Windows cheaper than Linux, says Microsoft (vnunet)

Posted Nov 5, 2002 23:56 UTC (Tue) by mly (guest, #2171) [Link]

It would be interesting to know how they counted though...

Not much substance in the article. It all reminds me of when I suggested to the CTO of the company where I then worked that we should move a system from VAX to Unix, since the new hardware which we desperately needed was so expensive if we stayed with VAX. I was asked to get a quotation on a Unix machine, and the IT manager (who sat in DEC's lap) should get a quotation for a VAX. Initially the VAX was 5 times more expensive, but as he stalled us the price dropped quite a bit. Not getting close to the Unix price, but the IT manager claimed that there would be so many additional benefits with the VAX, so that from the perspective of my system, the VAX would have zero cost! We bought the Unix system...

Ok, licensing costs are only 5% they say. Is that the MS licensing costs, or does that include all licenses? In the practical case, a Linux server will probably run a lot of other free software, while the Windows box would have other commercial grade software as well, whether it's database servers or something else. Still no effect?

What about hardware costs? Surely they will be higher for Windows systems. Ah, but that's probably just a few % as well?

So, much better support for automated and remote administration in Linux, what's that going to mean?

They probably come to the conclusion that the Linux admins have much higher salaries (because they deserve it?) which will negate a more efficient system.

That the attitude that any clown can handle a Windows box, 'cause it's got GUI, will lead to all sorts of problems when underqualified people are made to handle systems they don't understand doesn't cause a measurable cost?

Security then, all problems with virus attacks and anti virus software that slow down or crash the systems. "Hm... separate issue, doesn't go into our TCO template!" ???

Maybe they compared average costs for all computers, and Windows machines were 95% desktop machines with just Office, and the time spent fiddling with configuration and wasted due to bugs by the users doesn't count since they aren't on MIS dept budget. On the other hand, most of the Linux machines were mission critical servers with thousands of users? That could certainly make the linux boxes more expensive per machine... :)

Posted Nov 6, 2002 1:56 UTC (Wed) by AnswerGuy (guest, #1256) [Link]

Unfortunately I cannot post a substantive comment on this particular article, because VNUNet's site isn't responding. (Slashdotted?)

I have a knee jerk desire to see things from "the other" point of view. So I immediately tried to see what TCO (total cost of ownership) factors could be included (and omitted) in order to reach the "MS Windows is cheaper" perspective.

Obviously initial licensing cost goes to Linux. The most expensive commercial mainstream Linux distribution is a fifth the price of a single system license for Win2K. Generally that single shrink-wrapped box is good for the whole site, for installation on an unlimited number of clients and servers.

Someone might complain that such a installation doesn't include media and user manuals. However, MS Windows also lacks printed documentation at this point, and most vendor pre-installations of MS Windows are now media-less or include these specially locked "Rescue Media" which are matched to a specific machine model or even to a specific machine (via CPU ID or motherboard ethernet MAC address?). These latter media are a nightmare for internal helpdesk technicians.

Initial deployment costs are separate from the licensing and aquisition of the software. In rare cases MS Windows might have the edge here because the OS is pre-installed on most systems that can be purchased. However the majority of sites, even small offices with less than a dozen systems, have to re-install the OS (or wisely choose "blank" --- no-OS preloaded) systems. So the question comes down to, which system is easier and cheaper to install.

Norton Ghost is the de facto standard for installation of MS Windows for corporate deployment. Of course it's a commercial product. There are several packages for automated installation of Linux. The most common is built into the Red Hat Anaconda installation package. The feature is known as Kickstart. While kickstart can be a a little bit quirky, it is powerful enough to fully automate the installation of clients and servers. Kickstart is pretty flexible, it can be run from floppy or CD, or (with a little extra work) it can be run over PXE (the Pre-eXecution environment supported by most modern ethernet cards), or etherboot.

Something like kickstart can be more difficult to use than Ghost, and there is a much smaller base of people with the expertise to fully automate it. Therefore, it can be view as the higher cost alternative, even though Kickstart itself is free and included with every copy of Red Hat (and the Red Hat derivatives that retain some derivative of Anaconda. However, the one-time cost of configuring a server for the most advanced and difficult form of kickstart and the occasional effort to create new .ks files for new types of machines is amortized over all the system installations (and especially re-deployments). A $1000 spent on a top notch Linux consultant is still less than only two or three site license seats for Win2K. [I create custom kickstart configurations for my customers, the basics are generally done in one consulting day, including full training for a couple of local admins].

Of course this seques into the major TCO fudge factor: support costs.

The usual claim is that Linux (and UNIX) is far more complex than MS Windows, and therefore Linux expertise is at a premium. This has been the usual accounting trick to make Linux appear to be more expensive. However, there are now millions of Linux users. For the last several years Linux has been used widely at colleges and universities, and now it's finding increasing use in high schools and at the elementary level. The kinds of people that are naturally attracted to help desk, systems administration and tech support roles already know Linux.

Also it is generally recognized that the number of Linux systems that can be managed by a single admin is greater than the number of MS Windows systems. The two major factors that account for this are Linux' overall better robustness, security and reliability (once the systems are set up correctly, there is simply less support required of them) and Linux' remarkably good remote administration capabilities (notably ssh and rsync).

However this factor highly variable and difficult to quantify. An average help desk technician or sysadmin lacks the training and expertise to really benefit from Linux' scripting and administration features. In general we can expect that there will be additional training needed to allow technical support personnel to take advantage of many features that Linux has to offer. Something as simple as shared remote screen sessions (like the old PCAnywhere "co-pilot" mode) are surprisingly underappreciated by Linux support people. [Hint: the screen package can easily do "multi-user" sharing for text mode (curses) applications, and VNC is the tool of choice for shared X sessions].

One factor that is clearly in Linux' favor relates to requisite hardware costs. Linux runs adequately on older systems with less disk space and less memory than similar tasks under most versions of MS Windows. In addition there is a markedly slower "inflation" of these requirents with newer Linux releases, and there is markedly less pressure to upgrade the full OS on most Linux systems. Generally there continue to be official security upgrades and patches to a Linux systems for a couple years after the subsequent versions have shipped. Thus a Linux system is fully supported with security patches for about 4 years, can be unofficially updated indefinitely, and an upgrade the the newer version wil generally be less disruptive and require fewer hardware upgrades than similar circumstances under MS Windows.

Related to this is the descreased need for ancillary equipment. It's possible to run Linux systems "headless" (with no keyboard, monitor or mouse) and to use serial consoles. This can lead to significant savings by eliminating the costs of KVM switches, and cables, not to mention the expensive rack space and additional labor consumed by those extra cables.

Serial terminal servers are generally more compact and less expensive per port, and only take one slim cable per machine. Modern terminal servers such as the Cyclades, run an embedded Linux and support ssh. Modern motherboards and chipsets even offer serial console support right into the CMOS setup (though it's immature compared to the classical UNIX servers and workstations).

Obviously Linux offers compelling advantages on the server side. However, the trend towards Linux dominance in the server room is almost unquestioned. The major threat to that dominance relates to server interoperability to the desktop. Microsoft's ActiveDirectory services were/are one major attempt to force customers to use their OS on servers by limiting client side support to a set of proprietary protocols. Eventually Microsoft will come up with the mixture of client side features that will require Microsoft servers. Eventually they will limit the interoperability of those servers with "impinging" Linux and UNIX servers (that's what TCSA is all about to them).

Thus the real battle is on the desktop.

Linux on the desktops still seems like an imposing goal. However, it is no longer dismissed as an impossibility. [I personally have used Linux as my desktop system for about a decade, but I'm unusual in that regard.] The recent releases and advances in Mozilla, StarOffice (and OpenOffice.org, Evolution, Gnumeric, Abiword, and the GNOME and KDE environments, along with the packaging of WINE into Lindows, Xandros, etc., has been covered in numerous places.

How does this relate to TCO considerations?

Most of these are free packages. StarOffice (the only commercial application among those listed) is a tenth the price of MS Office. The Lindows and Xandros distributions are signifantly cheaper (half to a fifth?) the cost of their MS Windows competitors.

So, the initial outlays are comparably in favor of Linux. However, a Linux desktop system running these feature-laden "big" packages does require hardware that's comparable to, possibly even slightly more extensive than similar MS Windows desktops.

Microsoft proponents will quickly point out that the patchwork of Linux applications lacks "integration" (for "ease of use"). The criticism is well taken. Linux isn't going to force it's user base into a proprietary "integrated" approach that forces all Linux desktop users to it's form of single sign-on directory services, to force them to use a specific mail transport agent or limit their access to group scheduling, etc.

This is an area where Linux needs considerable development, however. It would be nice for some company, such as Red Hat, to offer "enterprise deployment kits" --- a sort of distribution that installs a whole network rather than just a single workstation or server. Such a kit could provide an integration of all the common services (LDAP, DNS, mail, intranet/web, database, SNMP/network management and monitoring, asset tracking, ssh/Kerberos, VPN, etc) into one pre-integrated mixture. Naturally, such an offering would entail alot of assumptions --- it wouldn't fit all the needs of many customers, and probably wouldn't fit all the needs of any major enterprise. However, it would be open sourced, extensible and customizable. Other companies and groups would create competing mixtures with variants (OpenNMS vs, Big Brother, ssh vs. Kerberos, NIS vs. LDAP, etc).

Still, the key element to TCO in current offerings relate to lost productivity due to learning curve and "non-integration" issues.

In one TCO "analysis" (or "propaganda piece" as the some may deem it) there was considerable expense attributed to lost productivity as a result of incompatible file formats. In other words, the major TCO discrepancy was caused by the projected need for users to constantly import MS Office .DOC, .XLS, and .PPT files into OpenOffice.org or similar applications. A large part of that lost productivity stems from the notorious limitations of these import and conversion filters.

This is a clear case of artificial legacy. It becomes a prisoner's dilemma. If enough people switch to open document formats than the conversion and import tools (and native file format support) will set us all free. Until a critical mass of mainstream users do so, there will be a considerable, possibly overwhelming, cost attributed to the constant issues that are raised by this activity.

Naturally Microsoft has been aware of this for years. They've used new releases as opportunities to incorporate file format changes and thus force many sites to upgrade ALL of their clients when a few key users upgrade their office suite. Of course this has lead to many sites adopting a policy prohibiting use of newer Microsoft Office versions pending IT review, aquisition and deployment processes.

The major issue in TCO that is probably deliberately omitted from any propaganda that puts a Microsoft dominated solution ahead of Linux has to do with Microsoft upgrade costs and risks. It is clear that Microsoft will continue to increase their prices, and that it will exert further pressure on its userbase to force them to adopt their subscription model. It is clear that Microsoft will continue to use updates as leverage to force companies to purchase unrelated upgrades to other systems and subsystems. It is likely that some "security" patches will entail ancillary upgrades (non-free to non-subscribers).

Basically, there are huge financial risks that should be added to any MS TCO analysis. That is where TCO propaganda (FUD) differs from honest analysis and where this article is probably worthy of Enron's accounting.

Posted Nov 6, 2002 13:31 UTC (Wed) by QuaintRcky (guest, #7518) [Link]

Posted Nov 6, 2002 15:12 UTC (Wed) by dbhost (guest, #3461) [Link]

Now on to Windows.
Windows 2000 Professional Full Version CD $299.00
Microsoft Office XP Standard edition (Word, PowerPoint, Excel, Outlook) $469.99
Norton Anti Virus 2003 Professional to keep $39.99
Adobe Premiere 6.5 $629.99
Adobe Photoshop 7.0 $699.00
The total for a Windows system to do the basics of what I do with a computer is...
$2136.98
The cost of training for Windows is $0.00, I still have access to plenty of books at the library.

So let's run these numbers out over five years.
Annual donation to my favorite distribtuion company $120.00 X 4 not already accounted for $480.00
Steady software upgrades from vendor and vendor's parters $0.00 as they are included in the club membership...
Total over 5 year ownership $748.00, again being very generous to the vendors.

Initial investment for Windows $2136.98. New versions of Windows every two years at $199.00 per upgrade, or $398.00 over five years, Annual NAV subscription renewals of $15.00 per year, or $75.00 over five years, Premiere and Photoshop upgrades every 18 months at $189.00 per application or $1134.00 over five years. Also assume in that time at least one Microsoft Office upgrade at $260.00 That would bring the tco for Windows over a five year period to $4183.98.

Now divide the cost of the Microsoft + popular apps over five years by the cost of being generous to Linux vendors and the Microsoft option costs almost 5.6 times more than Linux! And that is with me chosing a more expensive route than is neccesary. I could have opted to not go the club route, (poor netizenship though...) I could have used a free Anti Virus for Linux, and I could have used Broadcast 2000 instead of Main Actor, I simply like Main Actor better...

So after reviewing the basic math invovled in this basic example, and if Microsoft honestly believes Linux is more expensive to use than Windows, then anybody that would trust anything that runs on Microsoft systems with Mathematical equations is a fool. Because it is obvious that they simply cannot do the math.

Posted Nov 14, 2002 9:17 UTC (Thu) by akukula (guest, #3862) [Link]