Cloud-based design
- Thread starter nitindeo
- Start date
Daniel Payne
Moderator
By your post title I assume that Cloud-based design is one new way to deliver batch-oriented tools to EDA users who don't want to install, maintain and hire a staff of CAD integrators.
It's a noble direction to pursue.
Larger design companies would be attracted to cloud-based design if they owned the cloud and the data never left their secured internal network.
Smaller design companies may not be so paranoid about their design data being stolen when using a cloud-based EDA tool flow.
The other idea that I hear bantered about in EDA is success-based pricing. I can only assume that it means if my chip works then I pay the EDA vendor a royalty, and if it doesn't work then no royalty.
Any other new ideas out there?
It's a noble direction to pursue.
Larger design companies would be attracted to cloud-based design if they owned the cloud and the data never left their secured internal network.
Smaller design companies may not be so paranoid about their design data being stolen when using a cloud-based EDA tool flow.
The other idea that I hear bantered about in EDA is success-based pricing. I can only assume that it means if my chip works then I pay the EDA vendor a royalty, and if it doesn't work then no royalty.
Any other new ideas out there?
Hiroshi Komoda
New member
I surely think that the business model that used a cloud is one choice.
It may be a nice thing for the user to pay money depending on the work of the tool, but how will for EDA vendor?
It may be a nice thing for the user to pay money depending on the work of the tool, but how will for EDA vendor?
R
Randy Smith
Guest
Cloud-based computing is likely the future of EDA design but there are some significant challenges - especially in physical design where data volumes are tremendous. Clouds need to be tailored to each user since jumping between vendor clouds would mean passing huge volumes of data over the internet. No single vendor has the state of the art tool in each category. The best chance at a complete solution, even if only for physical design, would need to come from a foundry where they could support the multiple vendor tools in their reference flow. This is not a trivial task and is far removed from the core expertise of the foundries. While I would hope to see it soon, I anticipate this will take many years and likely will slowly evolve from the Mask Data Prep (MDP) area and proceed up the tool chain. Cloud computing can be much more easily be supported in the ESL space where data volumes are generally smaller. The challenge there is a lack of a standardized design flow. This, however, can be addressed by the major EDA vendors.
hedley rokos
New member
Head in the sand?
Cloud-based processing readily makes sense for mobile applications where local processor power (both senses) has to be limited - but I don't see much of EDA going this way.
I can also see the benefit for situations where the associated hardware is used in peaks and is a large proportion of the total variable costs (and the cloud simply runs the same software as you do internally).
But I don't really understand the benefits of the cloud during the bulk of the design process where variable hardware costs are relatively small, and licenses are already paid on a usage basis.
In principle it would be nice to be able to "share support staff", but I don't see the cloud as being necessary for this - though it could I suppose act as a catalyst. On the other hand, one important differentiator of EDA suites is how far software support is unnecessary except when you wish to do something that is specific to your requirement; I suspect many of us have seen attempts to share this requirements-specific support - and I would be interested to hear of any successes.
Am I wrong simply to view the cloud as a new opportunity for outsourcing? If I'm correct, I would recommend the greatest caution in going down this route - at least until the undocumented parameters of the internal support can be included in the outsourcing specification.
BTW, what Randysan writes makes sense to me - until you need to transfer the shrink to a different foundry. I therefore fear that anything beyond provision of raw processing will need service specialists who can support a suitable range of foundries and of EDA tools - plus the customers' special requirements
Cloud-based processing readily makes sense for mobile applications where local processor power (both senses) has to be limited - but I don't see much of EDA going this way.
I can also see the benefit for situations where the associated hardware is used in peaks and is a large proportion of the total variable costs (and the cloud simply runs the same software as you do internally).
But I don't really understand the benefits of the cloud during the bulk of the design process where variable hardware costs are relatively small, and licenses are already paid on a usage basis.
In principle it would be nice to be able to "share support staff", but I don't see the cloud as being necessary for this - though it could I suppose act as a catalyst. On the other hand, one important differentiator of EDA suites is how far software support is unnecessary except when you wish to do something that is specific to your requirement; I suspect many of us have seen attempts to share this requirements-specific support - and I would be interested to hear of any successes.
Am I wrong simply to view the cloud as a new opportunity for outsourcing? If I'm correct, I would recommend the greatest caution in going down this route - at least until the undocumented parameters of the internal support can be included in the outsourcing specification.
BTW, what Randysan writes makes sense to me - until you need to transfer the shrink to a different foundry. I therefore fear that anything beyond provision of raw processing will need service specialists who can support a suitable range of foundries and of EDA tools - plus the customers' special requirements
Last edited:
Mentor has a "cloud-like" offering where the customers and Mentor share a common "space" ... and the customers can offload some of the simulation/verification work, etc to this "space". The simulation and verification tools are already capable of multi-core work ... but I guess will need some more work before they can be classed as cloud capable.
One added advantage of of this offering is that customer support can be handled more effectively as both sides can see the software bugs easily (instead of creating test cases, setting up webex sessions, setting up the right tool versions ... etc ...)
is this the a step in the right direction ? ... I think so ...
If it is a success ... I guess the EDA companies can charge by "time" or "processing power" usage ... interesting ...
One added advantage of of this offering is that customer support can be handled more effectively as both sides can see the software bugs easily (instead of creating test cases, setting up webex sessions, setting up the right tool versions ... etc ...)
is this the a step in the right direction ? ... I think so ...
If it is a success ... I guess the EDA companies can charge by "time" or "processing power" usage ... interesting ...
R
ron
Guest
One way to view 'cloud' resources are as another form of consulting. Not meaning to over-trivialize it, but I can imagine the 'give us your design and we'll clean it up' kind of scenario. Think of a 'cut my power consumption in half' button for example. The cloud makes sense when you only need to use it infrequently, and compensation is directly tied to results. If I can give you a 'bad' netlist and you can return a 'good' one to me, that could make for a pretty compelling value proposition.
T
treforsouthwell
Guest
I don't think a lot of companies would accept having their data leave the security of their own network. I think the major EDA licenses should come with one compute farm CPU machine per license to be installed on the customers premises. The cost of the machine is small compared to the license cost and would resolve a lot of the capital investment issues that big companies face.
C
Camille Kokozaki
Guest
Forecast: Cloudy, but with a chance of increased computing
Many good insights listed here.
Nitindeo:
Here is a great conversation starter:
Above the Clouds: A Berkeley View of Cloud Computing.
I took a stab at proposing some new EDA models in a GSA Forum article:
On-demand Tool and IP Grid: Examining the Merits of Fully Leasable On-demand Engineering Utility Resourcing
Randysan:
I think many challenges can be overcome if the environment is set in the cloud and the reference flows and libraries are all set up to minimize file transfers. You are correct in stating that the foundries are a natural repository of the work launch, and this addresses security concerns as well. The foundries already have your crown jewels in the most imaginable physical incarnation.
hedley rokos:
Interoperability is key and migration to other foundries issues are there regardless of the transport/repository location. If a model is set where the customer container can be plugged and unplugged, then file migration is possible (of course you have the design migration issues and the cloud bears no blame for this portion). So you design with retargeting in mind (easier said than done I know).
treforsouthwell:
I respectfully disagree with the premise of security being achievable only on-premises (cannot help it with my wordplay), even though some companies will insist on having this happen. I can summarize by saying security can be provided at every step in the transaction and IT costs will ultimately dictate this, so unless you have enormous resources, you have to start using external data centers anyway.
If you have not already guessed, I am a big proponent of cloud computing and novel ways needed for EDA delivery through the cloud. Eventhough there will always be ten reasons to avoid doing anything, there is always that one reason more equal than the rest to do it at this juncture and that is 'faster TTM with convenient peak usage at an affordable price in a secure setting'. OK. I lied, this 'one' reason is multi-faceted.
Many good insights listed here.
Nitindeo:
Here is a great conversation starter:
Above the Clouds: A Berkeley View of Cloud Computing.
I took a stab at proposing some new EDA models in a GSA Forum article:
On-demand Tool and IP Grid: Examining the Merits of Fully Leasable On-demand Engineering Utility Resourcing
Randysan:
I think many challenges can be overcome if the environment is set in the cloud and the reference flows and libraries are all set up to minimize file transfers. You are correct in stating that the foundries are a natural repository of the work launch, and this addresses security concerns as well. The foundries already have your crown jewels in the most imaginable physical incarnation.
hedley rokos:
Interoperability is key and migration to other foundries issues are there regardless of the transport/repository location. If a model is set where the customer container can be plugged and unplugged, then file migration is possible (of course you have the design migration issues and the cloud bears no blame for this portion). So you design with retargeting in mind (easier said than done I know).
treforsouthwell:
I respectfully disagree with the premise of security being achievable only on-premises (cannot help it with my wordplay), even though some companies will insist on having this happen. I can summarize by saying security can be provided at every step in the transaction and IT costs will ultimately dictate this, so unless you have enormous resources, you have to start using external data centers anyway.
If you have not already guessed, I am a big proponent of cloud computing and novel ways needed for EDA delivery through the cloud. Eventhough there will always be ten reasons to avoid doing anything, there is always that one reason more equal than the rest to do it at this juncture and that is 'faster TTM with convenient peak usage at an affordable price in a secure setting'. OK. I lied, this 'one' reason is multi-faceted.
Daniel Payne
Moderator
FYI - Mentor is hiring for an IT geek to help out their SAAS offering: IT Consultant (SaaS) at Mentor Graphics in Newbury, Berkshire | LinkedIn
Camille, you are correct that it is a cost analysis. Add the compute center cost savings to the perceived value of "faster TTM". And then subtract the cost of the service itself plus the perceived cost of security by allowing the data to leave the premises. It is not entirely different than the transition from IDM to fabless/fablite.
D
DigitalSystems
Guest
Someone has already been thinking in that direction: Teahlab (www.teahlab.com) offers an online EDA. The site focuses on undergraduate students so the EDA is for small circuits. But it is definitely the first step towards an online EDA. Teahlab not only allows users to create and save simulations completely online, but users can also embed simulations in forum posts for other users to simulate and provide effective feedback. Check it out; it's pretty cool.