Navigating EDA Vendor Cloud Options

Experts at the Table: Semiconductor Engineering sat down to discuss the challenges of cost-dependent cloud decisions, and how to navigate between different EDA vendor clouds options with Philip Steinke, fellow, CAD infrastructure and physical design at AMD; Mahesh Turaga, vice president of business development for cloud at Cadence Design Systems; Richard Ho, vice president hardware engineering at Lightmatter; Craig Johnson, vice president cloud solutions at Siemens Digital Industries Software; and Rob Aitken, fellow at Synopsys. What follows are excerpts of that conversation, which was held in front of a live audience at the Design Automation Conference. Part one of this discussion is here. Part two of this discussion is here.

SE: When it comes to working with tools from different vendors, how does this get worked out in the EDA vendor clouds?

Turaga: The Cadence cloud is an open platform. We allow any third party tools as long as [we all agree]. That’s how we are approaching the platform.

Ho: The asterisk, “As long as they agree,” is the key. We’ve gone to a direct-to-the-cloud provider because of that exact issue. We want to be able to use multiple vendors and so we have to go direct to avoid any issues where a vendor may say, “I don’t want my tool on one of the other EDA company’s cloud because there are easily things they can get access to.”  That’s a valid concern.

From left to right: Mahesh Turaga, vice president of business development for cloud at Cadence, Rob Aitken, fellow at Synopsys, and Richard Ho, vice president hardware engineering at Lightmatter. Source: Semiconductor Engineering

SE: Are there are ways of managing that for security concerns or IP security concerns?

Ho: One of the advantages of an EDA vendor cloud is they can provide support for their tools that are in there, which is very powerful and very helpful. But it also means that if you have a third party tool in there, they can go and access that, and yes, this probably brings up the question of comfort. Is a company comfortable installing their tool on one of the other company’s clouds?

Turaga: We have many instances of customers using third party IP on our platform, even Synopsys IP. It depends on what type of IP it is. We are open from our standpoint, but there are elements that need to be cached out to protect it.

Ho: Would you put your tool onto a Siemens cloud?

Turaga: It depends on the situation and the particular tool.

SE: Cost is another consideration of moving to the cloud. Let’s say I like one tool from this vendor, and another tool from another vendor. I could probably go cloud on both sides. But it’s so expensive, you have to make a choice of only one vendor. How should I choose?

Johnson: I think you’re starting to touch upon a different knob, which is what’s the extent of control that you want to have over those levels of detail from being able to pick a best in class flow, which means multiple tools from different vendors, to how you want to optimize it with your cloud vendor selection. You get to choose whether to give up a little bit of control for the simplicity of having it more homogenous from one of the EDA providers or go to the other extreme, which is deal with the complexities, but take full advantage of whatever infrastructure. I don’t think there’s going to be one answer.

SE: For users that run tools for, say, simulation and debug on a laptop, but then want to move the job off the laptop onto the cloud, how easy is it to do that today?

Johnson: If you’re designing a chip, I suspect that you’re probably running that on a server somewhere and launching it from your laptop versus running it locally?

Aitken: That goes back to the simplicity piece. If your design is simple enough, or what you’re doing is simple enough that it fits in laptop type storage and performance, etc., then it makes perfect sense to do that. The mechanics of transporting that environment into the cloud is definitely non-trivial. Maybe you can talk to a cloud vendor and say, “I want a laptop from you that’s the laptop cloud,” then maybe it would work easier, but I suspect that even then it would be a disaster.

Ho: The technology is not there yet.

Turaga: At Cadence, we’re looking at containerizing some of these applications to run in the cloud.  I think some of the verification tools have been asked for.

SE: What are some of the biggest lessons and takeaways in cloud today?

Johnson: Alignment of all of the various stakeholders on what the end objective really is. What’s important is what we are trying to prioritize and optimize for. We tend to have CAD, IT, design team, all with slightly different priorities of what the ideal solution would look like. Getting that group to align and then build a strategy based on those constraints is much more effective than those groups independently coming to EDA companies and asking for the optimal thing just for their portion of the solution.

Turaga: Another thing is cost. We’re trying to work with our customers to figure it out because it’s more than cost. At the end of the day, what’s the value of bringing your product to market ahead of the competition? How do you quantify the business value of reducing the time to market by a couple of months?

Aitken: The initial thought that people have is. “Oh, the cloud it will be much cheaper than what I do now. And it’ll be free and it’ll be magic and everything will be wonderful.” And then you go do it, and it’s a lot harder and more money and X, Y and Z. The other piece is that — even though this has been around for a while — we’re still in the early days on this. At the 70th anniversary DAC, I’m assuming that there will be cloud-native solutions to some of these problems that have been around for a long time, and that the existence of those will prompt new use models, new design models. What would happen if somebody built an EDA tool out of micro services? What would it look like?

SE: How long have EDA vendors been providing cloud services and how is this changing?

Turaga: At Cadence, even before cloud, we used to offer virtual CAD. That was where we used to do data services for our customers in our data centers. Then we moved that to public cloud in the mid-2010s and by 2017, we announced what we have now so we’ve been at it for a while.

Aitken: The nanoHUB TCAD stuff has been around for ages and ages. It’s effectively cloud-ish, not mature cloud.

Johnson: We’ve been using the cloud as part of services delivery for a long time, and it wasn’t really called cloud. We’d do this for emulation but also for software, for smaller customers that are interested in a full flow. The nomenclature has changed more recently as well as some of the practices. All of the solutions that you’re seeing in EDA right now that we’re calling cloud, that we are calling SAS, are still single tenant, virtual private chambers dedicated to a customer that’s generally been important for security reasons for the multitude of historical things. So, we’re not to the micro services yet, the full cloud native, but that’s something that obviously the industry is going to have to grapple with.

Ho: The biggest lesson is start with your chip design methodology. Figure out what your tool flow is, figure out which tools you want, best in class or single vendor, figure out where the data flows, then figure out how cloud fits into that because everything comes from your methodology. Time to tape out is what I measure. I think it’s the most important metric. How fast can I get? And that’s what I’m trying to optimize for, and cloud is a tool in that toolkit, but it has to be one.

SE: For the users here, what do you wish the EDA providers knew about cloud?

Steinke: That NFS is really darned expensive and doesn’t really work too well.

SE: Is the cost of cloud going to be a barrier to adoption?

Steinke: Yes, it’s certainly going to pose a challenge. It’s not just, “Oh, flip it on. This is the way to go.” You’ve got to consider how you’re doing it and the value that’s bringing to your methodology and the designs that you’re bringing to market. It all comes down to cost of the engineering resources, the people, the tools, and the compute that you need to bring that design to market. The cloud definitely has its own price tag on that. You have to see how that fits in with your strategy and how you’re bringing it all up.

Ho: I agree with that, but I do like the fact that we have the three major EDA vendors here offering cloud. Competition always keeps the price lower.

Turaga: Continuous cost optimization of cloud is the key, and that we have storage, compute, and we’re looking at data instances of an enterprise/on-premise, and continuously monitoring that.

Aitken: If price is the one barrier that’s left, the market will figure out how to deal with that. That’s not going to be the limiter. If there’s a limiter it will be some kind of technical issue.

Part one of this discussion is here. Part two of this discussion is here.