When I was studying advanced statistics as part of my graduate school program I learned an important lesson -- don't be like the boy with a new hammer who tries it on everything. Not every statistical technique is good for every situation. One technique may be perfect for one situation and another for another situation. I think the same thing applies to deciding when to choose a particular computer operating system or application software...
Linux may be like the new hammer, particularly for Unix users. In some cases, moving from Unix to Linux can make lots of sense. The move can enable an organization to save money on licensing fees. Or, it can help an organization take advantage of open source software especially adapted for Linux. Or, it might be an opportunity to upgrade to a specific computer box that is important to an organization's computer strategy. Another possibility is that it might position an organization to take advantage of known future requirements, such as when an organization's software provider is moving to Linux and plans to eliminate support of its Unix products.
But it is important not to be blinded by the advantages as seen in the eyes of others. You have to look at the realities for your own organization. When Windows NT Server began to take off in the mid-1990s, lots of organizations dropped NetWare and rushed to Windows NT Server -- sometimes at significant cost in terms of adapting software and retraining users. Other organizations stayed with NetWare and made their initial investment in hardware, software, and training pay off much, much longer. I know of a very successful urban community college that has stayed with NetWare and has won respect and awards for their stable environment and advanced networking capabilities. Over the years, NetWare suited the needs of that community college.
Here are some scenarios in which it does not make sense to migrate from Unix to Linux:
Making the most of the current investment
You purchased a new Unix box, say from HP, Sun, or IBM within the last couple of years. Your box runs well. You're software is meeting your needs. Your users are trained. It makes more sense to enjoy the investment until it is not meeting your needs or until you have a different business strategy that requires Linux in the picture to fulfill. I know of one organization that purchased new business systems, including the database. They spent a year transitioning to the operating system and database, at significant cost in personnel time. The database applications were tested and working well. Then they decided to change operating systems and database vendors to save on licensing fees. They did save on licensing, but the cost of database programmer and user time, plus the added user disruptions caused them to lose money.
Specialized hardware requirements
Your software systems are tailored to special features offered on an HP, Sun, or IBM Unix box. A migration means spending hours finding and adapting those special features in the program code, recompiling, and testing. The programmer and user leg work can be very expensive in terms of lost time. The disruption to users and to your business can also be serious and therefore expensive.
Big iron demand
You have a huge database that is critical to the functioning of your organization and it needs a powerful processor on bigger iron than x86. Leave the database on the big iron and Unix, but implement Linux boxes in a data warehouse environment for reports and other data access needs. Organizations in this category would include: 1) a company that does a large Web-based and mail-order business, that has a huge customer database, and that does intensive selection and sorting of customers for specific promotions and 2) a large university that has a huge database of alums and uses the database to raise money.
Application analysis points to painful transition
Your application software programmers are very worried about adapting the software for a migration. Their analysis of the required software changes sends a red flag that the transition could be extraordinarily disruptive and expensive. Instead of doing all applications and systems at once, consider starting with one or two systems as an experiment to see how things go.
Critical software not adapted to Linux
You use a software vendor and the vendor's software is intricately woven into your business operations. The vendor's software is not adapted for or supported in Linux. One example, is a small credit union that uses third-party software supported only in Unix and that has no on-site programmers (I'm thinking of a credit union in my town in this situation).
With all of this said, I think Linux will prevail over Unix within the next five years and is clearly growing very rapidly in all IT sectors. If I were a CIO writing a five or ten year IT plan, and my Unix systems were currently stable, I would still include Linux (and open source software) somewhere in the picture five years down the road. In the meantime, I would watch the progression of Linux and open source software in terms of how they might meet my needs -- and I would revise my plan each year.
ABOUT THE AUTHOR: Michael Palmer is a professor at the University of Wyoming and co-author of Guide to UNIX Using Linux, Third Edition (Course Technology).