Edge computing is a distributed computing paradigm bringing compute, storage, and applications closer to where users, facilities, and connected things generate, consume, and/or leverage data.
Edge computing is already used in ample applications, mainly in combination with IoT (the Internet of Things). Although edge computing has been around for some decades, various types of IoT applications that require low latency, rapid analysis and/or fast action (intelligence shifting to the edge, including real-time decisions), and can’t (wait to) be processed in the cloud for different possible reasons have been driving the attention for edge computing in recent years.
Edge computing addresses the limitations of centralized computing (such as latency, bandwidth, data privacy and autonomy) by moving processing closer to the source of data generation, “things” and users (Gartner)
However, edge computing is about more than IoT alone and concerns IT deployment in a connected and data-intensive world in general, among others for mission-critical applications and customer-facing situations where speed and uptime are crucial. The edge IT deployment model is complementary to cloud, overcomes some of the challenges encountered with cloud computing, and includes data centers, physical infrastructure, and software, as used in various sectors that already can benefit from an edge approach today.
Edge computing is more ‘deployed in particular markets and segments because the use cases to leverage it with a proven ROI, benefit, and/or need already exist. Many of the future use cases that will need edge computing and are most often mentioned, on the other hand, don’t exist yet, certainly not at scale.
The usage of edge computing in specific markets and use cases goes hand in hand with vendor strategies whereby edge solutions and services are sold for applications that can directly benefit from it. Today, these market segments and use cases are often divided into three categories, depending on the types of locations and applications.
In the context of applications and use cases for the edge, some vendors distinguish between commercial applications (e.g., retail chains), industrial applications (oil & gas, smart manufacturing, also called the industrial edge, in the sphere of Industrial IoT) and the telco market which is perhaps the most promising but also still in full development (whereby 5G is an important factor).
Others distinguish between telco, the industrial/enterprise edge with a focus on IoT and remote facilities/offices/locations (comparable to commercial).
What You Will Learn
What is edge computing?
The precise definition of edge computing and where ‘the edge’ actually is (it is a much-debated topic) depends on whom you talk to. Ample vendors, organized in consortia with leading research institutions, and working on topics such as standardization and reference architectures, come with different views and definitions.
The reason for the current confusion that arises as a result: edge computing is big business and its full potential lies in the future. Vendors of all sorts are fully preparing to get a major share of the edge computing market as it develops. One could say that, for once, the market of solutions for the edge is ahead of the actual ‘state of edge computing’ in the field.
There are ample definitions of edge computing, but they’re not always clear. The same goes for the edge whereby some go as far as to say that edge (computing) is everything that is not in the cloud. In an older article on edge computing and IoT, we mentioned the definition of edge computing according to Gartner: “Edge computing is part of a distributed computing topology where information processing is located close to the edge, where things and people produce or consume that information”.
Edge refers to enterprise-hardened servers and appliances that are not in core datacenters. This includes server rooms, servers in the field, cell towers, and smaller datacenters located regionally and remotely for faster response times (IDC)
Most edge computing overviews also point to a 2018 Gartner prediction which we mentioned in the same article that by 2025, 75 percent of enterprise-generated data will be created and processed outside a traditional centralized data center or cloud (instead of 10 percent at that time).
In ‘The Digitization of the World – From Edge to Core,’ a white paper (PDF opens) and series of research-based material which IDC presented end 2018 (with Seagate) under the name ‘Data Age 2025‘, the interplay of edge computing and cloud computing is well explained (more below). Do note that as we update this edge computing article, end 2019, most workloads and compute aren’t happening at the edge at all. On the contrary: even a lot of IoT data are still not processed/stored in the cloud but in a company’s data center.
However, things are changing with cloud becoming the core and the fastest growth for data creation happening at the core and the edge. At the center of it all: in these days of ongoing digitization and digital transformation across industries, we have more sources and volumes of data. More importantly, however, there is an increasing need to leverage (analyze) data faster in ample use cases and scenarios where such makes sense and/or where customers/consumers demand it, as mentioned previously.
And sometimes, this means processing the data closer to its source, especially in distributed types of applications and scenarios. So, it’s also simply a return to a decentralized model (edge), in combination with a centralized one (cloud and its large data centers). The types of shifts we’ve seen many times before in IT: cycles.
Why edge computing?
With the increasing need for speed and low latency, ever more data and new use cases where edge computing makes sense we’re closer to the benefits of edge and examples of how edge computing is used. Let’s look at one.
There are many drivers for edge computing, other than sheer data-related and IoT-related ones. Moreover, most of the so-called next-generation applications that will really need extremely low latency and extremely high availability are not here yet. And it’s not certain if they will be here anytime soon, even if 5G is what many of them are waiting for. There simply isn’t any guarantee that things like autonomous vehicles or virtual and augmented reality at scale will indeed become a reality soon. In fact, even if 5G is really here there’s totally no guarantee that true self-driving cars will ever be a reality except in specific areas; there is far more to it than meets the eye. VR and AR is another story.
Vendors of edge computing equipment of course aren’t waiting because of the earlier-mentioned use cases and scenarios in which edge already makes sense and since it’s not certain what use cases we’ll see. In fact, just as there are use cases we probably won’t see happening at all, others will pop up that we don’t think about today. Yet, that doesn’t change the case for edge computing overall. And there’s quite some low-hanging fruit already in customer-facing areas such as retail and commercial applications where edge computing vendors are focusing on.
Let’s explain, including an example in the commercial sphere which was briefly mentioned in an interview on the evolutions in retail facilities: fast-food chains (with in-store retail being a big focus in edge computing now).
With fast-food chains we have one of many examples in retail and customer-facing, distributed, enterprise initiatives. As you know, many companies have started automating and digitalizing processes, including these customer-facing ones, thereby relying more on digital applications that become more mission-critical. However, in decentralized scenarios where the optimization of customer experience and cost savings, for instance, led to innovation, other issues arise, and the limits of cloud became visible. And here edge computing is positioned as the answer.
The example of the fast-food chain. Not that long ago, ordering something in a fast-food restaurant was a matter of standing in a queue until it was your time and checking you effectively received what you ordered. Today, you have kiosks close to everywhere. Less staff needed, an increased chance that you get what you ordered and so forth. For fast-food chains, this, however, means that these kiosks and the systems behind them have become more mission-critical. Vendors are positioning commercial edge computing solutions here, with, among others, so-called micro data centers and a potential centralized, cloud-based management service.
Benefits of edge computing
Let’s summarize some benefits of edge computing and where it can be applied: use cases where speed is essential (low latency), distance to a data center is an issue, there are applications that need to be a bit more secured and need to run constantly in customer-facing situations (regardless of a connection), the list goes on.
Many also tout lower costs and fast deployment time, for instance when deploying small micro data centers in those environments such as retail stores or fast-food restaurants where you also can save space and typically don’t have too many IT people around.
There’s scalability (e.g. across various dispersed locations, of course if you use compatible gear and can deploy and grow fast without compatibility issues), faster response times are especially important for some really critical applications, there’s the need to keep things – critical things – running in difficult and remote environments, typically what you’ll have with the industrial edge. And with customer-facing edge deployments we of course shouldn’t forget the customer experience. Finally, cost savings are also realized on a level of people, like it or not – remember the fast-food chain.
Where is the edge?
Now, where is the edge? Not that easy to tell and many definitions. And, again, it depends on whom you ask but also on the type of application and environment. You can imagine that the edge in the example of the fast-food restaurant that’s part of a chain, looks different than the industrial edge.
With the industrial edge, we’re often in remote areas, which are also further away from (regional) data centers with lots of devices in the field (think oil & gas, for example). Moreover, what do you precisely define as being part of the edge? In the fast-food restaurant, whereby a cloud application is used for remote monitoring of the various IT edge systems in all restaurants, you could say that everything is at the edge – close to the consumer.
An Edge location is a computing enclosure/ space/ facility geographically dispersed to be physically closer to the point of origin of data or a user base. In other words, for an Edge to exist there must be a hub or a core; therefore, dispersion of computing to the periphery would qualify as ‘Edge computing’ and the physical enclosure/ space/ facility can be defined as the ‘Edge facility’ (Smarak Bhuyan)
If you start digging deeper and go more in detail regarding the precise devices and so forth, in, for example, oil & gas, you can get lost in semantic details on what is precisely edge and not though. And the exact same thing goes for telco.
So, it’s time for standards and reference architectures. Here as well, a lot is going on, but again with multiple players and approaches. Did we say that the edge computing market is on the agenda of many vendors?
Sometimes you will also read that edge computing will replace cloud computing. That is simply nonsense since edge computing needs the cloud, among others, to have that visibility you need when things get distributed, for instance And there simply isn’t a business case to use edge computing everywhere. Moreover, it’s not as if edge computing doesn’t have challenges. Some of the touted benefits, for instance, regarding security, at the same time, are also still issues on other levels. And let’s not forget compatibility woes. So, the edge computing story certainly isn’t finished yet. In fact, it still really has to start.
The edge, endpoints and the core – cloud
A few more words on where the edge is. You could, for instance, follow IDC’s view.
In the before-mentioned paper, IDC distinguishes between:
- core (so, cloud, with colocation, hyperscale data centers, enterprise operational data centers and all forms of cloud computing),
- edge, being the enterprise-hardened servers and appliances that are not in core datacenters and
- endpoints which include the devices on the edge of the network such as sensors and wearables.
The overall trend is that data generation increases on the level of cloud which IDC sees as the core and of course at the edge with edge computing being that decentralized paradigm that drives digital transformation across ample use cases, especially in the coming years. And, as said, cloud, which is complementary, is still doing very fine for a long time to come too. According to IDC, in 2025 nearly 30 percent of data across the globe will need real-time processing with the role of edge continuing to grow.
Summarizing it all a bit, the graphic below from IDC shows the place and interplay of the endpoints, the edge and the core in the view of IDC. Note the presence of branch offices, cell towers and gateways in the edge part which again shows the three categories or clusters of applications we see.
The place and interplay of the endpoints the edge and the core according to IDC – source IDC Datasphere whitepaper – download here (PDF opens)
Edge computing and the data center market
Within the edge layer, we find server rooms, servers in the field, cell towers, and smaller data centers located regionally and remotely for faster response times and this is also where you find micro data centers for the edge or edge data centers which are often, well, ‘micro’.
Some vendors distinguish between ‘regional edge’ as the large compute and storage in central or urban areas (closer to the user/consumer than large central data centers but still) and local edge as those smaller micro data centers where data is generated and consumed.
Edge computing obviously has an essential impact on the data center market. And that brings us back to those previously mentioned cycles or paradigm shifts in computing. Think about the shifts in computing paradigms from mainframes to the client-server model and then to the more centralized cloud model again with colocation and – since late 2019 – over 500 hyperscale data centers.
Edge computing is a foundational enabler for several emerging technologies such as 5G mobile communications, which will allow IoT and other edge devices to take advantage of faster connectivity to data and compute resources with single-digit-millisecond network latency (Equinix)
As Jeff Groat, from US-based building system integration company Wadsworth Solutions that is upping its activities in data centers and micro data centers with edge and 5G in mind put it: the pendulum had to swing one way with everyone going in the direction of mega data centers, and now edge or micro data centers becoming vital, certainly with 5G on the horizon, as the pendulum starts to swing in the other direction. One can also say that the pendulum swings in two directions at the same time, especially with hyperscale data centers continuing to increase. Again, it will be a mix: edge and cloud.
Now that we mention 5G again, it’s perhaps a good moment to emphasize that edge computing isn’t just an enabler of 5G, Industry 4.0, industrial transformation and industrial IoT but also of IoT in other areas, of artificial intelligence and most certainly of digital twins and augmented reality, to name a few.
From a data storage perspective, the core (cloud) is expected to become the main repository with more than double the data stored in the core than in the endpoint by 2024. Moreover, edge storage is predicted to see significant growth as latency-sensitive services and applications proliferate per IDC.
Among the many edge computing definitions, here is one from Infrastructure Mason Smarak Bhuyan, Machine Learning Infrastructure PM at Google, that might appeal to the infrastructure and data center people: “An Edge location is a computing enclosure/ space/ facility geographically dispersed to be physically closer to the point of origin of data or a user base. In other words, for an Edge to exist there must be a hub or a core; therefore, dispersion of computing to the periphery would qualify as ‘Edge computing’ and the physical enclosure/ space/ facility can be defined as the ‘Edge facility’.” More here.
Edge and data center outlook
In its ‘FutureScape: Worldwide Datacenter 2020 Predictions’, IDC said that it expects that by 2023, more than half of new data center infrastructure deployed will be in increasingly critical edge locations rather than corporate data centers, up from less than 10 percent today.
Interconnection and data center company Equinix that cites the prediction for its ‘Top 5 Technology Trends to Impact the Digital Infrastructure Landscape in 2020’, adds that by 2024, the number of apps at the edge will increase 800 percent.
By 2023, more than 50% of new infrastructure deployed will be in increasingly critical edge locations rather than corporate datacenters, up from less than 10% today (IDC)
Distributed infrastructure and edge computing will accelerate hybrid multicloud adoption, Equinix says, expecting this to be the case across every business segment in 2020.
Referring back to IDC (you can listen to a replay of all FutureScape 2020 webinars from the research company here), Equinix points out that businesses must modernize IT to become virtualized, containerized and software-defined to support the edge and should consider new data center partners that can bolster edge build-out and prioritize infrastructure optimization and application communication costs.
Two more edge-related data center predictions from the same IDC FutureScape webinar: 1) by 2023, 75 percent of major edge IT sites will leverage ML and AI-enabled controls to transform maintenance (predictive maintenance indeed) and improve the efficient use of energy resources and 2) by 2024, over 75 percent of infrastructure in edge locations will be consumed/operated via an as a service model.