It's applications, not OS, in driver's seat

It's applications, not OS, in driver's seat
There's high drama in portraying the computing industry at war over operating systems. "Unix vs. Windows NT" makes far more compelling headlines than "Engineers need tools that make sense and work." Enhancements to Unix are automatically interpreted as maneuvers to defend against the incursions of the Windows NT "threat," or even an eventual Windows NT victory over Unix. Let's face it, it's simply more interesting to cast the evolution of software as a raging debate instead of what it is: a development of tools to accomplish specific jobs.

In the EDA market, for example, evidence suggests that the "inevitable victory" of Windows NT is not imminent. The reason is simple: Electronic-design engineers today commonly depend on Unix because it reliably supports complex simulation tasks that can last for days. Unix has the capacity to handle the size and complexity of data in compute-intensive applications like formal synthesis. Many agree that Windows NT will develop the robustness, seasoning and maturity required of an operating system for such applications. But for now, Unix and Windows NT coexist in this market with mutual respect, each doing the jobs it was designed to do.

What is actually more interesting than dramatizing the OS wars is looking at real-world applications: Who's building them? And for what operating systems? In the EDA market, it takes years of experience working with software engineers at independent software vendor (ISV) companies to fine-tune application performance. We've learned that it is not reasonable to throw technology "over the wall" to our software partners and expect them to implement it. What does work is diving in and collaborating to deliver the applications and technology that customers want.

HP's Unix operating system on PA-RISC supports a very large portfolio of EDA applications. And with the 64-bit capabilities in HP-UX 11, ISVs porting to this operating system are delivering advanced benefits to their customers. Initially developed to exploit the large-memory high-end servers from HP's Convex acquisition (with 16 GB of RAM and 16 processors), HP-UX 11 currently delivers 64-bit capability on workstations while retaining full 32-bit binary compatibility. It accelerates traditional 32-bit applications by implementing performance-optimized page sizing, a technology that is instrumental in accelerating performance in some applications by 65 percent without changing the hardware.

Binary compatibility and performance enhancements are major advantages that HP offers its ISVs and customers. In addition, applications running on HP-UX 11 will migrate seamlessly to EPIC, the next-generation computing technology that underlies Intel's Merced microprocessor.

Binary compatibility is not enough, however. ISVs know that, and they look to system providers to work closely with them. That's how they know they'll achieve next-generation performance and technology improvements. The important work-recognizing interdependencies and working in close collaboration-continues to be the best recipe for success.

So where does that leave the Unix vs. NT debate? EPIC is a whole new computing paradigm engineered by HP and Intel that will host both Windows NT and HP-UX. Maybe with this development it will become clear that application requirements and real costs-not operating systems and perceived costs-will drive customer choice. In this brave new world of computing, the winners will be customers who exercise their options in the most efficient ways and the companies that help them do that.

— Diane Sullivan is worldwide EDA manager for Hewlett-Packard Co. (Palo Alto, Calif.).