Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
Performance, Security, Behavior: how Silicon Cloud’s driving the IoT
February 26th, 2015 by Peggy Aycinena
In a recent phone call, Silicon Cloud co-founder and CEO, Mojy Chian asserted that the IoT has inspired a new set of initiatives from his lively startup enterprise.
“If you look at the Internet of Things in its entirety, it includes the transmission, aggregation, processing, cloud services and so on.” Chain said, “But it all starts with the nodes, the endpoints, which are the sensors.
“Today we are positioning Silicon Cloud to provide a design-enablement infrastructure for the IoT, a large part of which is the chip design. But now we are extending our services to include design tools for sensors, concentrating on the node itself, and providing a complete infrastructure for node design enablement. We don’t design the semiconductor or the sensor, of course. We provide the design enablement for others to use to design these things.”
Chian spoke about drivers for the semiconductor supply chain: “What predominantly drives semiconductors are the mobile apps, the stuff that goes into our smart phones, tablets and so on. There are not a lot of different variants, however, among the semiconductor devices that [support] those mobile apps.
“Look at the processors from Samsung, Qualcomm or Apple. They’re all essentially the same, manufactured in volume, with the companies spending hundreds of millions of dollars to produce these things. They use leading-edge technologies, but there is little variety and therefore the semiconductor industry has consolidated.
“That’s why the big guys are getting bigger and making even more money, while the small guys are [being pushed] out of the game. There are very few small chip-design companies and few VCs to fund them. Again, that’s because of the type of devices for this space.”
Chian suggested shifting the search for opportunities: “If you shift your view to IoT devices, they have a completely different set of demands. There are hundreds and thousands of types of IoT devices out there, and most do not need leading edge process technologies. In fact, 65 nanometers are good enough. These devices are simple to design and don’t require hundreds of millions to build them. They’re closer to half a million dollars to design.”
“These attributes,” he continued, “will cultivate and bring about a new breed of chip design company, both through the variety of devices needed and the low cost of development. There is no need for leading edge process technologies or cutting-edge design skills. Instead, a lot of IoT device design skills will be cultivated through the universities.
“Taken together, all of this is relevant to our value proposition, why chip design in the cloud is important, and why it will grow in relevance over the next 5-to-10 years.”
His line of thinking clarified, I asked which universities Silicon Cloud is targeting with their proof-of-concept cloud-based IoT device design enablement infrastructure [try saying that 10 times fast] and what are the costs of participation.
Chian said, “Back in September 2014, we launched a pilot program with 11 universities in Singapore, Malaysia, and the UAE, in particular Abu Dhabi, with over 30 people participating. The intention was to give our user base the look and feel of the system, to get feedback, and to demonstrate that the system is operational and working – not just a demonstration vehicle. That was a milestone for us, which we achieved.
“Our second major milestone was to establish a second cloud center in Singapore – our original center is in South Carolina. With funding from the government of Singapore, that second cloud center is set to be operational this month.
“How do we charge universities?” he asked but answered instead, “First, full engagement with universities [has helped] build our initial user base and global footprint, so we see the universities as enablers of our technology.
“But universities are not long-term customers for the company. Our systems are designed to provide in-cloud services for a range of engineering and scientific applications. Long term, we plan commercial offerings that will not just be [targeted at] IC design, but will concentrate on the entirety of IoT enablement.
“This will include mechanical analysis, sensor analysis, MEMS design, and pressure analyzers – a collection we call multi-physics tools – that will have good synergy with chip design tools in the IoT space. It’s very logical to provide design enablement components for the sensor part of it, so designers of IoT devices can design all the way from sensor to the chip.”
As an aside, I asked Chian, “When did the IoT emerge; when was the term first used?”
He laughed: “All I can tell you is that large companies like Cisco have used [the concept for a long time] as part of their company’s overall mission and agenda. This is where they see future growth, and so does the rest of the world!”
Turning to a thornier question, I asked: “Security is such a problem today, everything from recent hacks of huge commercial and government organizations, to Twitter and YouTube. What can you do to prevent such incursions when the economic motivation is so vast in the areas of chip and IoT sensor design?”
Chian answered, “To clarify, some of what you mention here is about the end application, things like Twitter and YouTube, and not about the design. Nonetheless, security is an important issue and we know that.
“What we are offering is far more than just taking an exciting design infrastructure that’s been put on a server for remote access. [Our offerings] include a lot of high-performance computing, but also security – the two major technical issues that need to be addressed for us to succeed.
“High-performance computing in areas like banking or e-commerce require large machines, sometimes with as much as 256 gigs of RAM. But for chip design, we require machines with an even a larger footprint and ones that can run simulations for 2 or 3 days. So high-performance computing [at that level] must be provided to our user base and is a very important part of our value proposition.
“The second major issue is security, as you suggest. In the chip design space, issues of data protection and data tampering are equally important. In chip design, people have access to other people’s IP. When you are doing your banking online, you alone have access to your own data for downloading. When you’re doing chip design online, however, you have access to other people’s IP, their tools, and critical information from their silicon foundries. So data protection is very important!
“In our system, the entire EDA/PDK/user data all reside in the cloud, but the users cannot use just any old machine to access that data. They cannot use their own laptops or their own desktops. Instead, we issue controlled clients to our users, a small machine that’s registered securely in the cloud.
“It authenticates itself in the cloud, plus the user is authenticated. The machine we issue will not allow the user to have freelance download of the data. And the user cannot download data, PDKs, or tools through the machine if the work is done in the cloud.”
Chian elaborated: “One side benefit of using such controlled clients, is that not only can people not download, they cannot upload either. No viruses, no trojans, no malware can be uploaded into our system.
“[Additionally], we provide a sophisticated technology that saves users’ behavior in the cloud. The analogy is with shoppers on Amazon, a system that suggests other products the shopper might like based on their shopping history.
“[In our case], if the user goes to the cloud, uses a PDK from a particular foundry, and runs a particular simulation, our system then creates three interconnected nodes between [those ‘likes’] in our user data base. It notes the user used this simulator, this PDK, and this data at this particular time. Six months later, that same user can connect these tools again, because we have saved the behavioral data to help our users [with their productivity].”
“Behavioral data can also be useful for design compliance and auditing,” Chain said. “Have all of the necessary steps been done before sending the design off to tapeout? Have the appropriate number of regressions, LDR, DRC, etc., all been done?
“Additionally, the data can be used for provenance and back-tracking: Who generated this data, using what tools, and at what point in time?”
“As you can see,” he continued, “we are offering a very interesting combination of technologies that creates a powerful tool for the user. An infrastructure, by the way, that the user can use as much or as little as is needed. This combination of high-performance computing and design framework technology is so unique, we have applied for 9 patents covering our entire system.
“Some of our patents are related to high-performance virtual machines. Generally virtual machines have had problems and have not historically been able to perform at the same level of efficiency as bare metal machines. So some of our technology is about making virtual machines as efficient, allowing you to create on-demand a particular machine with a particular configuration, specifying the number of cores, the amount of memory, the amount of disk space and so on.
“Some of our other patents are associated with our user-behavior data base, how it’s generated and how it’s applied more specifically to applications related to IC design and other scientific and engineering problems.”
I was impressed: “All very leading edge, but perhaps you have some competition?”
Chian said the answer is good and not so good: “Currently, we have no competition in the space, which is a positive. But it’s also hard, because we have to prove our value proposition – users have never seen something like this.
“There are some entities today that provide remote computing; you log into their systems for to do scientific work or simulations. But no companies, to the best of my knowledge, are providing turnkey design infrastructures and associated user-behavior tracking.”
“This migration to the cloud,” I said, “has been a pretty tough sell in EDA. But it sounds like you might be offering a solution to that conundrum?”
Chian said the reality must be faced: “Whether it’s accepted or not, the migration to the cloud for chip design is inevitable. It is happening, it will happen, and we all need new technologies to make it secure.”
“It’s true, no system is totally secure,” he added. “But think about the fact that for most of us, our most important non-human valuables are in the cloud – our 401Ks, our checking, savings, CDs are all in the cloud. They’re out there and the potential that any of them could be tampered with is real, but that has not stopped us from moving our financial transactions to the cloud. It’s true for me, and probably for you, everything I own other than my house is in the cloud.
“Plus, look at all the apps in the cloud. There are so many! Non-secure apps that people use like Amazon Cloud, Rackspace, and so on.
“Of course, for more secure compute environments like banking, that is done in a private cloud. All of your data is protected there by the bank. It’s not in the public cloud. But depending on the type of app and the level of security required, we are seeing a variety of different infrastructures to meet those needs.
“[Not surprisingly], at Silicon Cloud we also have a private cloud. These are machines that we own and operate, and have under lock and chain. This is in response to some users’ needs for a private versus public cloud.”
“Overall,” Chian said, “we know that people in both science and engineering are very sensitive to security issues. But our continuum of technology developments means we will make [the solutions] happen. We are providing a trusted environment with protection of the users data – even we do not have the access to look at their data – with the entirety of their file security being our responsibility.”
“However,” he warned, “the users also have a responsibility. They have to sign software licenses with their EDA companies, and IP license agreements with their IP providers. The users have to be cognizant of the dangers of unauthorized use of those tools, the IP, and the data.”
Chian ended with a look into his crystal ball: “As we work at Silicon Cloud to move beyond the universities, our potential customer base will see we are talking about much more here than just putting tools and data on remote servers. We’re talking about users being able to log into our systems to access high-performance computing and high levels of security in a trusted environment. That, combined with our user-behavior profiles, is the future of cloud computing.
“We are going to be the leaders in making cloud computing a reality, everything from the chip to the sensor and the IoT. It’s happening, it will happen, and we are developing the technology to support it!”