OPINION: The total cost of the unending Linux-Windows argument is a great big headache--this latest attempt leaves us none the wiser.
A few weeks ago an Australian company called Cybersource released a study which claimed to compare the total cost of ownership of running a 250-person organisation's IT resources on commercial (mostly Microsoft) versus open-source/free software.
At Technology & Business magazine we're constantly reminded how important the issue of TCO is to IT managers, and that just looking at the upfront cost never gives you the complete picture.
This story was reported in local IT media such as the AustralianIT site, and even further afield, like The Register in the UK. For some reason, Slashdot ignored it.
Most people upon reading a report like this would--or should--wonder what sort of organisation Cybersource is.
According to the company's Web site, Cybersource is "Australia's leading provider of IT services in the areas focusing on Linux, Unix, TCP/IP, and Windows". Ahem. If that's not enough of a giveaway, Cybersource's CEO Con Zymaris has in the past offered to write a Linux column for ZDNet Australia--for free. Impartial? Hardly.
Most people would also want to look at the assumptions made in the study to see if they were realistic.
The study claims that, all other things being equal, the savings on software client licenses minus the additional cost of Linux-savvy staff would come to US$251,393.55 over three years. All other things being equal. Right.
Let's say every bit of commercial software in an organisation can be replaced with free-software alternatives. Leave aside that the study doesn't even consider support contracts, and claims Linux is not vulnerable to viruses. Ignore that it quotes off-the-shelf prices for all the applications, when a company buying 250 copies would pay considerably less. Why quibble over details?
The 245 desktop systems in this organisation would have Linux OS installed, OpenOffice instead of Microsoft Office, GIMP instead of Photoshop, and so on.
Sure, the savings on licences would be enormous, but what of the productivity losses in retraining all the staff to use an unfamiliar interface and applications? Does the study assume everyone in the organisation could make the transition with no training, without ever having to call helpdesk or read a help file?
Then there are the ongoing productivity losses when staff discover that even though OpenOffice is "compatible" with Microsoft Office, that doesn't mean "compatible all the time for all documents".
Likewise, all the commercial server software in this organisation is replaced with free-software alternatives: Windows 2000 Server with Linux, IIS with Apache, SQL Server with PostgreSQL, Exchange Server with Sendmail, and ISA Server with iptables. The savings on client licenses would be impressive, but there are hidden costs the study once again conveniently ignores.
It doesn't take into account, for example, how much more time IT staff would have to spend installing, configuring, maintaining, and integrating all this software using Linux.
Friends who work in IT have told me horror stories about having systems down for days while trying to work around library incompatibilities.
They've told me about finding an obscure reference in a French Linux discussion group about a configuration file in a totally illogical location without which the application won't do what it's supposed to.
Given there are only so many hours in a year, if the same number of staff are expected to maintain the same sized organisation, they'll either be working much longer hours, or the company will have to hire more staff. Staff who are overworked are more likely to leave, thus there's the enormous expense of recruiting replacements.
These are some of the more obvious holes in the methodology. I'm sure there are others.
Before you Linuxheads out there start bombarding me with hate mail, I'm not saying Linux has no place in the enterprise, just that there's a tradeoff--the software is free, time isn't.
I'm also well aware that Linux is fighting a propaganda war against an enemy with vast marketing resources and the social conscience of a uranium mining company executive on a dirty weekend. But fighting FUD with FUD leaves you with as much credibility as The Enemy.
In response to this piece regarding our Linux vs. Windows: Total Cost of Ownership Study (which showed that Linux as 34% less costly than Windows to run) I wanted to make a few clarifications and highlight some issues:
1) Cybersource is not a firm of analysts, but 'in-the-trenches' techies who have over a decade's experience in managing and supporting both platforms.
2) I can categorically state that Cybersource received no payments to research and produce this or any other report. Few supposedly
'independent' analysts can do likewise.
3) Josh claims that the report was anti-Microsoft FUD, but cannot fault
the numbers given nor the process by which they were arrived at. In return, he proffers once-off opinions provided by friends. In short, it's not FUD if it's actually true ;-)
4) Further, he misread the Study vis: "IT staff would take more time to deploy Linux...". Linux-trained staff were enacted to operate these systems, being efficient in their deployment and support, thus obviating
this issue.
5) Josh claims a "time and training burden" for Linux on the desktop, but offers no research or proof. My own research and discussions with a number of organisations in the US, Korea, Finland etc. indicates that
"minimal or no retraining" of end-user staff is needed to migrate non-power-users from Windows to Linux. Specifically, discussions with the IT Manager for Mastertrade (an appliance wholesaler with 47 branches
across New Zealand) indicate that for their migration of 300 workstations, no retraining was required.
6) Josh claims that "a company buying 250 copies would pay considerably less" due to volume discount. I disagree. Take for example the recent Victorian government licence deal, for 40,000 Victorian public servants, to use essentially the same software as was reported in our TCO study. We calculated a cost of around AU$2,000 per user for the Microsoft solution. The Victorian government deal, which one would expect for 40,000 users should attract the highest possible discount,
worked out to cost AU$80 million, or AU$2,000 per user. I don's see much of a volume discount at all.
In closing, a proposal to the staff of Technology and Business: conduct an independent study into how easily a group of Windows users can perform a set number of day-to-day business office tasks on a modern
Linux workstation, with no training and with no help. You may find the answers are a pleasant surprise.