Last week, ARM launched it ‘Total IoT Solution’ based around virtual models of hardware and software from Project Centauri.
His colleague Tran Nguyen, senior director of design services at ARM, last week detailed the way the technology will be used, with a developer spinning up an instance of a virtual model on a virtual machine on Amazon’s cloud service. But who is going to use this? EDA vendors have been offering virtual models for years for early software development, but this opens up the technology to a broader market of application software developers rather than embedded developers.
- Boosting IoT rollout with virtual hardware
- Virtual modelling toolchain to boost AIoT
- Virtual models speed up ASIC verification
- World’s fastest deep learning inference software
“The way we think about this is four different types of developer personas,” said Mohamed Awad, VP, IoT at ARM and, importantly, a board director at board maker Arduino.
“The first is the classic silicon partner developers that are doing chip bring up and low level drivers and it’s not aimed at them. Then there’s the classic embedded developers, the example I like to use is a pool controller with hardware and firmware, it’s a single device. They can benefit from this, and we have historically serviced them with Keil Studio.”
“Then you have got two other classes. A big OEM building out a software and service offering for the IoT may have lots of different devices that will be part of that offering with hardware and software under their control. There’s an element of convenience and scalability that ARM virtual hardware and software offers,” he said
“Then there is the fourth persona which is where the company is going and it’s the most exciting, That group of developers is developing some type of functionality to deploy across a broad set of IoT devices. A great example is Amazon and the wake word engine.”
The Alexa wake word has on operate on all kinds of chips in many different types of devices, from smart speakers to TVs and even lights. Rather than having to develop different software for each different device, ARM sees more of an ‘app’ ecosystem of software blocks that are easily integrated and shareable, in the same way that Docker or Kubernetes software containers are used by cloud developers to re-use software. However, these are used on ARM Cortex M55 microcontrollers that don;t run Linux or high level opeating systems, although A-class cores that will run Linux and Linaro are also planned for this IoT tool.
“There is no pre-requisite for Linux and containers,” said Awad. The portability comes from recent development in packaging embedded software through the Common Microcontroller Software Interface Standard (CMSIS).
“Project Centauri is Open CMSIS to open up CMSIS packaging with ST and NXP and put that into Linaro so there will be bits of software that will standardised, we used them as a mechanism to open this up so that it supports other workflows.”
“That fourth persona is the most exciting as it changes the economics of the IoT. The most sophisticated developers were using these pieces of software to deploy common software, with continuous integration and continuous development (CI/CD) in the cloud.”
He points to partners Impulse and DSP Concepts working on software development and deploying it in operation, hence DevOps.
“There will be lots of interesting areas particularly in DevOPs as the expertise is still pretty specialised and the people that really understand how to do this have their own workflows in mind,” he said, “Expecting them to be embedded developers is not really realistic and I think that’s one of the biggest near term advantages.”
“When you look at cloud native developers they really don’t have the expertise or the desire to develop for an embedded platform one device at a time,” he said. “It’s about scale. We will continue to work with those partners and working out how to work with virtual models, from the more cloud native tool developers through to our more traditional ecosystem partners.”
- Bosch looks to AIoT as revenue falls
- AIoT filter library is microcontroller agnostic
- Low-cost AI processor targets AIoT
- AIoT startup closes funding to commercialize AI-in-sensor chips
Other articles on eeNews Europe
- Nextchip licenses aiWare4 for automotive AI processor
- DuPont in $5bn deal to focus on electronics and automotive materials
- ProLogium raises $326m to bring solid state batteries to Europe
- Amazon details KuiperSat broadband satellites
- Cycle-resolution timing boosts real time operating system