November 2018

BTL Mark: Resolve interoperability issues & increase buyer confidence
BACnet Testing Laboratories

(Click Message to Learn More)

What is Edge computing?

Edge computing is the movement of decentralised intelligence that reduces core computing environmental needs, reduces volumes of data that has to be moved and brings machine learning to the edge.

Parikshit Joshi
Parikshit Joshi
Lead - IoT and Data Science

New Products
Securing Buildings News
Site Search
Past Issues
Secured by Cimetrics
Control Solutions, Inc

Edge computing is the movement of decentralised intelligence that reduces core computing environmental needs, reduces volumes of data that has to be moved and brings machine learning to the edge.

While most implementations that we see today are around power intensive and process intensive IoT gateways; the inventor of LoRa, Nicolas Sornin has a different perspective. Nicolas believes that the future of edge computing lies in low power consuming AI hardware.

RISC-V is one of the most promising forefronts of for edge computing in IoT. And, as we move forward, you will see exactly how RISC-V can potentially do that.
Edge computing for IoT is expected to have a huge impact on how large-scale sensor networks of the future will be designed and managed.

RISC-V and IoT Edge

As we move towards decentralized and democratized processing for our business and consumer needs, we find most edge solutions being tested to the limits of engineering and creativity. Given the wide usage and industry-focused nature of edge computing, while a solution might just work fine in one of the areas, it may or may not work well at all in hundreds of other similar use cases. That’s why we need the ability to modify and customize what we already have. RISC-V is one of the biggest open source movements in the hardware world, and Edge computing can certainly benefit to a huge extent from it.

RISC-V for modular edge processors

Embedded machine learning requires open source. But, when we try to embed machine learning on the edge, we have to face issues where MCUs are incompatible or too closed sourced to innovate. Microprocessors, on the other hand, are often not preferred for edge devices as they incur additional power and costs.

We are seeing a steady rise in the number of edge processors like Greenwaves, that are building edge computing intelligence using RISC-V. Rather than throwing away the MCUs that they were already using in their own devices, these manufacturers are now using RISC-V for building embedded machine learning units to work alongside with their existing MCUs. This is an extremely lean approach, where it doesn’t require to rebuild everything from scratch and utilizes everything you already know to build a better, intelligent edge compute-focused device.

In order to save even more power, these manufacturers are paying close attention to dynamic voltage and frequency scaling. Most such RISC-V based implementations also own their own clock and voltage domains specifically for embedded intelligence units and general MCU.

Separation of concerns has proven its worth into software and hardware engineering. With RISC-V and MCUs owning their own voltage domains, they both don’t have to remain active all the time. This helps reducing power consumption for edge.

As we have access to instruction sets with an RISC-V based unit, we can implement hardware driven convolution engines that can carry out vital edge computing functions like:

  1. Object recognition
  2.  Sound and vibration analysis
  3. OCR
  4. Anomaly detection

RISC-V’s adoption in the intelligence compute module is being led by the fact that it has the ability and flexibility to support a lot of machine learning algorithms and bring revolutionary intelligence features to edge.

Edge crypto engine for IoT using RISC-V

A huge problem with IoT devices having their own crypto engines embedded was the cost and energy it takes to build one. If not done properly, it could consume an outrageous amount of energy to render it as a low power edge computing device. With closed source or licensed technologies, you can just do enough when it comes to optimization. You can’t get deep into the modules and optimize where you clearly see something isn’t right.

Also, cryptographically secured modules incur additional costs and still require someone with deep technical know-how to oversee the integration implementation.  This is a barrier to IoT adoption - as we can’t reduce costs, nor can we optimize for energy consumption.

RISC-V, on the other hand, gives you the ability to customize everything. Be it how the lowest level of processes is being executed, or how much energy your edge hardware consumes. When energy consumption and performance optimization is done properly, it makes developing crypto-engines a viable reality for low power edge computing devices.

When perfectly optimized, these devices take as little as 30 mJ of energy, which is extremely well suited for edge computing situations.

How to increase RISC-V’s performance at the Edge

  1. Reduce the on-chip memory size: By default, RISC-V has a 32KB memory and uses some form of external flash memory. You can completely eliminate the instruction memory and move it to the external flash. By doing so, it also shrinks the board size required from 0.06mm square to 0.004 mm square.
  2. Use 8-bit configuration from various viewpoints - This will be adequate for simpler edge applications.

These tips easily speed up RISC-V based hardware performance on the edge and are definitely something that needs to be looked into while building Edge hardware.

Target applications for RISC-V on the edge - IoT and Blockchain

So far everything that we spoke around was embedded intelligence for edge. Blockchain is another domain where edge devices have a lot of potential to be explored yet.

It is not too long since RISC-V’s launch, but we already see blockchain manufacturers signing up for RISC-V and building on top of it like Blockstream and Skynet.

Some of the common uses of RISC-V for blockchain on the edge:

  1. On-chip cryptocurrency hardware
  2. Decentralized M2M economy models for IoT
  3. License-free core design modules for  blockchain adoption

Going platform agnostic

This might go way beyond our current perception of perfectly working systems, but edge requires us to be platform agnostic. Edge is the future of decision making for IoT, and we expect millions of highly interactive devices to follow the edge paradigm very soon. What this also means is that these devices would bring in significant fragmentation in terms of how they communicate. From SMS to APIs, large-scale edge networks require us to communicate with a layer of abstraction. There are organizations like UnifiedInbox that are doing a pretty great job of bringing communication agnostic IoT infrastructure.

As edgi-fication of IoT increases, platforms like UnifiedInbox will be mission critical to the success of large-scale IoT initiatives.

RISC-V definitely has a lot more potential to disrupt edge computing industry. The next few years will be critical in defining how far RISC-V can go.


[Click Banner To Learn More]

[Home Page]  [The Automator]  [About]  [Subscribe ]  [Contact Us]


Want Ads

Our Sponsors