I was reading through some new postings on a local LUG mailing list here in Louisiana, and saw a note from Chris J. about Microsoft’s Windows vs. Red Hat page. Chris brought up some good points, and I wanted to respond to them.
CJ: Now, let’s rip this FUD apart. First, Microsoft acts as if RedHat is the only option that enterprises would ever go with, and they say that while RedHat itself is cheap, it’s $2500 a year for support. Okay, that’s support, Microsoft. Why don’t we compare apples to apples, and point out that Microsoft’s support is somewhere around $700 per incident? To me, $2500 a year is FAR cheaper
The cost of Red Hat. Okay, in all seriousness, Red Hat is a little expensive on the front end. We often deploy it for clients, and the base cost of RHEL is a little high, especially since it’s a subscription model (you pay it every year) and not a one-time purchase (as with Windows Server). That said, Windows CAL licensing can add up very quickly. So, personally, I would have argued the point based on CALs. But as far as upfront purchase-the-shrinkwrap costs, Windows tends to be cheaper than Red Hat. As far as support costs, well, the RHEL subscription does get you support, but it’s not like your local ABC IT company. And most people don’t call Microsoft OR Red Hat for support.
CJ: I also find a lot of issue with the fact that Microsoft claims that every distro of Linux is so different that migrating from, say RedHat to SuSE is very difficult, if not impossible. One of the key strengths of any UNIX architecture is the portability of files. The file structure is based on an open standard, and you could very easily take files from something like Turbolinux, and easily bring it back up on any other distro of Linux, or perhaps BSD, Solaris, OSX, HPUX, etc… Linux admins tend to keep the data files on seperate drives/partitions from the OS, so you could simply install another OS on a new hard drive, and mount the old data partitions under that OS, and continue right where you left off. If you need something like a database, it’s not hard to dump SQL to a file and reimport it on the new server. And the configuration files are generally flat text files, so how is your data somehow married to the OS/distro that it originated on?
File System Differences. To me, this is a valid point actually. The directory structure across UNIX systems, or even across Linux distros, may be technically something of a standard, but in reality it’s not. Even within the Linux eco-system, it can be hard to remember what is where. Are installs in /opt/ or /usr/local/? How are my rc files organized? Where are my network configuration files? The Linux Standard Base (LSB) group is working hard to address this, but the cold hard reality is that it’s in fact a pain if you are managing more than just a few Linux servers.
Migrating SQL databases. Good point. That is pretty easy (thanks SQL). It’s also very easy to copy a MS-SQL database from one server to another.
CJ: Also, they make the claim that Windows 2003 has fewer published vulnerabilities than Linux. We all know that more bugs will be FOUND in Linux, and they will of course be squashed rapidly. But, due to Windows’ closed nature, how many bugs actually EXIST but have yet to be FOUND?
Vulnerabilities. There has been a bit of a fuss these days about vulnerability counts in Linux. The core of Linux, i.e., the OS proper, is stable and generally secure. It’s rather rare to see a published vulnerability for the kernel or any of the base operating system programs. However, most Linux distros do commit the cardinal sin of installing everything and the kitchen sink, and it’s an entirely valid argument to say that a vulnerability in an installed-by-default application is a point against Linux. This is very similar to how people group vulnerabilities in IIS and Exchange with “Windows”. Tit-for-tat. That or we need to all step back and stop grouping vulnerabilities in this way.
CJ: The only valid argument that Microsoft brings up in this article is about the management interfaces. They hands down win in that department, but that’s why you hear of UNIX guys working at places like NASA, making $200,000 a year. UNIX OS’s are definitely not easier, and you do have to know what you’re doing to accomplish the same thing that you can do in Windows with a mouse click. So what? It is what it is. I also love how Microsoft neglects to mention the fact that Windows Server 2008 is playing catchup with the UNIX world by adding a new feature called Windows Server 2008 Core. The core mode basically turns Windows Server into a GUI-less command-line-based server OS. That way, it can run faster, without the bloat and massive overhead associated with a GUI. Sound like any OS you’ve ever used? Oh, that’s right…UNIX/Linux/etc… And of course, once you are using Windows Server 2008 in core mode, you suddenly lose that one advantage that Windows has: its GUI based management interfaces. Those are some great arguments, Microsoft.
GUI. Actually, I tend to strongly disagree here about the focus on the Windows GUI. First, I think that most Linux servers, especially those used in large, commercial deployments, have pretty good GUI management tools. Second, whether in Windows or Linux/UNIX environments, if you have more than several servers to manage you usually manage a lot of it via scripting and automatic deployments (again, this applies to both Windows and Linux/UNIX). That said, you are right that Windows is pushing a more “scriptable” environment (e.g., with WMI, PowerShell, etc.), although even back to NT4, there was the ability to script a lot of tasks if you could live with the pain of using Windows shell scripting and/or of WSH.