The Breker Trekker 
Tom Anderson, VP of Marketing
Tom Anderson is vice president of Marketing for Breker Verification Systems. He previously served as Product Management Group Director for Advanced Verification Solutions at Cadence, Technical Marketing Director in the Verification Group at Synopsys and Vice President of Applications Engineering at … More » Should EDA Still Have Its Head in the Cloud?July 2nd, 2013 by Tom Anderson, VP of Marketing
A year or two ago, cloud computing was a hot topic at EDA events and conferences. The industry’s largest player, Synopsys, talked a lot about running their EDA tools in the cloud and claimed a great deal of customer interest in this area. Earlier this year, CEO Aart de Geus was quoted as saying that “Synopsys had made $0 on it.” What happened? Is the idea of EDA in the cloud dead? What are the issues preventing its adoption? In talking to customers and reading related articles, I’ve heard of several reasons why the cloud is not yet a factor for EDA. One practical issue is that EDA vendors are not sure how to price cloud-based licenses. If the customer motivation to move to the cloud is to more easily handle infrequently used tools or to provide peak capacity for frequently used tools, indeed the wrong pricing model could cost the vendor money. One technical issue is that some EDA tools are interactive and would be painful to use over the Internet.
But the most common issues raised surround IP protection. Engineers know the limitations of today’s computer security solutions and worry that competitors will be able to get access to design and verification IP in a shared cloud. RTL source seems to be the most sensitive form of IP. However, other aspects of the design and verification environments such as synthesis constraints, assertions or testbench code might also contain critical and highly proprietary information about the customer’s product. Specifically in the simulation domain, none of the major vendors has seen a push by customers to run in the cloud. The RTL source is considered too proprietary, and a compiled environment may be too large to transfer back and forth in reasonable time. In the case of Breker, our test-case generation is already very fast, so the only part of our solution that might make sense to run in the cloud is the simulation of our test cases. I suspect that we will just follow along when and if simulation moves to the cloud. My thoughts on this topic were prompted by a very interesting article on cloud support from formal vendor OneSpin. Their approach is to compile the design and assertions locally and then run the formal algorithms in the cloud. Their argument is that the abstract model would reveal very little about the design even if it were somehow compromised. Further, formal analyzes only a small portion of a chip at a time and so one couldn’t learn much about an SoC from such a portion. I found this approach intriguing and wonder whether it might be successful. If you’ve spent any time thinking about EDA in the cloud, please comment. What do you think of the OneSpin method? Could you imagine any sort of abstract simulation model in the cloud with which you’d be comfortable? Your thoughts are most welcome! Tom A. The truth is out there … sometimes it’s in a blog. Tags: Breker, Cadence, cloud computing, EDA, functional verification, SoC verification 3 Responses to “Should EDA Still Have Its Head in the Cloud?”Warning: Undefined variable $user_ID in /www/www10/htdocs/blogs/wp-content/themes/ibs_default/comments.php on line 83 You must be logged in to post a comment. |
|
|
|||||
|
|
|||||
|
|||||
EDA Cloud Computing (I hate buzz words like this, I cringed when the term “DEEP SUB MICRON” got thrown around) To me (been working in VLSI since 3 micron days – yes, microns 🙂 this is just an evolutionary step down down the technology road…
I’m hoping one day to be able to see (I do a lot of simulation at device level, IR drop, power analysis (SOC down to device), static timing, etc.) the ubiquitousness of my tools not only supporting multiple CPU’s, multiple threaded CPU’s, and even GPU’s. But, analogous to Google queries and massive data analytics, be able to run a single job across my network of hundreds of machines and query the results as they are being generated (e.g., portions of the parallel jobs may be not even started, incomplete, being moved to another server, etc.).
Alas, I question if the economics of putting that in place by any EDA vendor could ever be supported by customers like myself (i.e., we already balk at the price of some of these tools).
David,
Thank you for your comments. You’re touching on two related topics: the ability to run EDA tools on massive server farms, and whether the farms are owned by you or in the cloud. You specifically use the phrase “my network of hundreds of machines”; does this mean that you would not be comfortable running in a cloud? As for running in parallel, there are many EDA tools today that do take advantage of multiprocessor servers and server farms. A few even run on GPUs although this is still an area of research.
Tom A.
[…] took the industry so long? That’s a topic I first explored in a blog post almost exactly five years ago. At that time, there was quite a bit of buzz about EDA in the cloud […]