Should EDA Heads Be In The Cloud?

Consider the following two comments about cloud computing and electronic design automation:

• “Over time everybody will move to the cloud in EDA at least in some extent.”—Raik Brinkmann, CEO of OneSpin Solutions.
• “We put a substantial effort into that, and of all the things we’ve done in the last 25 years this is probably the single one where the result is essentially zero. I don’t mean some money, I mean zero.”—Synopsys CEO Aart de Geus on cloud efforts.

Synopsys’ efforts might be reflected in De Geus’ quote above; Cadence has edged toward to the cloud a bit, and CEO Lip-bu Tan said earlier this year “we’re keeping a very close eye on it. We’re ready, we just have to wait and get customer requests for it.” (Both those execs were on an amazing panel earlier this year about the future of EDA).

Somewhere between the poles of those experiences lies the future of EDA, electronics design methodologies and cloud computing, and the $64,000 question is what will be the tipping point?

Consider:

• 80% of new commercial enterprise applications will be deployed on cloud platforms, according to the market research firm IDC.
• The public cloud services market in 2012 was forecast to grow 19.6% to $109 billion worldwide, according to Gartner.
• Amazon Web services exceeded $1 billion in cloud services last year, according to IDC.

Cloud Computing Opportunity
We know that, at the very least, certain types of EDA tools lend themselves to massive multiprocessing clusters in the cloud; it’s a vision of step-function productivity gains that puts stars in the eyes of engineering managers.

Notes Dave Kelf, marketing director at OneSpin: “On the Amazon cloud system, you can have any number of computers, so any software that can run in parallel you can basically scale it up to maximum scalability. This is perfect for computationally intensive algorithms.”

We also know that a time-based usage model makes sense for some types of EDA tools and tasks.

And we know the cloud works for other industries.

We also know what’s not working: private clouds. In theory they can work for customers on a computation-scaling basis, but they’re not cheap.

“The cloud just by itself offers possibilities on the computation side that you just don’t have,” Brinkmann said in an interview. “Companies can’t afford to have big compute farms in the cloud and add resources to what you’re doing.”

Private clouds also get more expensive over time: As the node sizes drop, the hardware has to be refreshed every 18 to 24 months if you want to keep the same footprint in the data centers.

Perfect Storm Clouds?
So an EDA move to the public cloud sounds inevitable. In fact, you could argue that in 2013 there’s a confluence of two powerful factors that may hasten the tipping point.

• First, CFOs are still reeling from the Great Recession, and cost cutting and productivity improvements are always top of mind. (It doesn’t hurt that while Amazon’s cloud services revenue may have topped $1 billion, the company – amazingly – has cut its cloud service prices 28 time since the service’s inception).
• Second, parallel computing benefits in the cloud take on increased significance as time-to-market pressures build on engineering teams grappling with ever more complex designs.

But there are counter forces. Cultural change, even in a fast-paced innovation world like electronics design, comes slowly.

“The overriding issue is, ‘Well that’s not the way we do things,'” said Raul Camposano, CEO of Nimbic, which offers cloud-based simulation solutions for signal integrity, power analysis and EMI.

Rethinking Software Development
Another force is that the evolution will also require a reconsideration of how EDA companies build software. “You need to be able to simplify support to the point that you can applify,” Camposano said.

That may be easier said than done given the nature of the complex EDA tools and robust support infrastructure customers require. And if you think about it, EDA has spent the past decades evolving from “apps” (point tools) to integrated platforms and flows.

As an “applify” example, he said, “In an board design system, you click on a net and an app returns the parasitics (RLC) and perhaps the coupling with neighboring nets. It’s an ‘app’ in the sense that it only does that one thing, and it is trivial to use.”

While the vision and the potential are there, use cases need to be considered and that means that cloud computing isn’t for everyone. One of the tipping points for cloud computing, Camposano suggests, is the 20 percent rule. If you use your computers more than 20 percent of the time—say for RTL simulation—it’s better to host your own infrastructure at this point. And large EDA customers probably fall into that category (at least for now).

Security cloud looms
The biggest concern remains security: Will my IP get ripped off in the cloud? Then again those concerns may be overcome relatively quickly as they have existed since the dawn of the design automation industry, whether they applied to counterfeiting software CDs or hacking software downloads.

There’s also the issue of product verification across clouds. Just as we have to test across different operating systems, “we may have to test across different clouds,”

The true believers see a future in which those concerns are easily met, and Brinkmann serves as one of their eloquent spokesmen.

“The industry will change completely and people will forget about their concerns,” he said. “I say this because the cloud is an extension of your computer. EDA is
special but many other industries are special as well.”

He adds a compelling thought: “I believe other people will see other applications that you can only do in the cloud.”

Are these guys’ heads in the cloud?

Absolutely. Positively.