Framing the Edge Discussion
by Peter Christy, Industry Analyst
Peter Christy is a consultant and former analyst who has been working with MobiledgeX and helped in the creation and development of Seamster.
As Jason Hoffman has explained in the overview piece, Seamster was created by MobiledgeX to promote an open community discussion of edge computing, and by doing so accelerate the formation and growth of a robust edge market. My purpose here is to give a little more insight into the details and structure of Seamster.
The MobiledgeX team has been looking for good edge application and service candidates for the last two years, building on the earlier work Deutsche Telekom had done. That effort made us understand the importance of how we thought about the edge — our “mental model”.
The purpose of a mental model is to facilitate thinking about an important, complicated topic (which the edge unquestionably is). A mental model is a crafted abstraction — it’s simpler than the real world so that the complexity doesn’t overwhelm, but at the same time it strives to be open (not obscure any important aspects) and fair (not advocate one market position over another). (If you’re curious, Farnam Street has published a wonderful on-line book on mental models.)
Here’s where we ended up thinking about the edge, and where the model in Seamster came from.
Seamster is built around Use Cases. A Use Case is a concrete example of how an application or service delivers value to the user, where the value is commercial; an example of how the application tangibly helps the user accomplish key goals, typically improving their business. In an early demo of the mobile edge, MobiledgeX showed how use of the mobile edge enabled Niantic to have more locally-clustered PokemonGo game players in a common game, for example.
Next, we need to be clear about what we mean by “edge computing.” How can we decide if and how a particular use case is enabled or improved by the use of edge computing?
Edge computing is all about improving the performance of applications and getting beyond the performance attainable with a typical cloud and device application architecture. The value of cloud-based IT is so strong that it’s easy to forget that moving local applications to the cloud comes at a very significant cost in performance: nothing can beat the performance of a device-based, local application. But there are many cases where putting elements of the application in the cloud is essential (e.g., to get more computing power or storage that is reasonable to have on the device, sharing data between many users, or incorporating cloud-based services). Edge computing offers the promise greatly improving the performance of edge-enabled, cloud-based applications and services.
Deciding to edge-enable an application is non-trivial. In almost all cases, taking advantage of edge services has to be explicitly designed into an application. There is no simple and automatic way to adapt an arbitrary cloud application to run at the edge. Running the code at the edge is easy; the hard part is making the application run much faster with edge resources; that's what has to be designed in. The designer must choose between a device application (the simplest with respect to performance), a device/cloud application (the most common today), and a device/edge/cloud application (for applications that want to leverage cloud-based services or resources but want performance much closer to device performance).
Although there are new application structures to learn, application and service builders who chose to create edge-adapted, performance-enhanced device/edge/cloud applications don’t have to invent everything from scratch. There are known design patterns already to use, and more are being discovered daily. We plan to share these on seamster.io.
Location, Location, Location
As with real estate, performance at the edge is inescapably related to the location of a device. The speed of electronic communication is limited by the speed of light. So a better term for “edge” computing is probably “location-based” computing, the point being it isn’t a question of just using edge services — they have to be correctly located as well.
The Four “Edges”
A lot of the early edge discussion seemed to be arguments about what is and what is not the “edge.” Our answer to that is to be inclusive, and say that all locations where network-aware applications are run are “edges” including the device and certainly the public cloud.
As discussed above, the focus of seamster, and we think the commercial edge discussion, are applications that run components on “location” optimized platforms, with a cloud/edge/device or edge/device architecture. But we want the Seamster discussion to include applications where the ultimate answer may be to just run them on the device or use an existing cloud/device architecture. We want Seamster to be a place where we can have that discussion and debate, and in doing so all learn more about the new, location optimized edges and where and how they should be used.
In our model, as it stands today, we identify four edges: the device, the cloud, the “premises” edge, and the Telco edge. Those are the four edges. To explain the last two — the ultimate focus of Seamster — we need first to add some more details about our mental model.
More Important Details — Connectivity and Business Model
An important element in the modern edge discussion is the increasing prominence of Internet and application access through service provided by the global mobile ecosystem. That ecosystem is very different from the cloud and Internet. The simple reasons are that mobile connectivity uses radio spectrum licenced by the local government and radio spectrum is bounded (a limited resource) and treated globally as a critical national security asset and governmental revenue source. Whereas much of the Internet has been “free,” mobile connectivity (beyond private WiFi) has always been a managed, for-fee service based on a contractual, regulated agreement with the subscriber. All of this will carry forward in Internet access, sometimes in a frustrating way (“why isn’t this as free and open as the Internet?”) but with a lot of positive new (to the Internet) attributes as well: it’s managed service with service guarantees, it is more secure and private (by design), and it is already well integrated into local and regional regulation (e.g. privacy). For here let’s just say that the differences are quite important to the edge discussion
So when you get to the edge and Telco-provided mobile connectivity is added to the mix, connectivity (licensed or unlicensed) and business model (owned or service) necessarily are part of the discussion.
The final two edges differentiate by the service model — private or public service (there are lots of little corner cases like a private cellular system but we will stick to the big picture here.
The Cloud Edge
Transition to the public cloud has dominated IT for the last decade. Use of the public cloud outsourced infrastructure acquisition and operation, and moves more application cost from CapEx to OpEx. Furthermore the hyperscale service providers have rapidly advanced the state of development tools and (maybe most importantly) operational automation. All of that has great value.
From our perspective, the edge should be like the public cloud in terms of development tools and advanced operational automation. The standard tools chains that work on the cloud should be directly leverageable and the automation even more important given the scale and scope of edge deployments. The point is only that although the location-based edges are materially different from the public cloud, there are lots of good things about the cloud that should be inherited.
The location-optimized edges are typically different. They won’t be dominated by a few services. Proprietary assets will be much more important at the edge as will the use of regulated communications services.
The Premises Edge
The third edge — the existing location optimised edge — is the Premises edge. Technically the premises edge is like the cloud (server computing). It differs because the resources are proprietary (possibly more secure and controlled as a result), and the business model often CapEx, not OpEx, with scaling and technology the responsibility of the user, not a service. Like the Telco-edge, the premises edge can be more secure and private than the public cloud.
The Telecom/5G Edge
The final edge — the newest — is the telecom/5G edge — services delivered within the global telecom/mobile infrastructure. It is usually a service, like the public cloud (although there are a growing number of private cellular systems with “premises” like values), it’s more secure and private than the public cloud (because of differences between the Internet and cellular system). And it leverages the massive trillion dollar existing investment which already has a100’s of thousands of “edge” computers in place (in contrast to the Internet).
Applying our Edge Mental Model to Use Cases
So that leads us to the final and most important aspect of our mental model: how we understand the relative merits of using the different edges. We’ve identified four key edges, differentiated by technology, connection mode and business model. The value analysis for a given choice of edge starts with communication quality (latency and bandwidth — the “physics”) but rarely stops there. Next we have to determine and think about how the physics results in a better and more valuable application — how does the application or service builder take advantage of the better quality connection? More importantly perhaps, how does an edge service enable use cases that just didn’t make work before? Finally we have to think about the business model. Is an OpEx service key to delivering the value (as it was for many smart phone applications)? How valuable is the increased security and privacy of a premises or Telco cellular edge location?
Tagging Use Cases
Finally, we’ve tried to make it easy to relate these Edge Use Cases to other trends and interests in modern IT. The edge is a way of improving performance but the application incorporates other important advances such as Machine Learning and Artificial Intelligence technologies and intersect other broad trends such as the Internet of Things. We’ve tagged the Use Cases with these other topics, as appropriate, in order to make searching more efficient for a visitor with a specific interest.
The hope for Seamster is that it will host a vibrant discussion about edge use cases that will help us all better understand which applications make the best use of each of these four edges.
Edge Visionaries Welcomed!
Take action and help drive the edge computing conversation inside Seamster. Use Seamster as an opportunity to engage with like-minded people, promote common interests, and leverage new findings. Your input could help drive next-generation experiences!