Course embedded linux

Linux is just another operaTing sYsTem, 50 what's The Big Deal? Reduced cost and fasTer Time-To-markeT-THese are some oF THe many Benests That are making LInux THe FasTest GrowInq operaTing system In The WorLD flOW now anD, many preDICT, For Years TO come.

Why would IBM, MSC.Software, and a whole bunch of other hardware and software suppliers get "pretty excited" about the Linux operating system (OS)? Try low cost and reliability, multi-platform support, world-wide popularity (and potential for high-volume sales), and Open Source accelerating innovation. For these same reasons, you might want to not only get excited about Linux, but consider it in your information technology (IT) strategy.

Linux is a free, Unix-like OS developed by Linus Torvalds while he was a student at the University of Helsinki in 1991. (In fact, "Linux" is a fusion of "Linus" and "Unix.") Software developers worldwide have been refining Linux ever since. Linux is now developed under the GNU General Public License and its source code is freely available to everyone, although its assorted distributions might not be free from software companies and other developers.

WHY So popular?

Linux is not Unix; it was written from scratch. Both follow POSIX standards (Portable Operating System Interface for Unix). And like Unix, Linux is a more technology-oriented OS; that is, it doesn't quite have the ease-of-use slickness of Microsoft's Windows and Apple Macintosh OSs. However, those OSs don't necessarily have Linux's true multitasking, virtual memory, shared libraries, TCP/IP networking, multi-user capability, and load balancing features. Nor do they, like Linux, intrinsically support computing environments that can scale horizontally and vertically-from embedded devices, desktops and Web servers, all the way up to cluster computing environments.

There are several reasons for the worldwide popularity of Linux during the past two years, according to Deepak M. Advani, vice president of Linux Strategy and Market Development for IBM Corporation (Somers, NY). First, the Linux total cost of ownership is relatively small. You don't have to pay a license or royalty for the OS. Better, it can run on inexpensive, non-proprietary, commodity hardware (read "Intel" and "AMD" desktop computers). User companies that lease their computers can trade in these machines every six months or so for a faster machine-and with Linux. "Computer users can actually save money by making a decision to install Linux tomorrow," says Michael Tiemann, chief technology officer (CTO) of Red Hat, Inc. (Durham, NC).

From a hardware perspective, Linux is very efficient; it doesn't require a lot of CPU resources. "The things people have been able to do with old and refurbished desktop computers with Red Hat Linux and the other utilities that come with it is just mind boggling," continues Tiemann.

Linux is reliable. Once you deploy a Linux-based solution, it just runs. No reboots every week or so to flush out memory, eliminate DLL corruptions, or reset pointers.

Linux preserves users' investments in skills and applications because it can be installed just about everywhere. Linux is the only OS that runs on all "Intel Inside" and AMD platforms; PowerPC, RISC, and Alpha processors; as well as on mainframes and IBM AS/400s. "Users like that; they're not locked to any given vendor," says Advani.

There's also a "macro-economic trend" to consider, Advani is quick to point out. "Universities are deploying Linux--pervasively and aggressively. Many of today's students will not only run the IT systems in the future, but they will make many of the IT decisions." Tiemann agrees, relating a conversation he had with the CTO of a Tier-1 automaker. "I asked him if he was running Red Hat at home, and he said, 'Of course!'"

As open-source software, the Linux source code is freely available for examination, modification, and redistribution. Nobody really owns that source code (although the name "Linux" is a registered trademark of Linus Torvalds). Anyone can use it; anyone can change it, fix bugs, add new features. The only caveat is that software changes are relinquished for everybody to share and use.

"One of the interesting things about Linux and the Open Source development model is that programmers are reluctant to write poor implementations that will reflect poorly on their abilities," says Tiemann. To this, Advani adds that IBM sees Open Source accelerating innovation. "It's a mechanism to bring together some of the brightest minds in the software industry, collaborating through the power of the Internet, to develop a world-class OS."

A Hardware vendor's perspective

IBM's assumption is that if Linux is this ubiquitous, IBM should have Linux running on all its "e-server platforms." (These are rebranded IBM servers. For instance, the AS/400 is now the iSeries-"I" stands for integrated; mainframes are the zSeries-- "Z" for near-zero downtime.)

Of course, Linux runs on IBM's Intel-based Netfinity series (now xSeries). It also runs on IBM's other platforms, including IBM mainframes, Why would anyone want that? Universities, for instance, using IBM's VM (Virtual Machine) OS, can run tens of thousands of Linux instances on one mainframe. This not only gives the university's students their own dedicated Linux-based computer, it also increases the utilization of the mainframe.

On the industrial side, user companies love the reliability and manageability of their mainframes and the distributed computing model that IBM gives them. But these companies want to run modern, Internet-based applications on those very same mainframes. Linux makes that possible.

A software vendor's perspective

Unlike IBM, MSC.Software is pushing software--computationally intensive software. Specifically, finite element analysis (FEA). MSC.Software's goal is to take big computational jobs--crash analysis using FEA is an excellent example--and "parallelizing" them so they run much, much faster across multiple processors or some form of cluster computing environment. "By taking this really big problem, dividing it into smaller problems, and orchestrating that solution, we can do that job a lot faster," says Jay Clark, director of Marketing and Business Development for MSC.software (Costa Mesa, CA). However, "linking several computers together does not become cost-effective unless you use commodity hardware. Linux lets you use commodity hardware with all the traditional strength of the OS, such as multi-user and multitasking with low overhead and high security, so you can get the job done faster and a lot less expensively."

This is exactly what was found at Dana Corp.'s Spicer Axle Div. (Toledo, OH), the world's leading independent axle and component manufacturer. Running Linux-based MSC. Nastran and MSC. Patran across multiple Intel processors, Spicer engineers found that individual analysis jobs ran two to ten times faster, depending on the physics behind the problem being solved. In general, the engineers found the Intel-/Linux-based cluster computing environment ran approximately three times faster and at half the cost of other simulation environments.

(By the way, herein lies Linux's real threat to Microsoft. Forget "better" software, though that's a hard one to forget. Forget less expensive software; ditto. Ground zero is in Web-based and distributed computing environments, including cluster computing. Microsoft very much wants to own the defacto OS that fuels these environments.)

WHAT ABOUT UNIX?

Unix has traditionally, and technologically, provided high-end CAD/CAM, points out Tiemann. But it does so with a price tag five to ten times what you would pay using Linux running on Intel hardware. So not surprisingly, traditional Unix applications are now showing up on Linux. Happily, migrating Unix applications to Linux is a fairly easy port.

Does this then spell the end of Unix? Hardly. Computing will continue to be a heterogeneous world, says Advani; all platforms will have a key role to play. And, alas, Linux still has some limitations. It is not yet ready to be your high-volume, mission-critical OS, continues Advani. "I look at OSs from a workload perspective. What workload do you want to deploy? Is Linux ready for that workload? For high-end databases requiring 16- or 32-way symmetric multiprocessing (SMP) capability, you need low levels of failure over time. That requires sophisticated workload management and scheduling. Some of those capabilities aren't in Linux today."

LinuX In auTOmoTIve