The Operating System Quandary for Electronic Design

If you asked Microsoft if they believed that there is an OS quandary in the EDA market, their answer would likely be a question. What quandary? Microsoft expects NT to become the tsunami in operating systems overtaking UNIX in all of its various flavors. Will Microsoft prevail with NT in the EDA environment?

I would agree that NT has made great strides as a mainstream operating system for EDA applications such as printed circuit board design and PLD design. NT has captured an impressive share of the mechanical design market. However, it can be argued that most designers of the higher complexity boards, FPGAs, and mechanical design still bank on UNIX as their OS of choice. Are the users of NT for board and FPGA designs happy campers? Are designers using NT for medium to high complexity chip designs satisfied? We will consider addressing these questions in a future column if readers would like this type of data.

From a business perspective of those who port EDA tools, or make their living at offering workstation and server products, they have found that only a small fraction of the 1998 EDA market revenue was from NT based solutions. Without a question, the primary source for EDA tool revenue is still UNIX. Workstation and server revenue in the NT space was primarily from sales for mechanical, printed circuit board and PLD/FPGA design and for some desktop or client installations for ASIC design. As reported by the EDA industry via the EDA Consortium, NT based revenue grew at a higher rate than UNIX revenue, but NT was still left in the dust overall.

If NT hasn't captured a significant portion of EDA related revenue, why are the buyers and users continuing to make their statement with purchase orders for UNIX based applications and hardware? It's the high end of EDA products that users as yet do not trust to NT in the electronic design space. High end products are applied to high risk, high visibility, and high leverage design projects. It's also for reasons of the difficulty and sizable resources to port legacy code from UNIX to NT. It's from knowing the history of UNIX's stability compared with observations of WINDOWS systems and sometimes NT systems that causes doubt.

Will Linux become the white night for EDA users choosing Intel based computers? Will we have LINTEL computers? We have a ground swell of interest at this time. In my opinion the industry is yet in the early trial phase of Linux. A few products, particularly Verilog simulation products, are now available from large EDA suppliers to run on Linux. Major workstation and server companies have announced support including Compaq, Dell, Hewlett-Packard, and IBM. Expect additional Linux support annoucements in the June, 1999 Design Automation Conference time frame.

Linux provides for a more familiar and appreciated user interface to past users of UNIX. The NT user interface is considered 'closed', and too restrictive by many design groups of complex chips. But there are some chip design organizations that are satisfied with clients running NT networked to a server farm running UNIX.

The Linux ground swell has Microsoft's attention. The Wall Street Journal has reported that Microsoft has a small team of Linux stalkers. International Data Corporation, an industry research firm, reports that the Linux server unit sales rose to 17% last year from 7% in 1997 for the broad commercial industry.

For EDA applications the port from mainstream UNIX systems to Linux appears to be simple except for applications depending heavily upon the Graphics Users Interface (GUI) unique to the flavor of UNIX already being supported. Solutions are being evaluated for native compiled code and code that adapts to parallel processing. It is a major and expensive undertaking for a provider of EDA applications to initiate and maintain support of another operating system. There must be an attractive market opportunity for their products to make such a commitment and compelling reasons to internal EDA developers to undergo this expense.

The desire of chip design engineers to run 'Office' applications has been one reason to favor NT desktops most often networked to UNIX server compute farms. This argument is about to disappear. For the engineer not wanting to forego office applications, there is a solution to run the Microsoft 'Office' applications suite on Linux.

If someone were to quantify the cost of the current OS divergence to EDA developers of proprietary and commercial EDA applications, the support cost for divergent operating systems, the design inefficiency caused by this divergence, the expense to the electronics industry would be shocking. If Linux can meet the demands for complex chip and system design, and it will help us converge on a common, or nearly common, OS to reduce this cost, financial benefits will result for buyers of EDA systems.

We have already mentioned the GUI issue. Linus Torvalds, the father of Linux, keeps reasonable control of the Linux kernel. Except for the drivers, perhaps a least common denominator will develop for the Linux suite. An independent company such as Red Hat may have an opportunity to bring this under control. Oh, control is a bad word to the Linux community.

My analysis is that for Linux to become the choice as a dominant operating system for chip design applications, the industry will first need to address a well defined, consistent and managed software set. And the supplier of a Linux suite that includes the kernel, utilities, drivers and GUI support must have a proven support infrastructure and service plan.

Are we in a quandary at this time when it comes to choosing the best strategy for operating systems to be used for chip design? Without a doubt, Yes.

Copyright Summit Innovation Inc, 1999

Workstations and Design Tools Archive

Copyright © Chipcenter 1999

		Click here to get your listing up.