Poll: Do you want your EDA tools in the cloud?

[Daniel Nenni]

At the EDA CEO Outlook event EDA in the cloud was resurrected yet again with Dean Drako saying now is the right time. Is it? If you vote yes or no please explain.

View attachment 21488

Here is a discussion we had on SemiWiki back in 2012:

Chip in the Clouds - "Gathering"

 

[CharlieD]

I think security (control) will be the overriding issue. You would have the same limits on simulations in the cloud unless you have unlimited budget.

 

[Tanj]

The premise requires the tools to scale up and out. The benefit would be you get huge resources to reduce your latency. 30 hours of tools completed in 30 minutes. Since you share these resources with many other users and only pay for what you use, it allows huge bursts. The precedent for this is that software engineering in large companies has converted to this model. Of course, they have been investing in it over the past decade in some cases, and huge progress has been made in parallel scale out.

For it to really work you need silicon tools to scale out too. Asking this question without having tools available is not going to get a good answer. We can't even be sure how much scale-out is feasible - but in my opinion since the system elements are parallel, and the test/verification cases are huge in number, and the design rules can be checked independently it seems clear that scale out can be done.

 

[Daniel Nenni]

The question of cloud being good or bad for EDA financially always comes up. One of the companies I worked for offered tokens for peak usage on customer demand since the tool (verification) was not used daily. We offered the customer one license annually and batches of tokens. Revenue actually increased because engineers could get approval for tokens much easier than licenses. I would expect the same to happen in the cloud. It would definitely be good for EDA financially. Even if the tools did not have peak usage issues the EDA companies would just price tools accordingly with the customer saving money on the server side of the business.

 

[Bernard Murphy]

Another view is that yes, we want tools in the cloud but a private cloud. For EDA an exception is emulation where capital cost is very high and there is some openness to run jobs or spill overflow into a vendor cloud.

Another area to watch for this phenomenon (public cloud versus private cloud) is IoT management

 

[Daniel Payne]

The EDA cloud market may start out first with low-cost tool vendors serving the maker, IoT and startup niches.

 

[Staf_Verhaegen]

I voted yes for open source EDA in the cloud to make open silicon chips.