doesn't mention that it costs money to train people to use Unix. It doesn't have anything to do with how smart they are, they'll ay least need time to adjust. If you've ever read an ad in the newspaper looking for a secretary, you know that MS Office is pretty much the prerequisite. All of your employees know how to use Windows coming in, not so for Unix. Retraining people costs money.
:)
This article seeks to use "average" scenarios to make its point. I would say that Unix would be a lot more beneficial in specialized situations, where employees use a lot of custom or specialized software (e.g. POS stations, industrial settings). They're going to have to learn anyway, so why not have them learn it on a cheaper, more stable platform?
In the college scenario, the article takes no account that many colleges make these decisions based on what the students use. Usually, that's Windows. Sometimes Mac. Almost never *nix.
In the corporate scenario, no mention is made of the need to share files with other companies. This requires Windows. No corporation really cares about the evils of closed file formats until they get in the way. Besides, how are any pitches going to be made without PowerPoint?
To be realistic, both situations should have compared the cost of a Windows setup vs. a mixed Unix/Windows setup, since that's how it work in the real world.
Windows -- Grudgingly useful for desktop/secretarial environments, and you'll also find that most of the accounting packages out there, as well as many embedded systems packages, require it. Windows is also, like it or not, the OS of choice for hard-core gamers. Sucks, but true. Generally not a good choice for server environments due to cost and MS lockin (stability issues were all but eliminated with Win2K). Limited to x86 platforms; all other versions died of user apathy.
Unix -- Useful for light-to-medium duty single server environments (especially file-sharing and WWW), as well as clustering; Solaris, AIX, Irix, and occasionally even Linux pop up on high-end (i.e. mainframe or supercomputer class) systems. Also the system of choice for cluster computing (though MacOS Classic can make a credible case for being a viable cluster computation environment as well). Unix's traditional timesharing environment is a very small niche in the modern market, but still useful. Also a major scientific computing platform. The downside is that the proliferation of standards makes generalizing about anything above the command line difficult and/or pointless; Solaris != Linux != BSD, and it's going to stay that way. Runs on everything concievable, from a Commodore 64 all the way up to gigantic Cray supercomputers and Linux clusters.
MacOS -- Don't run a publishing house or recording studio without it; the Mac is the platform of choice for the creative industry. Also a good choice for education, but a weak gaming platform. MacOS X largely eliminates instability from legacy code. AppleScript as a scripting platform makes VBA and Unix Shell look horribly primitive (and MacPerl is available as well). Limited to PowerPC hardware.
That's my summation...
/Brian
Preface - I am fairly agnostic to what is on my desktop, although I do prefer Unix to Windows.
Reading this article, being a very informed technical user (one who has done both Uni unix sysadmining and Windows sysadmining because, well, what Windows machine hasn't needed it?), I found it very hard to buy any of Murphy basic assumptions or trade-offs.
First off, why does a Dell 2100 cost so much in the Windows solution? I went to www.dell.com to price the same thing and got US$1262.11 (40GB HDD, 256MB, 1.1Ghz Celeron, 17in head, net card, 2000/XP with Office academic). Mind you, I went in the Academic pricing door, because he is pricing for a school. The Office/2K software adds about $280 to the bill. Thus, the only thing he should have noted is that each computer buyer shells out $280 more for Windows. In other words, for the 900 computers (500 school, 400 home) in his first example, that's $252K - no chump change). That assumes no school licensing. If he isn't getting those basic numbers right, you know the rest of the article is bent...
The idea that "Smart Displays" would cut it in school is OK for some (terminal rooms, where many go to just read mail and surf), but forget it for heavy work. I've not heard of these being satisfactorily used in practice.
Also, I hate to say it, but I don't think this guy has ever seriously used Win2K. Many may not like to hear it - but I've only seen the BSOD once while using it. I've been actually pleasantly surprised myself at its reliability. I am now able to run these things for months without reboot (OK, so I had a solaris machine that went for a little over a year once until we upgraded the memory...). In any case, either system properly maintained is fairly reliable.
Point 2 - administration. At my old Uni, the CS systems (not the general machines) were maintained by 2 full time Unix sysadmins (we actually had very few Windows machines at the time) and a horde of cheap or free volunteers. The systems ran 24 hours, but only with help (because beginning CS programmers can do all sorts of weird things you don't anticipate). Either way, it's at least one full time person for Unix or Windows. I think the real cost will be in all the tech support needed for these students that grew up on Windows at home (at least 95% of them). That will need 4 full time people in and of itself.
I'll buy point 3, but everyone likes to upgrade.
I'm a little less able to gripe about his assumptions in the 5,000 manufacturing environment, but I'll add in some thoughts...
The last company I worked at had over 5000 all over the world. It was a mixed Unix and Windows (mostly Windows, since tech is always smaller than marketing and sales), and the whole organization didn't have but 50 tech support total. They worked hard, but they had a pretty efficient setup, and things went pretty smoothly. I'm going to assume he got his 30:1 Windows user:support ratio from some informed source, but he doesn't cite one, and I've never seen it that bad in practice.
Anyway, no need to beat the horse. There is one reason I do like the article. It is totally biased for Unix to win. However, there is so much crap that says the opposite (in Windows favor), that I guess you have to have the CIOs read both poles of crap to come to a decision in the middle.
What is often missing in these formulations is the investment in legacy software. This is why Microsoft won and Apple lost in the late 80's. Sure the Mac was better... but it didn't run all of the custom developed DOS software that Windows did. Then in the early 90's it was Windows NT vs OS/2. Although OS/2 had a compatibility layer, it wasn't "Windows". And thus, once again, all of those custom windows applications came to play.
.NET stuff. But how likely is that? Not. And so we go round and round the treadmill. As corporate lock-in grows deeper and deeper -- tough luck Linux.
Now we want companies like Ford to adopt linux? It isn't going to happen. They have, I am sure, billions of dollars invested in 16 bit and 32 bit windows software (Yes, there are still many VB 3.0 applications out there.). Until Linux provides proven, reliable, backwards compatibility here it's no dice. The lock-in cost is just too high.
Now. This may be possible in 10 years from now. As long as corporate developers use plain ole HTML plus well-supported Javascript and don't use ActiveX and, worse the new
Absolutely correct. The uptime on my w2k was interrupted only by hardware problems and service patches.
How about someone rating the MTBP (mean time between patches)? The MTBP is a far bigger problem than the MTBF.
I did have some problems with w2k initially. All of my problems were due to PC Anywhere, a bad Matrox driver, a bad SB Live! driver (SMP bugs), and a stick of memory that went south.
Maybe the reason my w2k box ran so well was my Enermax 350W power supply. I think people who run Unix also tend to build better boxes.
Despite the impressive months-at-a-time stability I experienced with w2k, this machine is now running Debian. After my memory went bad and I contemplate rebuilding my software environment with all the correct patches and drivers I came down with a serious case of patchitis.
Let me tell you, though, that dselect is no walk in the park either. Ever installed ext3 on Potato and then discovered that the XFree on Potato needs some extra TLC to run dual-head, so you go ahead and run Testing anyway?
What's the net cost of Potato running two years behind the times?
Unix guys are like the people who spend two weeks at the beginning of summer painstakingly ridding their yards of every weed and vermin, and then spend the rest of the summer drinking beer in their hammock hurling abuse at their neighbors who have to spray their Dandilions every other week.
Let's face it. For a UNIX admin, running UNIX on the desktop is a no-brainer. But if you expect the average user to run UNIX on their desktop, you're dreaming.
Believe it or not, our users are not stupid. Most of us couldn't (honestly) do their jobs anymore than they could do ours. This being said, asking them to suddenly use an OS and applications they have never seen because of some idea from an obviously biased article is crazy. Users will be most comfortable using an OS and apps they know, even given that the systems may sometime crash...
I know I will get flamed for this, but let's face it, Linux (and other UNIX-type OSes) can't win the desktop war. M$ has won already... You will never see the day when *nix is more pervasive on the desktop (at home or at work) than Windows. Unless you have a reason to spend months learning the intricacies of UNIX (and face it, our users don't), most people will go with what is easy. Most people (meaning non-admins) want PNP and point and click, not compilers and command lines...
I'd rather see all of the creative energy focused on trying to install *nix on the desktop be applied to making the back-end server functions rock solid. That's where UNIX has it's best potential.
And as for the reliability issue, my Windows Desktop has been up for weeks without a problem, and a 2 minute reboot fixes it when it does crash... Same with my W2k desktop at work... Of course I love knowing that my DNS, DHCP, and mail servers running on Solaris have been up for years at a time without a glitch too...
You only think that's true. One of the key discoveries in the science of human-computer interaction was that users frequently perceive easy tasks as being slower than harder ones, even though the reverse is true.
All the "power users" who think CLIs are more efficient because it seems like it takes less time would do well to try making some objective speed measurements with a stopwatch. It might come as a surprise that GUIs are actually faster, even though it seems like they are slower.Free Hans!
Why don't the lower levels of education use Linux based systems? As a tax payer I shouldn't have to pay for software licenses if Linux gives equivalent functionality for free. How about someone doing a study on that?
Among the truly staggering costs left out of Paul's analysis are:
1. Training end-users in an entirely new interface.
2. Retraining staff and hiring experienced Unix sys admins.
3. Migrating user documents from full-featured products like Office to stripped-down freeware like StarOffice.
4. Recoding, from the ground up, many custom apps designed to run on NT using premium-cost Unix developers. Testing, debugging, documenting, and implementing all these apps, and...retraining users (again)!
I think a recalculation is in order.
Haven't you heard of switches and VLAN's? You can overcome the issues you mention using moderately priced switches (assuming you're buying extreme or HP or Foundry, and not Cisco) and some not-too imaginative network configurations.
If you really need extra inter-segment security, throw in a transparent bridging linux firewall between segments and you've got a pretty tight setup.
== Just my opinion(s)
This is not true in reality. (I post from a sunray so I think I'm entitled to speak) These terminals speak IP so can surely route them. IP addresses are given out through normal DHCP.
:-)
But it's true that they consume bandwidth, peak 30 Mb/s. Although we've tested them with as low bandwidth as 512 kb/s, and they actually still work, although somewhat slow....