THE EMBEDDED ANTHOLOGY PART XV (EPILOGUE) – Edge Computing & IoT

It is no surprise that the connected objects are getting more intelligent and more autonomous, a trend that in turn has led to a huge growth of the edge IoT market, the total market value of which is expected to exceed $60 billion by 2028. This edge-fever has even led to the somehow exaggerated assumption that the days of cloud computing are counted.

Anyhow, maybe it is about time to think beyond cloud computing whose logic is being turned inside out by edge computing. In other words, while funneling everything through a centralized data center might be a smart model for scaling web search, social media, and media streaming, it’s not so smart for latency-intolerant applications like autonomous cars.

It is equally of great importance to reflect upon the implication of edge computing on IoT, when moving processing, storage, and analytics to the edge. How will this trend impact Cloud, IoT device vendors and OEMs?

The pros and cons of Edge Computing

In a cloud computing scenario, the connected device (or a network of the latter) gathers data and transfers that data to the cloud for processing and analytics. The powerful computing functions, such as the analysis of captured data is carried out by AI and machine learning algorithms residing on cloud platforms where the machines have the resources to complete these tasks. In such a scenario, with little to no processing taking place on the device itself, for an IoT device vendor, the spend on the physical device is relatively lower and the spend on cloud resources relatively higher.

However, edge computing by moving that processing away from the cloud and onto the device, while cutting the cloud platform investment, increases the spend on the physical device. Put differently, in such a scenario the IoT devices need to be more powerful and better resourced, in order to be able to store, process, analyze, and share data with cloud servers. Having said that, one should not forget that IoT devices that primarily rely on edge computing to perform, still maintain an important cloud connection for updating software, updating machine learning and pattern recognition algorithms, etc.

The benefits of Edge Computing for IoT

• Low latency and improved response times :With processing taking place on the device instead of in the cloud, response times are more rapid, and latency significantly reduced. For connected devices that rely on split-second feedback such as self-driving cars, this is essential for operating effectively.

• Offline availability: While cloud connected IoT devices offer incredible computing power, that powerful processing is only available for devices that are online. Where devices are deployed out of internet range or where networks go down, these devices are of little utility. Edge devices, on the other hand, have greater availability should networks go down or when deployed in remote locations.

• Improved security and privacy: With data storage and processing on the device instead of in the cloud, security is improved and the risk of losing data between the device and cloud is reduced. While there is still cloud connectivity, the amount of data transferred is diminished as is the security risk.

• Budget savings: Cloud computing is expensive and devices that rely on the cloud can be locked into contracts that represent a significant budget line for device vendors. Moving processing, analytics and storage to the edge reduces this cloud spend and delivers ongoing savings month-on-month.

The Drawbacks of Edge Computing for IoT

The main drawback of moving to the edge is hardware costs. Despite a significant decrease in the cost of IoT hardware, the extra processing power and storage requirements that edge computing demands means higher hardware costs. It is estimated that a typical IoT device capable of edge computing will cost between five and ten times that of the non-edge hardware. This is a significant up-front cost for vendors to pass onto their customers and has an impact on the go-to-market plans of those same vendors.

Though this up-front cost can be substantial, it needs to be balanced against the reduced costs of connectivity, data transfer, and cloud computing. For many vendors, this cost/benefit calculation makes the decision to compute on the edge profitable and it is this calculation that is helping to drive the growth in edge IoT devices today.

Edge Computing a must for latency-intolerant applications

Despite the above-mentioned cost-benefit analysis, the business case for edge computing on connected devices is quite strong for autonomous cars, remote monitoring of energy assets, medical technologies and patient monitoring, and predicative maintenance. To know why, let’s take the example of a self-driving car.

Autonomous vehicles rely on on-board computing power to make the split-second decisions that allow the vehicle to operate safely. The radar, lidar, sonar, GPS, odometry and inertial measurement sensors that enable vehicles to drive on public roads and in traffic demand processing that cannot rely on connectivity to cloud servers. The latency that comes with a cloud-powered solution – even if only measured in milliseconds – is too long for safe driving; all processing needs to be happening in real time on-board the vehicle. One has also to bear in mind that latency put aside, the system needs to remain functional in case of no connectivity. In other words, you need to make sure that every decision is taken on the edge without cloud dependency.

The same goes for remote maintenance, where you need to record / process data even if the connectivity is not fully available. In that case, it is essential to be able to re-sync once the cloud connectivity is re-established.

It results from the above, that edge computing which enables autonomous vehicles doesn’t mean there are no cloud connections at all. Indeed, manufacturers investing in self-driving technology typically rely on stable cloud connections such as home or office Wi-Fi to pool data from individual driving experiences and deliver updates over the air. Thus, while a self-driving car might primarily rely on edge computing to perform, it also maintains an important cloud connection for updating software, updating machine learning and pattern recognition algorithms, and improving the on-board experience of an individual car based on the pooled feedback from an entire fleet.

No matter if you are analyzing sensor or vibration data on a connected device or deploying a market-leading autonomous vehicle, edge computing is a great option for improving security, privacy, offline availability, latency, and response times. Thanks to AI accelerators, edge devices are now capable of running embedded deep learning applications at full scale more efficiently, effectively, and sustainably.

Conclusion

It would be wrong to assume that the days of cloud computing are over. It is of great importance to understand that cloud and edge computing are two different and non-interchangeable technologies, and each serve a specific purpose. Edge computing is used to process time-sensitive data, while cloud computing is used to process data that is not time-driven. Hence, cloud computing will still remain a crucial part of an organization’s IT infrastructure.

There are also hybrid scenarios where edge- and cloud computing work hand-in-hand. This is the case of IoT devices with computing power attached to them, along with cloud functionality. More precisely, the device-deployed code responds in real-time by shutting down the IoT machine in case of a damaging failure condition, while the rest of the application runs in the cloud.

As mentioned, edge computing is helpful in situations where one wishes to bypass the latency (or unstable or temporary unavailable connection) caused while communicating information from the device across the network to the cloud. Besides this latency concern, edge computing is also preferred over cloud computing in remote locations, where there is limited or no connectivity to a centralized location.

With edge devices increasingly affordable and with the edge market exploding in popularity, now is the time to embrace the next stage in IoT and connected technologies and “live on the edge”!

In case you’re contemplating to venture into the edge computing sphere and need help, let us know since at Witekio we cover all (full-stack) the edge-to-cloud embedded development services that you might require.

Embeddedly yours,

Cirus Coliai (BDM at Witekio for France, UK & Northern Europe)

#embeddedsoftware #IoTsoftware #softwaredevelopmentcompany