A few months back, in my “Shutdown” column entitled “A Mile in IT’s Shoes” (http://www.linux-mag.com/2005-03/shutdown_01.html
), I relayed the various technical challenges I was encountering while working on a pilot project to deploy a thin-client Linux
desktop for a large, institutional customer. At the time, the pilot program was ongoing and the jury was still out. Now, I have some bad news: the pilot failed and the customer decided not to deploy a Linux solution.
While the solution was sound and my customer liked what was delivered — indeed, the IT director of the company and many of his staff praised the solution for its innovativeness, its superior performance, and its excellent stability — the killing blow was the myth of superior Linux Total Cost of Ownership (TCO) and misconceptions about the ease of migrating users.
A lot of people are under the impression that because Linux is a “free” operating system, it should be cheaper to deploy than an equivalent Windows solution, with its pile of add-on products such as Microsoft Office, virus scanners, network client access licenses (CALs), and so on. But as it turns out, after everything is said and done, in a large desktop deployment, guess what? Linux costs the same!
How can that be, you ask?
Sure, if you use OpenOffice.org, you don’t incur an expenses for a desktop suite. And desktop licenses of your average Linux distribution are, what, $100 apiece? And with Linux desktops you don’t need virus scanners or CALs. So, in theory, those savings should be easy to realize, right?
Well, as it turns out, your PC hardware costs about the same. In large organizations, the per-seat cost for Windows and Microsoft Office are reduced because of the large volume Select agreements. And in my customer’s case, since OpenOffice.org was an unsuitable choice due to file format issues, it would have cost an extra $60 per user for Crossover Office to run Microsoft Office XP. Ka-ching! The discounted price for Windows was pretty much the same. And lest I forget — the Active Directory integration software that was chosen, Centrify (which I highly recommend if you need an identity management consolidation product), would have run $50 per seat for full blown desktops, or in the thin client configuration, $600 for the server and $1,000 for the Windows-based management console software. Oh, and access to Citrix applications? Yep, more licenses.
And then there was the issue of technical support. Novell, who admittedly was excited about a potential testimonial customer, offered special pricing for the pilot project: $40,000 for a special thin-client pilot deployment for 100 seats, with a non-dedicated engineer and a central call-in number usable during normal business hours.
Now a semi-dedicated, locally-based customer support engineer, who typically would divide his time between my client and a half dozen other clients, would have cost $150,000 per year — an expense that was out of the question.
And all of this before hiring a regular, on-staff Linux desktop and server support person to support the pilot. That salary would likely have been $80,000 per year.
Suddenly, a Linux solution is starting to put a large dent in my client’s wallet. And to top it all off, there’s the issue of company politics and the fear of the unknown. Windows, for whatever it is, is a known quantity, and end-users and their overworked IT support staff like it because it’s familiar.
People who aren’t familiar with Linux but who’ve heard great things about it assume it can replace Windows, with a nearly identical feature set and similar desktop experience. This is all fine and dandy, except that when Linux behaves in a different way or something is implemented in a different way, or the application feature set is not complete and requires workarounds (like dealing with remotely-mounted home directories on existing Windows servers), or if it’s expected to interoperate with a pre-existing Windows network and with other Windows clients in a “seamless” and fully transparent manner. These are precisely the areas where you can run into problems, because it’s those grey areas where Linux isn’t completely polished yet.
Although its paradigm is arguably similar, a Linux desktop is not a Windows desktop, and it requires a different management methodology with a new learning curve. You can try to communicate this to potential end-users as much as you’d like up front, but until they actually experience it for themselves, they aren’t going to fully understand it.
Are all Linux desktop deployments doomed in the same fashion? I think those of us who expect to be able to shoehorn Linux desktops into an IT environment and replace an entrenched Windows infrastructure are probably a bit deluded. There are many areas where Linux still requires a lot of work to integrate.
But I still feel there is a place for Linux on the corporate desktop, and arguably, with more time and more polish, the issues I encountered will become moot.
Don’t worry. Jason Perlow is still keeping the faith. You can reach Jason at