Fuel for the Next Generation of Chip Design
Q&A with Silicon Cloud CEO Mojy Chian
Goliaths, here come the Davids: How IoT is changing a world in which “large semiconductor companies have become larger, and small semiconductor companies have almost vanished.”
Answers were certainly forthcoming from Mojy Chian, Silicon Cloud International’s hyper-energetic CEO during our recent conversation. But Chian also generated questions, rapid fire, on the topic of security, as you’ll read below, along with Chian’s perspective on sensors, the IoT, the challenge of implementing a nontraditional model and more. Edited excerpts of the interview follow:
EECatalog: Let’s start with a brief introduction to Silicon Cloud.
Mojy Chian, Silicon Cloud: Silicon Cloud offers a complete turnkey design platform in the cloud, one based on private clouds [rather than Amazon cloud and the like]. Our design platform as a service encompasses the entirety of the design environment, that is, EDA tools, manufacturing data, Process Design Kits (PDKs), design IP, design flows, the computing machines, storage—everything is in the cloud.
It’s similar to Facebook, in that the user opens a browser and goes to a dashboard. And in that browser he or she performs design activities and tasks.
Users do not use their laptop or desktop computers to connect to our cloud. We issue a specific client, a controlled client. It is only through these controlled clients we issue that users can access the cloud. A major [reason] that we do this is for applying security and data protection.
Another component of our solution that sets us apart is that we are using entirely virtual machines for our cloud servers. We do not have any preconfigured machines. Instead, users create [virtual] machines. As a [Silicon Cloud] user, one of your primary tasks is to say what machine you want.
You might, for instance, specify that you want a machine with eight cores, 128 gig of RAM and two terabytes of data. The system builds that machine for you right at the time you use it, and when you are done, the machine goes away, and the resources go back to the pool of the virtual machine resources.
EECatalog: Is there anything else like what you are doing out there?
Chian, Silicon Cloud: This is the first time in the semiconductor business that a company is offering a turnkey design platform in the cloud on a pay-per-use-model. We have a multivendor environment not tied to any particular EDA company or to any particular foundry.
EECatalog: In what ways is the “vision” of moving the engineering flow into the cloud no longer a vision, but rather a reality?
Chian, Silicon Cloud: Answering that question brings us to the announcement we made at DAC, where we announced a commercial offering of our platform on a pay-per-use model and accepting customers starting in September of this year.
And in July we will announce our initial set of foundry, EDA and cloud technology partners
EECatalog: What hurdles are involved in this process?
Chian, Silicon Cloud: The biggest hurdle in this process is the interactions with the partners and making partners part of this new model because this is a nontraditional model.
And [within this set of partners] the partners are at different stages of recognizing and appreciating this new model. Some of the partners, independent of us, had been thinking about this model, so when we approached them, they said, “Great! We had been thinking about this already, so let’s do something together.”
And there were other partners who were very far from this model, so that when we approached them and started talking about it, they said, “Well, we are not quite sure we like this. What’s the impact to my traditional model?”
We’re experiencing a very wide spectrum of partners and different levels of enthusiasm and interest with regard to embracing this new model.
EECatalog: How much of a catalyst to getting folks on board with this is the IoT?
Chian, Silicon Cloud: It is the biggest lever, because in the mobile era, large semiconductor companies have become larger, and small semiconductor companies have almost vanished. So if you look at today’s semiconductor space, which is primarily driven by mobile devices, when we approach partners and say, “here is this new business model for small companies,” they say, “Why bother? Whether I like this model or I don’t like this model, the user space is so small, it is not even worth it for me to even try to build a new model.”
Now, with IoT, the equation is changing, because IoT has specific attributes.
One of those IoT attributes is that the devices are much smaller, and it costs much less to build chips.
Two, there is a large variety and diversity of devices. [The companies in this space] don’t need to use leading-edge process technologies. They can use process technologies that are two- or three-generations old. When you put all of this together, this is a huge motivator to create the next wave of small chip design companies, not only in the U.S., but worldwide.
EECatalog: What are you seeing as this IoT lever takes hold?
Chian, Silicon Cloud: We are seeing small chip design companies coming up in the U.S., in Singapore, in China, in India and other places. This is something that is helping this mission [substantially] because we all realize that these are small companies, their budget for chip design is very small; with a few hundred thousand dollars they have to develop the chip and send it to manufacturing. They can’t spend a lot of money on the fixed cost of building a semiconductor design infrastructure.
EECatalog: As the next wave, as just noted, of small chip design companies comes into play, what is Silicon Cloud’s role?
Chian, Silicon Cloud: Silicon Cloud is positioned to provide IoT design enablement. So, [although] we are not an IoT design company, we provide the platform, infrastructure and workflows for IoT semiconductor design.
Our core technology is based on two components. One is high-performance cloud infrastructure, the other component is a modern workflow for engineering and scientific applications. Our core technology has been specifically designed for engineering and scientific applications, rather than small transactions. We designed both the workflows and the infrastructure for where the simulations or transactions are very compute intensive and require high performance and require security, like, [for example] semiconductor design.
But [semiconductor design] is not the only space that has these requirements and attributes. For example, for IoT, other than semiconductor design, you also need to do sensor design. So in our system, in addition to semiconductor design, we also have tools for 3D modeling.
Going forward we will add MEMS design tools, multiphysics tools for tasks such as electo-thermal simulation and so on. This system can be applied to any other engineering or scientific applications that require intensive computing and security. And that could apply to, for example, DNA sequencing, because the type of simulation and compute requirements are similar to those of semiconductors. We don’t intend to do that right now because we are a small company and a lot of our knowledge base is related to semiconductor and IoT, so we are focusing on semiconductor and IoT, but the core technology can be applied to other spaces.
EECatalog: Security certainly has a huge part in all this.
Chian, Silicon Cloud: Yes. Many people ask, “If this cloud-based design platform is so good and renders all these benefits, why hasn’t this been done before?”
Historically there have been technology impediments for doing design in the cloud. Two of these were security and high-performance computing.
Security concerns both data protection and tampering.
If I am an EDA company or a semiconductor foundry and I put my data in the cloud, I am worried about data protection. How can I be sure that there is not unauthorized use or unauthorized downloading? Or if I am a designer, and I put my data in the cloud, how can I be sure other people don’t steal it? And on the issue of tampering, how can I be sure that data won’t be harmed, by the insertion of a virus, for example.
The foundries are very concerned about data protection. Especially for leading edge process technologies, companies like GlobalFoundries or TSMC spend hundreds of millions of dollars, sometimes a billion dollars or more on developing their process technologies, and the information about that process technology is embedded in the process design kit. Today, in order to do design, the users have to download that PDK.
So think about it, all these small companies and all these universities around the world each individually have to download these PDKs to their local machines. [That means] there are hundreds and potentially thousands of instances of this very expensive data that resides all over the place on small company’s machines. How do you protect this data? How do you even track it? Who has it? Who doesn’t have it? Who gave it away? If a competitor gets hold of it, how would you even know who gave it to them?
This is why the foundries are very, very selective about whom they engage with, so if you are a small company, you cannot necessarily get manufacturing information from these guys. You have to convince them that you are a trustworthy entity to get that, and many small companies cannot get leading-edge process technologies or any process technologies the foundry cares about. So that is a major issue. So that was a major component of our thinking that when doing the design in the cloud, you are not going to have hundreds of thousands of instances of that very important piece of data all over the world, it’s all in one place, and nobody can download it. Users can go in the cloud and use it, but they cannot download it.
EECatalog: How does Silicon Cloud address security issues?
Chian, Silicon Cloud: One of the measures is that we don’t allow users to use their laptops and desktops to connect to the cloud.
[Instead] we issue the users a controlled client. Among many security features of this controlled client are controlled downloading and uploading, which means that the system, by default, cannot download or upload anything. The system does allow for downloading and uploading with specific authorization. For example, if initially the user has some design he wants to port to the cloud, it can be uploaded, or, [at the end of the design process] when the design in done and they want to download it; we allow that—but it has to be authorized. We believe our authorization scheme for uploading and downloading significantly improves data protection.
The other piece is our Virtual Private Machines. The concept here is that just like VPN, which creates a secure and exclusive tunnel. If you use VPN, you create a secure and exclusive tunnel between your machine and the servers—no one else can come into this tunnel, and this tunnel is only for you.
We have the same concept for our Virtual Machines. As the users get on our cloud, they create one or more Virtual Machines themselves. And those Virtual Machines are exclusive to that user and can only be used by that user. When the job is done the machines go away.
Security was at the foundation of our technology development from Day 1. We knew if we did not address security the industry, users and partners are not going to accept us.
EECatalog: How has working with universities informed what you are doing now?
Chian, Silicon Cloud: Our first market was universities, and from [a security] perspective, universities are extreme cases, because universities are not in the business of data protection, they are in the business of information dissemination. It is not in their nature to have a secure environment, they want to disseminate everything. So we had to build a system that is university- and student-proof from the get-go.
Universities have been a great partner for us, they are not only the early users of our systems who can point out to us how we can improve the system and the user experience and so on, but we also have partner universities. Universities are providing content for training modules, design flow development; example design development and many other areas.
EECatalog: Are EDA tools where they need to be to take full advantage of the IoT?
Chian, Silicon Cloud: Generally speaking I would say, “no.” But there is a spectrum, and the answer would also depend on which EDA vendor you are referring to. There are two shortcomings, one being the business model and the use model. The other one is the on the technical side.
We will have thousands or tens of thousands of these small companies downloading PDKs to their local machines, EDA tools to their local machines, building their own design environment, buying their own computers, this is just not a workable model—there is a lot of duplication, many cannot afford to do this; many don’t know how to do this.
On the technical side, if you look at IoT devices, it’s really a combination of two pieces—the semiconductor piece and the sensor piece. There are not a lot of EDA tools and capabilities that can do a total system analysis, simulation and design, so you need a different set of tools, capabilities, ways of doing things, and an environment for the combined sensor and chip design.
This is something that over time, needs to come together much better, the tools need to be able to interact better together.
Anne Fisher is managing editor of EECatalog.com. Her experience has included opportunities to cover a wide range of embedded solutions in the PICMG ecosystem as well as other technologies. Anne enjoys bringing embedded designers and developers solutions to technology challenges as described by their peers as well as insight and analysis from industry leaders. She can be reached at email@example.com