Slashdot Mirror
Reducing the TCO of IT with Linux?
Bil Simser asks: "I've been asked by upper management to look at the feasibility of replacing our current Windows IT infrastructure with Linux. Basically someone has said that Linux is free so now we're off to see how free that really means. A full replacement is probably impossible, but I can see some benefits coming from selective replacement of specific technologies (e.g. application servers, web servers) that might be feasible. This is both from a cost reduction standpoint and increasing productivity when it comes to system management. I've already looked at a few studies done on TCO reduction on this and they look good so now I'm turning to the Slashdot community to see if anyone has either practical experience or informative insight into a problem like this? The objective is to determine the TCO of deploying Linux as a core part of our operational environment so what does that mean in the sense of hardware, software, middleware and management impact?"
The problem in this situation may be the perception that open source is free.
This means that it is easy for people upstairs to asssume that there will be correspingly less budget for training/ migration, dooming it to failure.
OSS may not be free as in beer, but at least its not like secondhand beer.
Migrating to a different OS costs money, whether from an expensive one to a cheaper one, or from one expensive to another. When you make the calculations, make sure to include a calculation of when the migration will start to pay off.
Setting up multiple alternatives might also be a possibility. Set up a matrix, and figure out how much effort goes into the various boxes that could be migrated - and at what risks.
If you have a simple IIS web server serving static, migrating to Linux will most likely be rather painless. Replacing a file server likewise (unless it is running XP - I'm not soo sure about that one).
It is extremely hard to give any more specific advice than this - since you aren't very specific yourself. How much business logic is tied up in Microsoft products - and how much can easily be replaced by open or java-based alternatives?
Stop the brainwash
Total Cost of Ownership
TCO = Total Cost of Ownership
Purchasing prices + the costs of system management.
please proff read !
Total cost of ownership
factors in all training/ upgrade/ support costs.
Assumes that the vendor doesn't suddenly change the rules (like Microsoft just have) and roger you senseless over a barrel.
There's a linux-service company called kangaroot which offers free auditing about linux-migration.
I don't know about the quality of the audit, but you can always try it?
IBM adressed similar problems to one of the biggest companies in austria
The New York Linux Scene has held business demos regarding Linux for business, TCO, desktop applications, Databases and more. There are audios available for download at the web site that include presentations made at the CUNY Graduate Center in NYC, and more recently at CUNY/LaGuardia in Queens, NYC.
Since you give no information on what you do with your NT infrastucture it's hard to say. But as far as I have expierienced, using Linux won't be considerably cheaper if you don't switch entirely.
The heterogenic fuss of administrating an NT/Linux mix of 500+ PCs is a pure pain in the but. And knowing how crappy NT 'networks' are, I suggest you ditch it entirely.
I guess you are considering a network wide update anyway, so total Linux could very much be the way to go. Alltough you'd probaly have to start with "let's just change the servers and one or two Desktops" to get the people used to the Idea.
And finally, to answer your question:
A Linux only enviroment for Standard PC work will allways be cheaper than WinNT. Provided you know your way about Linux admining and there's no special software that only runs on NT. Which would only be something like special Video NLE software or something simular.
We suffer more in our imagination than in reality. - Seneca
Mandrake Linux Business Cases
Consulting Times
Open Source - ready for business
Open Source on the Business Desktop
Why Open Source Software / Free Software (OSS/FS)? Look at the Numbers!
A Hard Look at Linux's Claimed Strengths
OLPC Australia
Basically, my opinion is the following: move your file servers, proxy servers and print servers to Linux.
Hold it! if you have any NT4.0 boxes in your building samba print servers WILL NOT WORK. there is a nasty bug in NT4.0 that keeps samba from sucessfully letting NT4.0 use the printers. This bug is not there in 3.11,95,98,me,2000 or XP.
Also any NT4.0 boxes will require the user to have a login on that linux server due to another nasty bug in the NT4.0 SMB software that requires a username and password and a "it's ok" acknolegement from the server to even view a share that is public.
I have fought NT4.0 in my enterprise for over 3 years because of this... and it is the fault of shoddy code in NT4.0 not samba... W2K was actually a breath of fresh air for me, solving almost all the SMB problems in the company.
windows NT = really really crappy networkable operating system.... and it will seriousally hamper any of your migration efforts.
BUT,Linyux + samba can easily replace all your PDC's and BDC's... its implimentation of the microsoft domain model is better than microsofts. that alone can cut TCO by a large amount... no more having 2 to 3 licenses for every workstation because of the PDC and BDC. (yes to be legal, when you log onto your machine it uses the PDC and BDC, and therefore requires another serveruser license... otherwise you are illegally using the server software.... Gotta love Microsoft Gotchas.)
Do not look at laser with remaining good eye.
That's funny - I have here a box running Redhat 7.1 that provides a public share and a PDF printing service, both accessed almost exclusively from Windows NT 4.0 desktop machines. I certainly haven't created Linux accounts for the 200-odd users on the network.
With Open Source to the moon and back!
What the software community thinks about Microsoft
Be free!
At my company, we am doing the same.
We are in a Win2000 Active Directory environment, and we are slowly replacing the file servers with samba server (which are really easy to join to a win2k AD domain, and use the AD authentication if you RTFM), moving the web servers to Apache, the firewall is now a linux box with iptables, snort/acid and 2 network cards, and so on.
The whole linux section is running webmin over SSL, so changes in configuration are easy to administer and make can be done even by the microserfs in our company.
The only struggling point is Exchange (as another post mentions), there we go into the realm of broken LDAP implementations in AD and X400 connectors, so we have decided to leave that alone for now (to be honest, exchange 2000 performs quite adequately) (hiss! boo!)
We did not do the 'sudden switch', we just made sure there was a *nix alternative the moment another winbox went belly-up. The users don't even realise something is different, and the new structure performs well.
I find TCO a vague concept; there are so many intangible factors involved. I do know that
a)we are rid of MS's expensive server licenses
b)we spend less time troubleshooting
c)the hardware requirements are significantly lower
Plus, I get to play with *nix boxes all day long! yay!
-- No Sig is a Good Sig
There seems to be a general smattering of questions around how vague my Ask Slashdot submission was. Let me clarify by saying that I'm not looking for anyone to do the cost analysis for me, or even tell me what it would be. Obviously the numbers are based on what components we have or what software we're tied to. That's a complexity of it's own. And whatever technical challenges there are need to be raised as flags, but at a conceptual investigation stage nobody needs to know those details to get a 50,000 foot view of the world. What I am looking for is really the best way to determine these costs and if anyone has some experience in doing such a move from one platform to another and what are some of the "gotchas" to look out for when doing this type of study.
True, a system where users are only engaging email against an Exchange server means pretty much nothing in terms of swapping out Windows with Linux and serving up POP3, but even in that simple environment there are costs associated with support, maintenenace, upgrades, etc. If it costs $40 million dollars to replace a Microsoft technology with a Linux technology over 10 number of years, I'd rather stick with Microsquishy where the support is there and pay the $3 million/yr for it (or whatever those numbers are).
I think one of the key points that we're seeing from looking at this problem is the fact that we're seemingly tied to certain products, not necessarily the technologies. So depending on the product rather than the service is causing a lot of grief in any kind of cost reduction. One comment that stood out was that large corporations were tied to Windows not because of IIS but because of Exchange. This is only partially true as most large corporations are tied to a series of products rather than technologies. It's not as simple as Exchange vs POP3 but more like BizTalk vs ???, portal technologies, SAP, etc. There is no one single solution in a corporate environment for all services.
Thanks!
I can't speak to your print server comment, but you are mistaken on the requirement to have an account on the Samba server to access shares from an NT domain machine.
Check into winbind - it even allows you to chown and chgrp files to domain users. I have a web server set up that way right now. There are no local user accounts on that system besides root, who cannot log in remotely (the system is in the next room, so access is not an issue.) On the NT side, you can take ownership and change permissions to an extent as well.
See www.winface.com for a book some ideas from the Unix world on TCO of Unix vs. Windows. Most of the arguments and calculations will work, and the TCO will be lower because you are likely to use x86 hardware with Linux, giving similar hardware costs to Windows.
There are some annoying errors in this book, but it is worth the $35 - I got a copy and it is quite thought provoking.
Most windows printing drivers (e.g. from lexmark) can use an "lpr" network port. So you don't need to use samba for printing.
no more having 2 to 3 licenses for every workstation because of the PDC and BDC. (yes to be legal, when you log onto your machine it uses the PDC and BDC, and therefore requires another serveruser license... otherwise you are illegally using the server software....
Where do you get this idea?
You can use Per-Seat licencing to reduce the number of CAL's you require - in Per-Seat licencing mode, you only require one CAL per workstation, not one CAL per workstation per server.
Also, a connection is only made to one authentication server at logon, unless that connection fails. An NT/2K Client will only establish a connection to a secondary authentication server if the primary server is offline.
However, I agree wholeheartedly that SAMBA offers a much more stable implementation of the SMB Protocol than NT. I have been actively using SAMBA on a number of servers in the office for File / Print services for a couple of years, and have had very few problems.
From our point of view, the big difference between NT/2K and SAMBA is the hardware requirement. Linux / SAMBA runs very happily, with less than 5% average CPU Load, on a P233MMX with 64MB of RAM. Serving the same content, NT4 was grinding to a halt on a P-III-300 with 256MB!!
We also use Linux for Firewall and Proxy services, however there are a number of systems that we cannot replace, in particular, we require MSSQL, so we need to retain a couple of NT/2K servers. Also, our virus scanning solution (Trend Officescan) requires an IIS server for it's automatic pattern file rollout (though this is behind our firewall, and is only permitted outside access to collect the updated pattern files for internal distribution).
Linux is also not suitable for certain VPN tasks - FreeBSD is a much better all-round solution, as its IPSec implementation is a better match to the official standard.
Linux has it's place in the server room, but so does MS, and other OS'es such as Free/OpenBSD.
Disclaimer: I meant what I thought, not what I wrote! What? You can't read my Mind? Oh dear!
I'm sitting here looking at a perfectly good alphaserver vintage 1994 that no one at work had any use for and hence I've inherited.
We use some BSD to squeeze a little more life out of a few aging x86 boxes for mail, etc. but to the best of my knowledge, nothing really touches linux when it comes to a.) hardware compatibility coupled with b.) application support and c.) a snowball's chance in hell of getting a support contract that will keep your TCO south of the prison rape that is MS licensing. But then, I'm the small fish...
In any event, if you have a mixed platform environment and a few older systems the capacity to leverage that hardware's remaining capabilities given a homogenous OS environemnt, open standards and fine-grained configurability creates the potential (in my mind) for a very excellent short term return on investment.
Furthermore, its a relatively low risk scenario for you and other management members to get good, hard fact data about what the realities of the differences in TCO is using linux in _Your Environment_ .
Check out http://www.rhsd.net . The project aims to replace the Microsoft Back Office programs...check it out!
Back in 2000 i was given the opportunity to go to London, UK to help a growing swedish IT consultant business. My main task was to act as a Linux mentor , to help and introduce Linux to the MCSE sysadmin there. During my 6 weeks, I replaced their current "firewall" (nt4sp3 with winproxy) with a Slackware/ipchains based firewall. This allowed them to
1: Remove the proxy software on all the clients
2: Provide some level of security for the DMZ
The firewall was up'n'running for over a year after I left, and then replaced by a Solaris firewall with Checkpoint.
My point is that I managed to introduce the power of Unix into the NT environment and easilly replaced the NT "firewall", and everyone was happy, including the business guys in suits who pays the MS licence bills.
Begin with attacking simple services, such as web, ftp, fw services. This makes a basic understanding, even for the "civilians". When they feel comfortable with the Linux fileservers and firewalls, "Hey, that fileserver never goes down ?", you will have a lot easier to migrate the rest.
Probable impossibilities are to be preferred to improbable possibilities.
Aristotele
Yes, I am one.
The rider is that you have to find roughly 3-4x as many Windows admins, and that in itself demands more managing than 1/3-1/4 as many Linux admins.
Let's put it this way: shop with 25 assorted servers has a choice of six Windows admins at (say) AUD$80,000 PA apiece or two Linux admins at (say) AUD$120,000 PA apiece. Quick! AUD$480,000 or AUD$240,000 PA for the same services, you choose!
Now let's turn to databases and email. Say that this shop has 5 of each and fifty seats on each, that's 250 licences for each, at a combined total of roughly AUD$300 a seat for MS-SQL plus Exchange, or AUD$750,000 (or a free Linux admin for six years riding PostgreSQL plus PostFix). It's enough to make an accountant go, er, postal.
Maybe you're not a Microsoft plant, maybe you're a Microsoft animal? (-:
Got time? Spend some of it coding or testing
Standard disclaimers apply
The comments below are offered in the hope that they will be of some use to the original poster and are not intended to offend anyone.
The opinions expressed herein are those of the author and do not reflect the opinions of the author's employer or any organization(s) the author may be affiliated with.
The TCO of replacing Windows with Linux will depend on the following factors:
1. size of your organization
2. complexity of your organization
3. your budget
4. your hardware (including your network hardware)
5. your software
6. your human resources (minus your IT people)
7. your IT people
If you have a small or medium sized organization with a good IT dept., Linux compatible hardware, flexible management, employees willing to accept something that looks and works a little differently and you are not running any strange / proprietary software which does not have an open source or Linux equivalent then you could migrate your whole organization with the assistance of 1 Linux/Hardware geek.
If that is the case you could migrate everything for the cost of 1 full time geek (30-120K/yr).
If you have a very large+complex organization, in order to achieve the best possible TCO you would need:
1. at least 1 Linux Guru/Master Geek (60-200K)
2. at least 1 Hardware Master Geek (60-200K)
3. at least 1 Linux Slave / Hardcore Geek (30-80K)
4. at least 1 Hardware Slave / hardcore geek (30-80K)
5. a budget sufficiently large enough to pay for the migration costs (an incremental rollout would cost more)
6. a small development team to code new apps and or drivers if you cannot find suitable replacements for what you currently have (1 to 5 people at 30-80K a piece)
If you migrate everything overnight that will have the lowest possible TCO.
If you migrate gradually you will end up paying more over time but that would allow your people more time to adjust thus reducing the human resources problems/issues.
If you have a small budget and you are not authorized to kill the patient in order to save it then your only option might be to migrate a small number of machines at a time and retrain your people as you go whenever necessary.
The best time to upgrade your OS would be if you are about to purchase new machines anyway you could then replace the old machine with the new machines incrementally fixing problems as you go.
In general Linux is awesome when used as a:
1. file and print server
2. public web server
3. intranet/private web server
3. firewall
4. router
5. mail server
6. database server
7. DNS server
8. network management workstation / server
9. authentication server
This is by no means an exhaustive list.
I can say no more without knowing more about your organization.
If you can provide more details about your organization without divulging the identity of that organization please do so. I am sure there are many slashdot.org members out there that could provide you with a lot more information if they knew more about your systems and your internal structures.
You might want to consider contracting an independent third party with good Linux and Windows knowledge to come in and inspect your organization and give you a guesstimate of what it will take to migrate.
Live long and prosper iII II
Unix_Geek_65535
This comment is, in my experience, inaccurate.
I've logged onto Samba-controlled domains and printed to their printers from both NT4 workstation and NT4 server.
Not sure what bug you're referring to, but I'd love to see a reference.
The central point of this, though: that Linux can easily replace windows for many functions, is definitely still valid.
While I don't disagree, I would love to understand this better.
Unless I'm missing the boat here, nobody loves Exchange. What people like is Outlook for mail and scheduling, which I find incredibly primitive and unproductive, but there it is--people do seem to like it.
The problems with Outlook on Exchange are obvious--Outlook is better described as a sophisticated virus-distribution platform than it is an MUA or a scheduling client. But if people love Outlook, they can use it with a variety of other scheduling platforms, including Lotus Notes Domino servers (with an outlook plug-in thingy) or even the Byrani stuff (see www.byrani.com).
Am I missing something or is it really feasible to replace the Exchange server side with minimal client reconfiguration and still get what people want?
http://www.compaq.com/tco/models.html
n .h tml
http://eu.conecta.it/paper/Impact_open_source_i
... and we have found that overall, it was a good experience.
We found that with our business(high volume, low margin) that the new licensing from MS would cost us several thousand dollars every few years, not counting the initial cost to get everything we already had in house "current".
Also, the cost of the hardware to run Server 2000 supporting all of our functions was also cost prohibitive as we would have needed to replace our aging HP Netserver LH3.
In the end we wound up replacing the Netserver LH3 with a pair of Linux servers. One running SAMBA, the other sendmail/POP3.
Overall, the cost of our server hardware was roughly 1/2 of what we would have paid otherwise. Cost of OS, $0. Cost of time and training to come up to speed and trouble shoot all of the ins and outs: 12 weeks @ 20hrs a week of my life (insert appropriate salary here).
Miscellaneous savings: No more weekly reboots (though we still do a monthly to insure everything is still peachy), we have confidence in the stability of our server OS.
Nothing is obfuscated, we can look at anything under the hood that we want to and modify it for our business needs.
Wealth of knowledge: Every error that I encountered along the way was solvable by doing a simple search on the Web.
Expertise: In order to accomplish this task, especially performed by only one or two individuals in your IT department, you will need to cultivate in house expertise. It will not be such that all questions will be answered as if by an Oracle, or even a Guru, but it certainly will but them on par with many of the people running around with their MS certifications.
Downfall: This was not an easy task to just "do". All of our IT folks in shop (myself included) are UNIX systems administrators, at least in the basic sense, and it still took a fair amount of time to untangle all the bugs.
I could not have imagined being told "We need that new server up in 2 weeks." and just doing it. Now I could bring said server up in 4-8 hours from scratch, but in the beginning it was a lot of trial and error.
-- You don't shoot to kill, you shoot to stay alive.
It's clearly defined in the samba documentation, and in all of their mailing lists.. NT4.0 has real trouble doing SMB printing to a samba box. the TNG fork of samba was supposed to fix this but cince the problem with printing did not follow NT to the NT5.0 branch it's a moot point now.
All windows NT systems will refuse to connect to a server without first asking the user for a login or password IF the current user/password used for the local login did not work on the remote machine/server... This is the first part making it a pain in the arse to print to a SMB server, and causing the requirement for a login on the machine (smblogin with a smbpasswd) to access the shares.. even if it is a public share.
granted I havent done NT4.0+samba for almost 2 years now... so my experience is based on what was happening before our switch to Windows 2000.
has samba added code to get around the bugs in NT4.0?
Do not look at laser with remaining good eye.
There has been mixed success so far..
The good...
- Significant drop in costs.
- Users are able to do their jobs still
:)
- Retained document compatibility with other companies.
- The email viruses are history now.
- Etc, etc
The bad...- The older staff members had a hard time learning the new system. They are getting there though.
- Certain custom systems were not immediately portable.
Overall I'd say it has been quite successful, for us at least--- Commission free trading & free stock up to $500 - use http://share.robinhood.com/kelvinp6
WHAT WOULD/DO I DO?
===================
UNIX for mission-critical apps but use Open Source and plan for Linux
LINUX for file, print, web, email, firewall
LINUX for middle tier apps and big development (easy port to big UNIX)
*LINUX for EPOS or "process user GUIs" ie simple GUIs/web pages
WINDOWS for "desktop power users" - those who use Office creatively
Final word to the wise:
Students of sales/marketing know that people make decisions based
on a bunch of personal psychological stuff - not always on technical
merit. My experience is that the guys who sign the check have a big
problem giving up their MS XP laptop (or whatever) as it gives them:
- Status with their peers
- "Membership" of the MS "club"
- Familiarity (theyve been using it since their degree dissertation)
- Confidence (NT was a bit unreliable but to them BlueScreen = normal)
Also, it's a MUCH easier thing to do at an installation that already
has extensive UNIXesque skills - far harder for MSCEs to learn.
I have marked * the key one here. Most are already sold on 1,2,3,5 as
sensible but 4 gives an opportunity to cut costs in warehouses, nurses
stations, ticket booths, etc. Why pay hundreds of $ for an OS when the
user only uses a browser?
We have just finished moving the entire infrastructure of a large UK construction firm over to Linux and Open Source. The TCO gains are enormous and fall into three main categories:
- Staff. Less maintenance/repairs means staff are freed for productive work (or staff numbers reduced).
- Licences. Pretty obvious really.
- Hardware. The upgrade cycle is _drastically_ reduced. Companies depreciate hardware - if you depreciate 300k over three years it costs 100k per year. If your hardware lasts twice as long that goes down to 50k per year.
The Company has entirely eliminated its windows servers and is very happy with the cost savings. They went on record with this in last weeks 'Computing' (UK industry magazine).We have the full case study here and several pages related to TCO at here
Making changes to the core network at an organisation is usually best done gradually. Few company directors will be willing to replace a complete network in one go. The major benefit of free software from a directors point of view is likley to be the massively reduced software license costs. Other benefits such as TCO, reduced staffing, reliability etc. are secondary.
:)
:))
Most IT costs are written off over time, so a Win2k server costing £4000 may have £1000 taken off its value in company accounts each year. Therefore two years down the line your server is valued as an asset worth £2000 in the company accounts. After four years have passed the server will not show up on company accounts. The amount of time purchases are written off over varies from company to company, so check how your organisation operates. In my experience few finance directors are willing to replace assets that still have value to the company, so don't plan on replacing that new Exchange server just yet.
When you try to convince your management to go with free software try to honestly compare different products/technologies. I have successfully implemented several Debian GNU/Linux servers running Sendmail at a company simply by comparing the product costs in front of company directors. The comparison can be quite simple, for instance the one I used for proving email simply put the costs of Microsoft Exchange, Lotus Notes and Debian GNU/Linux Sendmail side by side like so:
License costs
Microsoft Exchange £50,000
Lotus Notes £15,000
Debiam GNU/Linux £0
This will of course provoke an argument popular with company management that the free option must be cheep and nasty. So you will need to be able to show that *NIX has been perfected over thirty years, or that the server sofware you are choosing is the best in its class for your purposes.
Once they are used to the idea you can introduce added extras such as increased reliability, improved staff motivation and management benefits like cost/performance improvement and less management overhead.
For some more hopefully helpful information look at http://www.siriusit.co.uk
Comment caveats:
1 I am not impartial, I spend most of my time implementing free software solutions so I may have a slight bias
2 The company I work for http://www.siriusit.co.uk specializes in free software implementation so they may have a slight bias too
Steve Peters
I'm sitting here at work browsing Slashdot with 'Zilla on my Redhat mail/web/SQL server next to my Windows 2000 box that's here simply to store files for about ten users and serve up a couple of web pages now and then. IIS blew up again a couple of weeks ago; now it runs Apache and that's one less thing I've had to check every day. I'm reading Slashdot because I have to keep an eye on the 2000 box while I re-reinstall Service Pack 2. Oh, look! It seems the installation has hung again. I'm installing SP2 because SP3 includes some things I don't like; mainly the ability to download and run code from Microsoft any time it feels the need to do so. So I have to apply each relevant patch as it gets posted. Nothing new there. I was doing that anyway. I'm reinstalling the Service Pack because yesterday the box decided it would pick and choose which apps it wanted to run at random, and a virus scan (Virus scan - Have I even done that on the Linux machine? Have I had a reason to?) yeilded no answer. Both of these machines are behind a hardware firewall, with only a handful of ports going to the Windows box, and yet the Windows server will still catch whatever the Malady of the Month happens to be most of the time. I'm venturing to guess this probably comes from all of our Outlook-happy staff storing files on it, but I'm thinking they'd never know the difference if suddenly Samba starts handling all of that and the worms have nothing to run on. I'm glad I was on an OSS kick when IIS failed, else I'd have probably been here all night trying to make it work before reinstalling the OS and restoring data from a backup. I've yet to know why this happened, but I don't care anymore. Ultimately, even when it's working like it's supposed to, our Windows server requires constant attention. Sure, I spent two weeks setting up this Linux machine to do what I wanted it to since I'd never been exposed to it before, but I haven't had to lay a hand on it since then, except times like this one when I catch up on Slashdot while waiting for this sad waste of hardware next to me to get its act together. Wow, it's already 10am, the 2000 box is still posessed, and I still haven't gotten any work done. I shudder to think what this would be like if I had to deal with these issues on a large scale. I have one word for you: Yes.
The company I work for saved a tremendous amount of money by setting up Linux Terminal Servers. So now our entire call center runs Linux and we don't have to pay the high costs of "Software Assurance" from Microsoft.
The biggest issue is, in my mind, use common sense. Make sure you have a better understanding of your current situation (systems and how people use them). In most cases, don't make all the changes at once - plan to do things in stages, test things out before you depend on them, then deploy - and examine how that stage went so you can adjust your plan for the next stage. Maybe you start by replacing a few servers, for example. If you're replacing desktops, maybe you start with just a few systems, or you replace Microsoft Office while keeping Microsoft Windows on a few systems.There's much to be said for incremental changes.
- David A. Wheeler (see my Secure Programming HOWTO)
Generally it's only at the server level, although we've installed a few workstations on a "try it to see if they'll like it" basis. Servers are generally a no-brainer unless they are running some server-side MS-specific utilities.
We've found that, when asked, virtually every Developer will claim that their application will not run if Linux is the file server. In all but one case they were dead wrong. (The one case was an application that ran on FoxPro on the server.) Most applications have no clue what OS the files are stored under and couldn't care less.
The downside to switching clients to Linux has generally been a reduction in our income from that client. One client even uses the "mail" feature of Outlook (mailing contacts and appointments to other members of the group) which generally sucked until we installed a nice Dell server which we loaded with SuSE Linux. Just like Exchange but without the costs. Also, unfortunately, without the headaches because they now call us for help only about twice a year!
From almost any standpoint you can mention (original cost, administration costs, utilization of platform, etc.) Linux comes out ahead. There's even damn little training!
Try it on a few workgroups at a time and see for yourself.
No one ever had to evacuate a city because the solar panels broke!
If you change the design from desktop computers on peoples desks to having smart terminals on peoples desks that just run an X window system terminal then you can save a fortune.
For 1000 people the costs break down like this:
1000 desktop machines with 17" monitor at $300 apiece. $300,000
20 application servers at $2000 apiece. $40,000.
OS, Free.
5 FTE's at $80,000 to maintain everything. 400,000 a year.
The application servers are all on a gigabit network for talking to each other and use a distributed network file system to ensure that the file system is always up. All computers on the network share processing power using Mosix or something similar so that all possible processing power is used.
Now, look at the windows solution:
1000 desktop machines at $1000 apiece. $1,000,000
50 servers at $2,000 apiece. $1,000,000
OS and software costs $1,000,000 a year
20 FTE's @ $40,000 $800,000
Project both over 10 years. Don't forget to replace all windows hardware every 3 years. The Linux hardware will be good the entire 10 years except for hardware failures.
Over 10 years:
Windows $24,000,000
Linux $5,000,000
So, as you can clearly see for 1000 desktops Linux will save 19 million dollars over 10 years.