Review Mike Krell's RCR article for possible definitions/clarifications to add
Closed this issue · 2 comments
https://www.rcrwireless.com/20181217/analyst-angle/iot-edge-analyst-angle
What is the IoT edge? Where is the IoT edge? I want to put a stake in the sand on my definition of this oft misunderstood term.
Having just completed a podcast with Qualcomm discussing IoT edge computing (take a listen here), I was struck by our need to define the term “IoT edge” before we even started the discussion. Qualcomm is a key player in the edge computing ecosystem, and I’m an IoT analyst, consultant and pundit who spends lots of time talking about the IoT edge. If we have to come to an agreement on a common definition of the term and get common ground before we start a discussion, you know the industry still has a problem.
At James Brehm & Associates, I work and follow key players in the “edge IoT market”, including Qualcomm, Dell, Cisco, HPE, Cradlepoint, Juniper Networks, Nokia, Ericsson, AT&T, Sprint, T-Mobile, and many others. Between them, there is still a tremendous amount of confusion over the term “IoT edge”, and each of them is attacking the edge with a different approach. Each of these enterprises is looking towards future IoT networks (LTE CAT-NB1, LTE CAT-M1, 5G, LoRa, etc.) with the belief that edge computing is not only the key to IoT, but a way to derive new revenues from traditional networks and networking equipment.
They all talk and think about edge computing and the IoT edge differently. HPE talks about edge as the demarcation point where IT meets OT and is pushing to move full cloud computing on premises. Dell thinks about edge as a gateway where you can standardize both southbound and northbound traffic. Cisco talks about the fog and edge computing being a subset of fog. MNO’s such as Sprint and AT&T use the term multi-access edge computing to discuss moving compute closer to the customer and to more local, premises-based compute and connectivity points at the edge of the mobile network.
What about where the sensors and actuators actually take measurements and the measurements are turned into actionable information, I would call the edge. So again, what’s the IoT edge?
As part of my end-of-year thoughts on IoT, I want to put a stake in the ground over the term “the edge”.
One issue before we dive in, realize that “multi-access edge computing”, “fog computing” and “edge computing” are all similar but not quite the same – let’s look at those first.
So, what are the differences between MEC, fog and edge? Small, but distinctive.
Fog computing: Fog computing is a superset of both MEC and edge computing and covers mobile and wireline connectivity. Fog computing includes the intermediate layers between the edge and the cloud. Fog computing also includes both storage and deep packet networking.
Multi-access edge (MEC): Multi-access edge computing (MEC) is the application of cloud architectural principles to compute, storage and networking infrastructure at the edge of the mobile network.
Edge computing: The movement of compute, storage and networking outside of the traditional data center. In IoT terms this can be either represented by edge gateways connected to edge devices connected to the internet, or by edge devices connected directly to the Internet.
In my mind, the edge is not a specific place. I repeat, it is not a specific place. It is a term used to discuss the movement of what have recently been cloud functions, i.e. computation, storage and networking, out of the traditional data center, and closer to where the data is actually gathered. The physical location can be an enterprise premises, or it could be at the actual IoT end-point (sensor). What we do know is, it is not in the data center where the traditional cloud resides.
The critical aspect of the edge is not where it is physically located, but what it means and what the job of the edge is…and that is to move the 3 critical functions of IoT — compute, storage and connectivity, closer to where the data is both gathered and measured.
Why is this important? IoT is all about taking action on data.
As an example, the data may have a “shelf life” and action often needs to be taken now for it to have real value. Waiting for data to move from device to data center to be analyzed and back to device can cost time and money. This quick time to action is but one example of the value the IoT Edge Computing can provide. To support other IoT applications, the Edge has a certain set of characteristics that make it valuable.
To me, the IoT edge must have the following attributes:
Local. Edge computing can run isolated from the rest of the network allowing local functionality to continue if the external network goes down. This allows higher resiliency applications (such as machine-to-machine) to function, no matter the state of the network, and later move data off premises for storage or integration with other business functions.
Local to the source of data. Key information is gathered by devices. Edge computing allows the analysis and reaction to information closer to the actual source of that information allowing new business models and the leveraging of business specific applications.
Low latency. Running close to the device allows for faster reaction. Specifically enabling Artificial Intelligence and Machine Learning algorithms with real-time response will enable a wide variety of new business models and applications.
Secure. Edge computing embraces local communications of private networks for performance, security and privacy. Edge computing embraces local communications of private networks for performance, security and privacy.
Analytical. MEC differentiates itself in the ability to collect real-time information from data collected and processed at the point of capture with the minimum amount of bandwidth expended.
Location awareness. Edge computing knows the location of each specific connected device allowing new business models, optimization of business processes and the ability to monitor compliance on a real-time basis.
These characteristics define the IoT edge. Where is the IoT edge? To me, that’s just not important, it’s the understanding of how it works and its value that allows enterprises to derive the value, no matter where it’s located.
One last note on “why we care” about the IoT Edge. Our preliminary market estimates are that this segment of enterprise computing hardware is approximately $1.6B today, growing at 30% annually through 2023. That’s something to think about.
Okay, that’s my stake in the ground. Let me know your thoughts.
The two points that stand out to me from this definition that we don't cover are:
- 'Edge computing can run isolated from the rest of the network allowing local functionality to continue if the external network goes down'
- 'Edge computing knows the location of each specific connected device'
Though these are good for resilient and location-aware IoT use cases, they don't apply to many other edge use cases. Suggest if we want to cover them, a single entry on IoT edge requirements. @mtrifiro ?
Adding the following glossary entry:
IoT Edge Application
An IoT application which requires the use of device edge or infrastructure edge computing to operate, for performance, cost or functionality reasons. May require the ability to run isolated from the rest of the network, allowing local functionality to continue if a network outage occurs; and may also require the location information of each device.
See also: Edge-Native Application