Home » Technology » Scientific Computing » Linux - Coupling Flexibility and Cost-Control

Linux - Coupling Flexibility and Cost-Control

© Adam Hughes

In this series of articles seeking to define the optimum operating system for use in scientific computing, we've looked at Windows® and Unix®, discovering benefits and pitfalls in the use of each. The opinion here, probably not so thinly-veiled, is that for the scientist looking to get the most from his computer time, Unix® offers tremendous flexibility in developing and using powerful applications.

That being said, one issue which hasn't really been addressed here is that of cost v. performance. Which kind of system is going to cost more, and which will perform better?

To begin to answer these questions, it's instructive to make some distinctions, and with them, an assumption or two. First, having determined what the scientist most needs from his operating system, we can say that, performance-wise, he'll choose Unix®. Proceeding on this assumption, we can then compare the costs of both the operating systems and the platforms, or physical machines, where those operating systems run. In general, the cost of software is nearly negligible for most users when compared to the cost of the needed hardware. Assuming this is true for operating systems, we're left comparing the performance and expense of the machines themselves.

Up until a few years ago, probably before the introduction of the Pentium® processors, Unix®-based workstations were the real desktop powerhouses of the scientific computing world. It may sound a little laughable now, but many computational science groups were built around several 100-MHz machines with 64, 96, or 128MB of memory. While they were faster than anything else widely available at the time, these computers often taxed research budgets to the tune of tens of thousands of dollars each. My, how times have changed!

These days, of course, personal computers typically have processors of 350 MHz or faster, with standard memory allocations approaching the limits that the Unix® workstations knew in the recent past. What's more, PC memory is cheaper and easier to install than in the past, adding even more powerful capabilities to the ubiquitous personal computer. And, of course, the price of these PC's, even if you want every flashy gadget known to man, won't approach the ransoms paid for the Unix® boxes of the not-so-distant past.

Given the above discussion, developing the optimal scientific computing workstation is a trivial matter, right? You grab a cheap, fast PC, choose Unix® as your operating system, and you're all set! Well, the first part of that recipe sounds OK, but the problem is, PC's aren't made to run Unix®. So, if you're "stuck" using Windows®, what choice should you make? It may be that the answer to that question doesn't really matter, thanks to Linux.

The copyright of the article Linux - Coupling Flexibility and Cost-Control in Scientific Computing is owned by Adam Hughes. Permission to republish Linux - Coupling Flexibility and Cost-Control in print or online must be granted by the author in writing.