So you've hired a new Unix administrator recently and she's pretty darn good with Linux. The network team also has strong Linux talent and is hankering to use it. Your budget-meister is dropping hints about lowering costs. Does that mean it's time to move to Linux?
In this article, we'll discuss how to properly evaluate a potential migration from Unix to Linux. We'll see when a migration is or is not a good idea.
Reasons to migrate
There are three reasons why companies may want to consolidate and/or migrate to a different architecture.
Something's got to change if your servers are crashing all the time; you're talking too much in reliability meetings; you can't meet service level agreements (SLAs); and technical support is telling you your operating system (OS) is no longer supported.
If your current operating system is outdated, then you may be forced to either upgrade or migrate. This might be the perfect opportunity to plan a modernization effort.
What are you spending to maintain your current environment? If you count hardware, software, application licenses, maintenance and storage, it's probably a lot more expensive than you think.
At the end of the day, the decision to migrate or not may depend on return on investment (ROI). Can a migration reduce your current expenditures by a significant amount? Project management offices (PMOs) will prioritize the projects that deliver the best ROI for the company, so you must be prepared to justify the investment.
Whenever a new CIO comes onboard, after a predictably short honeymoon/reflection period, you will start to see different vendors passing by his or her office. In many cases, regardless of how well the existing environment is performing, the new CIO will want to make his or her mark on the corporate technological infrastructure. And that may mean a migration.
Researching your options
If the decision has been made to re-architect your server farm, you'll need to research the options available to you. Naturally upgrading your existing architecture is a possible solution. You can leverage your existing staff's knowledge about the infrastructure as you upgrade to more powerful and reliable systems.
A second option would be migrating to another Unix platform. A third would be migrating to Linux.
Suppose you are with a large Sun shop, and your studies have shown that you could potentially save millions by consolidating and migrating to IBM. (This is just an example, and arguments could be made for Sun's advantages, too.)
In these hypothetical evaluations, you've seen that IBM's POWER5 architecture supports both Unix (IBM's AIX) and Linux (RHEL4 and SLES9) on the same server, so you would have more options at your disposal than you would have had by staying with Sun or migrating to HP.
If you already have strong Linux engineers, running Linux on this hardware platform might be perfect for you. IBM pSeries servers enable you to run multiple logical partitions on one frame (server) and even to micro-partition a single CPU to host both Unix and Linux applications (using IBM's Advanced Power Virtualization).
For database applications that need maximum reliability and performance, you might consider Unix partitions. For application and/or Web servers, a Linux partition might be the right call here.
The right reasons to move to Linux
Don't move to Linux because you heard it was free. Linux distributions from SUSE and Red Hat are certainly not, and neither is the phone support. Don't think about moving to Linux because you love the Linux PC you have at home or because your technical people like it. There must be a real, compelling reason to migrate to Linux from Unix.
If the direction of your company is to scale horizontally and/or you're uncomfortable with big iron, you may want to consider blade technology. A high performing, clustered blade solution running Linux might be the best solution for you.
One of the biggest selling points of Linux is its ability to run on commodity hardware in the form of PCs as well as the big iron.
With some flavors of Unix, you're forced to buy the vendor's associated hardware (at a premium price). Almost all Linux distributors do not sell hardware, so there's no hardware lock-in. With Linux, however, you may have limited peripheral choices on the client side because of the scarcity of drivers for some Linux distros. (Note: Some hardware vendors do package their own Linux distros with their servers or PCs.)
Your objective must always be to leverage technology solutions that will align themselves with corporate initiatives. Does your company want to use only one vendor's proprietary applications? If that vendor also sells Unix, then you might as well get the benefit of using an OS tuned to those apps. However, if your organization wants to choose best-of-breed applications -- whether they're proprietary or open source -- Linux is a good choice. Also, if your goal is to reduce costs, Linux fits.
Don't think that Linux is just for small applications in departments. Linux is definitely an enterprise solution option for all types of server consolidation and migration projects.
Do the studies, talk to vendors and chart out at least four different technology scenarios. You may find that your best solution will be less of a complete Unix, Linux or Windows solution, but a mix of all three. As long as you can justify the environment from an ROI cost perspective and you feel comfortable that you can support a heterogeneous environment, there is absolutely nothing wrong with that.
Did you find this tip helpful? Send us an email and let us know what tips you want to see.
This was first published in October 2006