IT Practices Stream over the Embedded Border: Q&A with Michel Genard, Wind River
For automotive, industrial, mil aero, medical, smart energy and consumer markets, the IoT is making demands—and making a strong case for simulation before deployment in the mainstream embedded world.
Editor’s note: Michel Genard, vice president and general manager of system simulation at Wind River, spoke to Embedded Intel® Solutions shortly before Wind River announced that it had released its latest version of SIMICS. Edited excerpts follow.
Embedded Intel Solutions: What are the challenges you see as customers shepherd projects through the life cycle?
Michel Genard, Wind River: It comes down to a few simple things:
One, the customer never has enough access to targets (boards to prototype) for their systems so that they can equip every single one of their engineers through the life cycle to the same system. That lack of access means that it then becomes a challenge to collaborate consistently across functions and departments.
The second challenge is: how can we be more efficient? Time and time again, we see that hardware and software complexity increases. From an efficiency point of view, you not going to be able to scale just by adding more engineers, you need to find a way to automate or bring efficiency in producing software so that you can release your product.
Embedded Intel Solutions: Where does the IoT come into all this?
Genard: The IoT use case that has been very interesting for customers is really to model out the system of systems. Designing the gateway itself is not that difficult. You can get from many different partners some software stack or whatever. It’s easy to get something that gets your platform gateway up and running. Connecting the gateway to some nodes is not that difficult either. What is much more complex is determining before deployment how you can set your system of systems.
Knowing where can you set your many gateways connecting to your field nodes, with all of these connecting back to your data center or cloud—it’s all extremely difficult. And we have seen a great use case where customers [use SIMICS] to simulate the system of systems gateway nodes—all in a single simulation, so, that on a desktop, they can debug those kinds of systems.
And by the way this is why, although automotive is a new area for SIMICS, SIMICS makes sense in that market because vehicles are becoming a network with a lot of devices. In other words, the car is becoming a system of systems (Figure 1).
Embedded Intel Solutions: You’ve noted that practices familiar to the IT world are crossing into the embedded world, with the IoT serving as catalyst.
Genard: In IT, the concept of “I simulate before I deploy” is something that has been there for years. Whereas in the embedded world, the idea of simulation prior to deployment, while it exists, has not been mainstream. It’s been used in some markets when the application is a bit more complex.
What’s changing is the rise of the IoT. If you look at IoT from an infrastructure and an architecture point of view, it is making the case for simulation big time. Fundamentally [with the IoT] you are never going to be in the position to really have a completed system that you can test.
As an example we have a customer using SIMICS to simulate a smart metering system. This customer’s utility company customer deploys gateways, and each gateway connects to thousands of smart meters. And this is something that fundamentally cannot be tested in the lab. They can test a couple of gateways with a few hundred smart meters, but there is no way to actually test the real system by deploying in the lab all of the systems together.
[Without simulation] they would have to test when they deploy, and this is where there is obviously high cost. What we are seeing with simulation and SIMICS is that with this kind of opportunity in the lab you can simulate the full, deployed system, with many gateways, thousands of smart meters, all embedded devices with the actual code, that is, with the software that is going to be deployed and from the software perspective see how the systems relate. You can force errors, look at “what ifs” and the like.
In the same way that IT has been using virtual machines and simulation before deployment to manage deployed systems, we believe that we are going to see the same thing in the IoT world.
Embedded Intel Solutions: What other changes do you see the IoT bringing?
Genard: One interesting dynamic I am noticing is the question of “who is going to manage IoT systems?” IoT systems can be considered an IT system in the sense that it is going to be deployed. So is it going to be managed by IT people who [traditionally] would manage a system to be deployed? Yet the point about an IoT system is that you deploy and you push update features and use cases all the time—so by definition it is never going to be done.
The IoT is really going to change not only how the company and users are thinking about the development of systems, it is also going to change how you think about the deployment of systems and there is a lot that can be learned from the IT world here.
Embedded Intel Solutions: How do you balance conveying that the world has changed and at the same time communicating what you want embedded developers to know about specific Wind River offerings?
Genard: Welcome to my world. It is always a balancing act, indeed. An engineer needs to connect to: “I have a job to do; I have a task to accomplish—what are the roadblocks and how can I find a solution to work through the roadblocks?” There is a use case focus. Engaging with an engineer is about focus on the pain points, on the things that hurt his or her success.
IoT’s advent tips the scales toward management discussions where the emphasis is: what is going to happen in three to five years. This is where we offer perspective: knowing that because a, b, c and d are going to happen, “e” is going to happen, and here are the steps to take to succeed.
Embedded Intel Solutions: What messages are you most concerned about getting across?
Genard: When I meet with customers, the one thing I hope will resonate with them is really around the idea that it is no longer about trying to put more people on the job, or using people for offshore development or trying to find the new debugger that will “by magic” find all the problems that you have not found for years.
Using simulation and tools like SIMICS will help customers think about how they can transform and how they can come back with a new way of doing things. You cannot deploy those complex systems that will be deployed in the IoT world in the same way that we used to develop 16- or 32-bit systems four or five years ago—it does not scale.
This is what I spend time on: helping the customer to think differently, rather than to think, “I need an engineer cheaper that can do more things,” or “I need a better debugger.” The industry has been working on debuggers forever.