Cloud, What's Next?....</Edge\>

So What's Wrong With Cloud?

Cloud is a computing style — self-service, automated. However, cloud today is very much about computing and storage centralization in mega data centers, focused on economies of scale, massive scale and back-end processing. This is ideal when data — and its compute — can be centralized. Today's cloud solved a major problem — increasing flexible accessibility to computing and storage for businesses, especially for the back office. But the front office remained a browser or an app, talking to data centers far away with discrete requests and responses. Content delivery networks (CDNs) sprang up in order to cache data closer to users — but the paradigm is still about streaming data in one direction, from the center to the edge. The explosion of things and immersed people will shift the value from scale toward agility — where interactions will need to be much more immediate and truly interactive, where delay is unacceptable, and where the bulk of the data is being produced and needs to be analyzed and consumed at the edge.

There are four major problems with today's cloud that must be overcome:

1> latency

2> bandwidth

3> autonomy

4> privacy and security

When the Edge Grows, What Does the Centralized Cloud Do?

While the edge will eat the cloud, the centralized cloud will continue to have a role, although less in real time, and less central

• Coordination: Provisioning and updating both the edge and the things at the edge. Cloud computing has rapidly advanced technologies for automated provisioning and remote management. These technologies will be used to provision and manage the intelligent edge.
• Aggregation: Federating multiple "edges" (or parts of edges) together — for example, all retail stores for an enterprise.
• Archiving: Recording processed data or logs from the edge for archival or later processing.
• Machine learning: While machine learning and training will take place at the edge, longer-term learning, analysis and training models can be done centrally — with lessons pushed back out to the edge.
• Traditional back-office processing: Where real-time processing isn't critical, where scale matters more than agility, the central cloud will handle it.
• Fallback edge: Where a more intelligent, autonomous edge does not exist, the central cloud will take on that role, albeit less efficiently and with less interactive agility. As the edge grows, there will be a distinct gap between the haves and the have-nots (locations and enterprises with edge capability, and those that rely on remote connections to a centralized data centers).

Cloud-based businesses will need to change. 

Amazon's Whole Foods stores eventually include edge micro data centers for virtual stores?

Cloud computing vendors, of course, will not sit idly by while requirements and revenue opportunities move to the edge. While their expertise is in delivering massive IT capabilities with economies of scale, they can also extend their locations closer to the edge, and leverage their growing relationships with developers to extend their programming models toward the edge.

Delivering Centralized Content With Lower Latency — Essentially Cached — Closer to Users

CDNs have done this kind of work for years. 

Extending the Programming Model to Devices at the Edge

Amazon introduced its Amazon Web Services (AWS) IoT platform in late 2015, and AWS Greengrass in December 2016. These solutions are more about extending the programming model to things, but still puts the majority of computing, storage, analysis and machine learning in centralized AWS data centers. AWS is making it easier to connect intelligent things to the central cloud, but is not pushing "enough" of cloud capability close to the edge. It simply won't be enough.

With Azure IoT Edge (announced May 2017), Microsoft is talking much more about putting intelligence at the edge itself, although with limited capabilities. Azure Stack, which will come out later in 2017, can be seen as a super-edge to Azure, but much richer than most edge locations require. Microsoft's approach to edge computing is more open to a heterogeneous edge that isn't running in Azure data centers — but whether Microsoft can deliver a single, functionally scalable and subsettable architecture that fits any and all edge requirements and use cases is far from clear.

Who Will Build the Edge?

For enterprises, the growing requirements at the edge will turn strategies of simply "going to the cloud" upside down, and create completely new "business moments" at a low-latency edge. Rather than from the cloud-out, the edge will more likely grow from outside-in, based on what's already in place at the edge, with entrenched vendors and products, such as:

• Homes: Gaming consoles, cable set-top boxes, large appliances, Wi-Fi routers, home PCs, intelligent personal assistants
• Mobile devices: Smartphones, 5G mini-towers
• Entertainment locations (theaters, amusement parks): Micro data centers, edge servers, campus Wi-Fi
• Transportation: Processors in cars, airplanes
• Cities: 5G mini-towers, smart stoplights
• Workplaces: Campus Wi-Fi routers, enterprise data centers
• Manufacturing plants: Micro data centers, edge servers
• Military: Mobile micro data centers (cloudlets)

The diverse field of players will include enterprise IT, gaming vendors, telecom and cable operators, smartphone vendors, and consumer appliance vendors. Software vendors skilled at dynamic provisioning, machine learning and analytics will find rapidly growing opportunity at the edge. User interfaces will require massive investment as immersive technologies and ambient user experiences become the premier and eventually the predominant way that workers and consumers will want to interact with businesses and each other.

What Does Edge Computing Mean to Business?

Edge computing will be a necessary requirement for all digital businesses by 2022. Forty percent of large enterprises will be integrating edge computing principles into their 2021 projects, up from less than 1% in 2017. Successful businesses will identify and compete for smaller and more fleeting business moments that depend on ambient user experiences and immersive relationships with customers and things (capturing a moment when a customer "looks" at a thing with their phone, when they walk by a location, when a thing measures a change, etc.). IoT and more immersive, interactive UIs will drive one-third of large enterprises to create or use edge locations by 2021. Automation at the edge, together with machine learning, will be critical in customer relationships. Application architectures will need to factor in latency, location, federation and autonomy in their designs — and technologies and services to manage sprawling computing on the edge will thrive — as well as balancing what takes place on the edge, versus the cloud, and everything in between. Latency and/or bandwidth requirements will become critical application deployment factors for 30% of workload deployments by 2021. Competition for edge computing will be intense and confusing, as vendors from the consumer world (and newly connected things) collide with the enterprise computing world — and businesses should expect to place a number of bets. Over time, the "shape" of edge computing will settle into specific different, overlapping technologies and use cases — it's unlikely that one size will fit all, and some vendors won't survive the transition or new competition. A successful business will have edge computing strategies (it may have several, depending on its different uses of edge) by 2020. It's time to start thinking about the next fundamental IT architecture trend after cloud computing. The edge will flip the computing paradigm, pushing processing and storage to the edge.

And over time, the edge will eat the cloud.