A Solid Methodology is the Margin of Victory

Designing complex systems is difficult. It seems to me that the difficulty has increased at a rate that is more exponential than linear with design complexity. Some version of Moore’s law for design is at work. The challenges for advanced designs just seem to keep mounting. A new item that is top of mind for many executives is the talent shortage. Even if you had the budget, you just can’t find enough qualified engineers. It’s a vexing problem with many proposed solutions. At the root of many solutions is a simple, elegant focus – a solid methodology is the margin of victory.

Semifore recently posted a press release – CommScope Selects Semifore to Accelerate Product Development and Improve Quality of Results. This sounded like one of those margin of victory strategies. You can read the press release at the link provided. It’s also on Semifore’s SemiWiki page here, with lots of other good information about the company.  In this post, I wanted to explore the story beyond what was in the press release. Does Semifore deliver a margin of victory strategy? Here is what I found.

Who is CommScope?

First, a bit about the company mentioned in the press release. According to the press release:

CommScope is a global leader in communications technology to create the world’s most advanced networks. Across the globe, the company is redefining connectivity, solving today’s communications challenges, and driving the innovation that will meet the needs of what’s next.

Networks and communications are indeed a critical enabler for a lot of what’s next. The company provides both hardware and software to enable broadband, enterprise and wireless networks. The product line, support strategy and overall focus on the big picture are quite impressive. After some digging, I wanted to learn more about how ambitious companies found the margin of victory in their markets, and how Semifore fit.

For that, I spoke with Semifore’s co-founder and CEO, Rich Weber. Rich has a storied career designing complex systems at major corporations such as Data General, Silicon Graphics and Cisco.

Exploring the Margin of Victory with Rich Weber

My conversation with Rich explored the bigger picture of what it takes to successfully build a complex system. The specifics for CommScope are outlined in the press release. I was after the bigger story.

Our discussion began with several stories of confusion and chaos. They all tied back to a lack of a solid methodology and timely communication. The stories Rich related were primarily focused on the hardware/software interface, or HSI for complex systems. The HSI is composed of many registers that implement the complex protocols allowing software to control the hardware required to implement the software algorithms.

Small errors in this part of the system can cause big problems. Beyond the visible failure mode of “it didn’t do what we expected” there is the invisible failure mode of “the new feature can’t be implemented because the HSI doesn’t seem to support it”. It turns out that errors in the HSI implementation aren’t the whole story.  Rich described several scenarios where a well-meaning engineer made a seemingly harmless change to a bit in the HSI on a Friday evening. That caused regressions to fail over the weekend and further cause the data that was driving the various teams working on the project to be invalid. Teams like documentation, testing and design were all victims. Monday morning was met with chaos.  The HSI is quite fragile – small tweaks can result in big headaches.

We also talked about specifying the HSI and the role standards play. It turns out Rich has been directly involved in developing many of these standards through organizations like Accellera. Standards such as SystemRDL and IP-XACT. What I discovered in my discussions was that these standards are certainly useful, but they all fall short of delivering a complete, robust solution that can express all aspects of the HSI. Solving this problem allows an executable specification of the HSI to be developed, something of a holy grail item for many design teams.

With this backdrop, I began to see a pattern. Semifore offers a way to fully specify the HSI with a robust language (the elusive executable spec). They also offer a way to create a correct-by-construction HSI with accurate, up-to-date representations for all team members. No more chaos.

It seems like CommScope figured out that a solid methodology is the margin of victory. If you dig into what Semifore is up to, you may find your margin of victory as well.