Choosing A Linux Flavor For Your Datacenter
There are hundreds of flavors of Linux, each with their own focus and opinion on how the soup of open source tools should be assembled and maintained into a workable operating system. Choosing one for your desktop can be fun, as you get to try different distributions out without a whole lot of investment. However, when choosing a flavor for the datacenter or cloud hosted environment, you may find yourself stuck with your decision for a long time.
The first issue that needs to be addressed is the matter of support. If you are part of a corporation that requires their IT systems to carry a support contract, your choices are drastically limited. The three biggest Linux vendors are Novell with Suse Linux Enterprise Server (SLES), Red Hat with Red Hat Enterprise Linux (RHEL), and Canonical with Ubuntu Server.
I’ve run all three in our datacenter at one time or another. I’ve generally been very happy with SLES; it has proven itself to have excellent resilience, and the YaST system is fairly easy to manage. Novell also has licensing agreements with Microsoft, with may be valuable to management.
I have had a love-hate relationship with RHEL over the years. Making the switch from SLES to RHEL was more difficult than I thought it should be. Almost everything was in a different place, and when we made the switch Red Hat was charging a large fee for having a local patch repository, something we had for free with SLES 8 and 9. However, over the past few years I’ve gotten used to the RHEL way of doing things, even if we have abandoned RHEL for it’s license-free clone.
Ubuntu is still clearly one of the biggest names in Linux, and appears to be the platform of choice for new projects. Where RHEL is focused entirely on the enterprise and ensuring their flavor is well supported for up to a decade, Ubuntu appeals more to those looking for cutting edge tech at the expense of long term platform stability.
SLES and RHEL require a licensing fee to download patches and get phone support. Ubuntu offers support through Canonical, but can otherwise be used completely free. My experience is that we rarely, if ever, call for support. I recall once calling Red Hat support just because we could, and that may have been the last time. As a systems administrator, you are the support.
If you can talk your stakeholders into giving up support on the operating system, you can expand your candidates to Debian, OpenSUSE, or, my personal choice, CentOS. Debian has a fantastic reputation as a server operating system, but I have found that tracking RHEL releases with CentOS gives us a stable, reliable system that we can make long-term plans for without the risk of falling out of date on security patches.
My experience with OpenSUSE dates back a few years, to the first time we decided to run Linux on IBM Power servers. I’ve had an old version of OpenSUSE running in an LPAR internally for years, and the system has never crashed, never experienced a kernel panic, and has very respectable uptime statistics. I haven’t tried the newer versions, but if they have kept the same stability and performance from previous years, OpenSUSE would require a serious look.
When you get that call at 2AM for an obscure problem that you never anticipated, you will be glad you chose a widely deployed, well supported, and constantly developed server OS. The odds are that even if you did not anticipate the problem, someone else has already been through it, and posted a solution. It might be interesting or intellectually stimulating, to run a smaller Linux distribution in your data center, but when the chips are down the only thing that really matters is keeping your application up and running, and you don’t want your operating system to get in the way.