Research Computing

Spending graduate school in Silicon Valley teaches you a great deal about computers and technology. You learn how to build your own computer from dirt-cheap components. You meet many people who go on to work for Oracle, Cisco, Intel, Yahoo, and a plethora of startups too diverse to name. And you learn about Linux and how fast, stable, and secure an operating system can be.

In contrast to the home PC market, the server and research computing markets are already very well suited for Linux adoption: Researchers are already familiar with the basics of Unix, scientific applications almost always come in Linux versions as well as Windows versions, and the enhanced security, speed, and stability of Linux are much more valuable to the researcher than they are to the general population. What follows is a more detailed analysis of the advantages of Linux for serious research computing users.

Security

This almost goes without saying. Bugs are inevitable in any code, and the bigger, more bloated, and more legacy-oriented the code, the more likely that bugs will appear. The recent outbreaks of Code Red (8/01), Slammer (1/03), Sobig (8/03), Blaster (8/03), Bagle (1/04), MyDoom (1/04), Sasser (4/04) and the recent IE keystroke logger (6/04) now seem to be the rule rather than the exception for Windows-based systems. If you use Windows, you were probably personally affected by two or three of these viruses. If you used Linux, you wouldn't have been affected by any of them.

Speed

To those of us who use Linux and Windows both, it is amazing to see how slow Windows is. The funny thing is that most people who use Windows are under the misimpression that it is their computer that is slow, rather than the operating system. The reference in the Intel ad at right to “frustrating lags” is a testament to just how widespread this misconception is. (Of course, the Intel ad is just trying to capitalize on this misconception to sell users more hardware.)

Like everyone else who uses Windows, I'm very familiar with the “frustrating lags” referred to in the ad—for example, when I'm dialing in on my laptop and mistakenly download a large attachment in Lotus Notes, my computer is competely hung until the attachment finishes downloading. Similarly, if I'm at work and there's a network problem, again Windows is slowed down to the point where it becomes very difficult to switch between applications or even to work within a single application. The basic reasons for these hangs (and we all experience numerous momentary hangs in Windows every day) is that the Windows operating system itself (even NT/2000/XP) is not truly multi-threaded, whereas Linux is (note that “multi-threading” is a software feature that is unrelated to the “hyper-threading” processor technology referred to in the ad). Essentially what this means is that, even though multiple applications can run simultaneously in NT/2000/XP, Windows itself can only “talk” to one application at a time, and if for some reason that application is slow to respond (or even worse, not responding at all), Windows is stuck and is not able to talk to any other applications until it gets a response, making it very difficult for the user to switch to another application.

In Linux, there are no hangs of any kind, and switching back and forth between applications is always instantaneous and seamless. It's also true that disk and memory access is faster in Linux, although this difference is generally not noticeable to typical users.

Stability

Linux doesn't crash. This used to be a very severe problem for Windows, although my experience with Windows NT/2000/XP has generally been pretty good (although I have coworkers who have had problems). For doing research, this can be extremely important: obviously, if you need to let a Matlab or Mathematica procedure run for a few days, you need to be sure that the application, the operating system, and the interface between the two won't crash during that time. While an application (such as Matlab) can crash in Linux just as it can in Windows, at least you can rest assured that the operating system itself will be stable (and that a problem with some other application, such as Internet Explorer or Lotus Notes, won't crash the whole system and force you to reboot).

Competition

Finally, the strength of the U.S. economy is built on competition: letting the best ideas and best products prevail. The fact that Microsoft enjoys a virtual monopoly over personal computer operating systems leads to enormous incentives for the company to absorb, stifle, or corrupt anything which could be a possible competitor down the road. In response to these incentives, Microsoft for several years prohibited computer manufacturers from preloading Netscape on their computers, and Microsoft continues to corrupt HTML and Java standards set by the W3 consortium and Sun so that documents created by MS products (Word, FrontPage, etc.), or optimized by other developers for how they look in IE, will then function poorly or not at all in competitors’ products. This both maintains IE's dominance in the short run and, more importantly, prevents the possible erosion of the Windows monopoly down the road if a uniform (uncorrupted) Java or browser standard were to emerge and greatly reduce the need for anything other than a bare-bones operating system for many users. These tactics have given IE a dominant share of the browser market, and now that dominance has been achieved, Microsoft's innovation in the product category has ceased. The U.S. economy loses as a result. (Note that next-generation web browsers such as Mozilla and Opera have continued to innovate and incorporate new features such as popup-ad-blocking, enhanced security, Google search bars, viewing multiple web pages simultaneously, and support for “mouse gesture” shortcuts). Web browsers aside, you'll also notice that Microsoft has yet to release a Linux version of MS Office, despite the fact that Linux users outnumber Macintosh users (see also this article). Why would Microsoft ever release such a product given that it would only hasten the demise of the Windows monopoly? (Macintosh is not regarded as a serious threat to Windows by anyone.) Competition in the U.S. software industry is being hindered by these practices, which ultimately stem from Microsoft's market power in the desktop OS market and its incentives to maintain that market power. By adopting Linux, you will be moving the U.S. economy one step closer to a competitive OS market, with spillovers into the entire computer software, hardware, and even technology industries.

Moving Away from Microsoft

The above advantages have not been lost on many corporations and foreign governments. Countries that are poorer than the U.S. and less beholden to the software giant are already beginning to migrate over to Linux and other open source applications. Some noteworthy examples include:

China
Korea
India
Asia and Asia
Brazil and an update
Israel
Spain
Europe	This is a long and in-depth article. See particularly the companion piece at the bottom
Developing Countries	BBC article talking about multinationals in general, some discussion of developing countries in particular.

Learning More About Linux

Linux is Unix for your PC. As I mentioned above, for most research users, a switch to a more Unix-oriented desktop environment is easier than for the general public. There are a variety of sites devoted to Linux and switching from Windows to Linux (you do not need to switch completely, you can continue to run Windows and Linux on the same machine). Some of the most useful are the following:

Linux Online	Provides a good introduction to Linux.
The Linux Documentation Project	Huge repository of information. They maintain the Linux FAQ.
Live CD Trial	If you would like to give Linux a try without actually installing it on your computer, you can run it off of what is called a “Live CD”. The link to Tuxs.org at left explains how. (Note that this will probably give you a more sluggish Linux experience than if you installed it on your hard drive.)
Novell/SuSe	Novell offers and maintains the SuSe Linux distribution, which is the most popular distribution in Europe and the second-most-popular in the U.S. Their Novell Linux Desktop is an add-on to SuSe that provides a very user-friendly interface. You can also download a free evaluation copy, much like the Live CDs above. The one disadvantage of SuSe/Novell is that it is somewhat bloated and thus slower than leaner distributions such as Debian.
Debian	If I were buying a new computer for work, this is probably the distribution I would use. Very customizable and more “lean and mean” than the other leading distributions, but probably not as good a choice for novices.
Red Hat	This is currently the most popular Linux distribution in the U.S. I actually don't like it that much because it's extremely bloated (i.e., contains an enormous number of files and features), which tends to make it slow to boot up, more sluggish than other distributions in general, and harder to customize.
Running Windows Applications in Linux	A brief article on running Windows applications in Linux. Of course, you should be trying to switch over to open source applications (such as Mozilla and OpenOffice) or native Linux versions of applications (such as Matlab and Mathematica) as much as possible and run a Windows app only if absolutely necessary.
VMWare	VMWare lets you run not just Windows applications, but Windows itself simultaneoulsy with Linux. It gives you the look and feel of having two PCs (one Linux and one Windows) running simultaneously and hooked up to the same monitor and keyboard.

Other Open Source Projects

Linux is but one of many prominent open-source initiatives. You should seriously consider the following products if you have not done so already:

Mozilla	They maintain the Mozilla Firefox browser, Thunderbird mail client, and Mozilla HTML Composer applications. Technology columnists across the country have almost uniformly panned the security holes in IE and recommended switching to Firefox: here's one such article.
OpenOffice	Word processing technology is simply not that complicated. As more and more governments and schools switch to OpenOffice, this will probably become the future word processing standard. As an added bonus, it also contains spreadsheet and presentation software.
GCC	The GNU Compiler Collection. This is perhaps the original open source project of any appreciable complexity. GCC is now an extremely good compiler for C, C++, Fortran, Java, and a number of other languages.
Octave	An open source Matlab clone.
The GNU Project	Umbrella organization for a wide variety of open source projects, primarily related to Unix systems but also some others. GNU Unix utilites are very often superior to their commercial counterparts.

TeX and LaTeX

TeX was created by Donald Knuth at Stanford between 1979 and 1981. It is a word processor/desktop publishing system for mathematical and scientific writing that for the first time made typesetting complicated mathematical expressions feasible to a researcher with a personal computer. In 1983, Leslie Lamport made a large number of his own TeX macros and definitions available to the general public as LaTeX, with the aim of making it easier for new users to get up and running with TeX typesetting. The commercially available package Scientific Word simplifies TeX even further by providing more macros and a graphical user interface to LaTeX.

Most research economists are familiar with LaTeX or Scientific Word. I switched from LaTeX to plain TeX several years ago because I became frustrated with a lot of undocumented behavior in LaTeX that would arbitrarily change how my documents looked and made it very difficult to do simple things such as single-space footnotes or quotations within a double-spaced document. Like a manual vs. automatic transmission automobile, plain TeX gives you a great deal more control than LaTeX so long as you are willing to spend a little more time learning up front. (The same can be said for switching from Scientific Word to LaTeX.) The definitive guide to writing in plain TeX is Donald Knuth's The TeXBook; it is a brilliant book providing absolutely complete documentation for every aspect of TeX. The TeX Users Group provides a good introduction to TeX and LaTeX and a very thorough set of references to information available online. Both TeX and LaTeX are freely available to the public.

TeX presentations: As a researcher, one of the most difficult and annoying aspects of PowerPoint is typesetting equations. Entering the equations is time-consuming and awkward, and they often don't look nearly as good as what you would get from TeX or its derivatives. The open source software package Prosper solves this problem by combining the LaTeX typesetting language with presentation software, allowing you to include graphics and animation in your presentation while at the same time rendering your equations just as easily and beautifully as in TeX and LaTeX.