Cloud computing is in the middle of one of the biggest hype cycles since the dot com days, and there has even been a recent flurry of interest in cloud computing for EDA. Is this also hype, or is there really something of compelling value that cloud can offer for electronic design?
First, we should distinguish between “EDA” and electronic design. Rather than viewing the situation from a narrow EDA industry perspective, cloud computing should be approached from the standpoint of how electronic design customers view, forecast and prioritize their technology and infrastructure needs, and how the electronic design ecosystem can step up to match these needs.
Electronic design customers face tasks that are essentially intractable—the computational complexity grows exponentially as design complexity grows. The electronic design ecosystem—primarily EDA tool vendors and compute hardware vendors—has delivered innovation and performance that, in combination, managed to support this explosion of compute resource demand over the past two decades. However, electronic design customers are being forced to spend even more precious CAPEX dollars to satisfy the never-ending demand for compute infrastructure. This situation is exacerbated by the rapidly growing complexity of managing this infrastructure. Possibly worse than the huge capital investment, electronic design companies must divert focus away from their core business imperatives in order to run data centers.
What cloud computing offers the electronic design community is an opportunity to fundamentally break this cycle. Cloud computing offers the promise of instant access to all of the compute resources necessary to match the highest demand peaks, and the instant elasticity needed to stop paying for those resources when the peaks are over. Moreover, as this benefit is delivered as a managed service, customers can refocus on their core competencies, while focusing less on doubling data center capacity every two years.
Of course, there must be a starting point use case, and the commonly acknowledged best use case for cloud computing in electronic design is functional verification. Functional verification is the top consumer of compute resources at most electronic design customers. Because functional verification regression testing entails running many thousands to millions of small independent jobs, the limitations of Amdahl’s law don’t really apply, and additional parallelization can be matched against ever-increasing regression test size almost without limit. Electronic design customers have occupied the majority of their data centers with verification compute farms for years, and this naturally extrapolates to matching demand surge needs with cloud compute resources. Today, the vast majority of customers have access to less than 100 simulation compute slots per day, and with our new VCS cloud offering, Synopsys can multiply this basic hardware/software resource with on-demand surge capacity. Thus, for example, an 8.5 day verification regression task that would cause a 7-day schedule miss could be brought back on track by purchasing a 5X surge of on-demand VCS cloud resources.
Industry analysts, customers and vendors all spend a great deal of time talking about how to evaluate whether and how any of this can be justified. Determining the ROI of moving to cloud computing is both complex and situational, but not really rocket science. However, because of the newness of cloud computing in this industry, there is also a large emotional component in the electronic design community’s thinking about cloud adoption. At this point, it’s likely the emotional factor will trump any objective assessment. Electronic design companies place extraordinary value on their designs and other associated intellectual property. IP is the lifeblood of our industry, and for now in any objective analysis, no economic benefit can come close to justifying any risk or perceived insecurity that cloud adoption may imply. There has not yet been a leading electronic design company that has reached the “a-ha” moment that, for example, led Netflix* to adopt public cloud wholesale approximately 18 months ago. As Netflix stated in their “Tech” blog, “Our CEO, Reed Hastings, has not only been fully on board with this migration, he is the person who motivated it! His commitment, the commitment of the technology leaders across the company, helped us push through to success when we could have chosen to retreat instead.”
The electronic design ecosystem is actively laying the groundwork for supporting cloud computing. There are intensive investigations, collaboration and testing activities going on, and much, much more on the horizon. In addition to all of this activity, however, it will take a leap of faith at the most senior level at a customer before real adoption commences. It’s incumbent on all of the vendors in the ecosystem not only to provide compelling objective value, but also to collaborate so that everything we do is aligned towards making that leap of faith a simple and comfortable decision.
*Netflix Tech Blog (http://techblog.netflix.com), "5 Lessons We’ve Learned Using AWS", 12/16/2010
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.