As early as 2017 Wired magazine’s Jeremy Hsu was warning the technology industry that it was time to think beyond cloud computing. Edge computing, he wrote, turns the logic of today’s cloud inside out. 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”.
Two years ago, Witekio’s own Kevin Sibué wrote that edge computing would help deliver the connected objects of tomorrow that would be more intelligent and more autonomous than those of the present. In 2021 edge computing is delivering on this vision with the edge IoT market enjoying compound annual growth rates above 35% and expanding to a total market value that is expected to exceed $60 billion by 2028.
But what does edge computing on IoT look like in 2021? What are the benefits of moving processing, storage, and analytics to the edge? What is the role for the cloud in an era of edge computing? And what are the impacts of the edge computing trend for IoT device vendors and OEMs?
What is Edge Computing on IoT?
Before moving to the benefits and drawbacks of edge computing for IoT, it’s worth spending a few moments defining exactly what edge computing is.
To understand the revolution that is edge computing, it is useful to first understand how IoT devices work outside of the edge. Typically, a connected device or network of devices gathers data and transfers that data to the cloud for processing and analytics. The powerful computing functions, including the analysis of captured data using artificial intelligence and machine learning algorithms, takes place on cloud platforms where the machines have the resources to complete the tasks. For an IoT device vendor, the spend on the physical device is relatively lower and the spend on cloud resources relatively higher, with little to no processing taking place on the device itself.
Edge computing moves that processing away from the cloud and onto the device. More powerful and better resourced IoT devices can store, process, analyze, and share data with cloud servers. Still connected to the cloud, the processing shifts to the devices which cuts the cloud platform investment required. At the same time, the edge IoT devices offer low latency and improved response times, offline availability that cloud-reliant devices cannot, and improved security and privacy.
Edge computing is integral to emerging technologies including autonomous cars, remote monitoring of energy assets, medical technologies and patient monitoring, and predicative maintenance.
So why are these industries turning away from the cloud and towards the edge?
The Benefits of Edge Computing for IoT
There are several advantages of moving to the edge for IoT vendors.+
1/ Low latency and improved response times
2/ 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.
3/ 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.
4/ 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 cloud reduces this cloud spend and delivers ongoing savings month-on-month.
The Drawback of Edge Computing for IoT
The advantages of IoT computing on the edge are clear and compelling but there is at least one significant drawback that should be considered before moving to the edge: hardware costs.
In recent years the cost of IoT hardware has dropped dramatically. Sensors that might have cost $1.30 each in 2004 could be bought fifteen years later for just $0.44. That price has continued to drop by between 10% and 20% year on year and has driven the deployment of billions of IoT devices worldwide.
Yet the extra processing power and storage requirements that edge computing demands means hardware costs necessarily rise, too. Witekio engineers estimate 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 oto 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 in Action
If the business case for edge computing on connected devices is strong and the trend towards adopting edge computing on IoT is clear, what does this edge computing look like in action?
A good example is 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.
Of course, the edge computing that enable 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. 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: the cutting edge today is, in fact, the edge.
Edge computing, Edge IoT, why it matters?