Cloud at the Edge: A Game Changer Waiting on Wings

What is Edge Computing?:- Edge Computing is an architecture in which tasks like data processing, analytics, ML inferencing etc. are performed much closer to the location where it is needed, in order to reduce latency and save on network bandwidth costs. A straightforward use case of Edge Computing is a real-time IoT application that needs to process huge amounts of data from sensors spread across a factory and send back instructions to Actuators or Applications on Edge devices.

Edge Computing and IoT:- Large Enterprises in Manufacturing, Energy and Utilities were the early adopters of Edge computing, primarily to solve challenges arising out of wide-spread IoT adoption. These Enterprises invested in self-hosted and/or self-managed IT infrastructure within their premises to run their edge computational tasks. One core component of this Edge Infra was an Edge Gateway that performed tasks like: storing data for a brief duration, queuing up events/messages, ML inferencing, enforcing policies etc. The Edge Gateway was also responsible for interfacing with applications hosted on Cloud for tasks like Permanent Data storage, Full-fledged analytics, Compute-intensive ML model training, Centralised Config management etc. This model worked well for IoT based use cases by allowing to deploy and manage latency sensitive IoT applications reliably at the Edge layer.

Advancements and Challenges in Edge Computing:- Rapid technology advancements in various fields led to Edge Computing being adopted beyond the precincts of IoT and opened up opportunities to use Edge Computing for several other use cases. The likes of Google, Amazon, NVIDIA and Intel produced highly capable and efficient hardware SoCs in tiny form factors that could easily fit into sensors, wearables, smart phones, bulbs and microwave ovens. Evolution of AI modelling and training techniques meant it became possible to even perform high computational ML model training on edge devices/gateways with minimal reliance on Cloud. Online gaming applications fuelled the need for real-time data processing and analysis with guaranteed sub-microseconds latency for users spread across the globe. Adoption of 5G technologies allowed for more network bandwidth with seamless mobility and roaming. These developments drove research and innovations at the edge computing layer and opened up unprecedented business opportunities for Enterprises. Naturally a self-hosted/self-managed Edge was no longer viable, simply because of the complexity and scale involved. A highly scalable, provider hosted and managed infrastructure platform was needed at the Edge, where much of the complexities are abstracted away by the provider through automation and orchestration.

The rise of Edge Computing as a Service:- Let's shift our attention to look at how the Hyper-scale public infra cloud providers (Amazon, Azure, Google and Alibaba) responded to these challenges and opportunities. They launched a slew of managed services primarily aimed at (i) providing bundled software solutions that could run on customer's edge infrastructure and inter-work with their public cloud services, (ii) a seamless centralized control plane for managing both the public cloud and edge cloud services. Some of these vendors developed their own specialized hardware SoC or even full-fledged Gateways and Devices for the Edge and made them available via partnership with hardware vendors. This reduced complexities at the Edge computing layer because customers were no longer required to build out and maintain the edge infrastructure all by themselves. At the same time, Telcos (like AT&T, Verizon) who sold networking solutions and bandwidth to enterprises, began to offer Edge services to those enterprises. Niche players (like StackPath) emerged, who partnered with Telcos and built out global Edge data centers offering compute resources on demand (much like a public cloud provider). Enterprises who had compelling needs to scale their Edge computing layer or who wanted a cloud-like usage model, subscribed to these services and deployed their edge workloads. This approach reduced complexities and enabled enterprises to scale their consumption on-demand. In short, they began to implement and consume Edge computing services, in ways very similar to consuming public cloud services.

Edge Computing at Hyper-scale proportions:- Speaking of Edge, all the hyper-scale cloud providers were offering Content Delivery Network (CDN) services much before large scale Edge computing adoption. CDN essentially started off as a caching service to allow web pages to be served from a location very close to the user who requested the content. The cloud providers often rented out data centers from Telcos or built their own facilities to host the caching servers. This also allowed the cloud providers to soon offer high-speed interconnect services for customers who required a low-latency connectivity between their data centers and cloud. The Cloud providers began to expand the number of Edge locations and at the same time recognized the immense opportunities with Edge computing, especially with the imminent large scale roll-out of 5G. The next evolutionary step was to open up these Edge locations to customers to run their own edge workloads, so customers have one uniform experience in consuming both cloud and edge services from the same provider. This way, enterprises could drastically reduce setting up and maintaining Edge infra of their own and rather choose to run their edge workloads on the Edge Cloud of their favourite Cloud Service Provider. In short, this brings Cloud to the Edge. In due course, more Edge locations would get added to the Edge Cloud, bringing affordable Edge Computing capabilities very closer to Enterprises. However, for this model to succeed at a global scale, the Cloud service providers need to partner closely with Telcos simply because Enterprises buy network and bandwidth from them.

Edge Computing and 5G:- Earlier this month, Google announced "Anthos for Telecom", which will bring its Anthos cloud application platform to the network edge, allowing telecommunications companies to run their applications wherever it makes the most sense. Though finer details of this offering are yet to be revealed, it is Google's answer to bringing Cloud to the Edge, starting first with the Telecommunications Industry. Telcos looking to capitalise on 5G's Multi-access Edge Cloud (MEC) would be enticed to leverage "Anthos for Telecom" for realizing the full potential of their Edge applications. Perhaps Google's long-term intention is to scale out and scale up their Edge Cloud across the globe, allowing a variety of Enterprises across Industries to run complex and sophisticated Edge applications at scale. AWS and Microsoft have also announced similar offerings and partnerships.

Conclusion:- IoT has been the prominent use case for wide spread adoption of Edge Computing and it's evolution. However, the future of Edge Computing will be shaped by other disruptive trends as well: 5G and AI/ML. As 5G roll out gathers steam and more AI/ML models get trained and deployed at the Edge, Edge Computing is poised for rapid adoption and technological advancements. And much of the future Edge Computing market will be dominated by the Hyper-scale public cloud providers and Telcos as Cloud at the Edge becomes the primary consumption model for Edge Computing. It is certainly a major game changer waiting on wings.

References:-

1. https://www.gartner.com/en/doc/3889058-the-edge-completes-the-cloud-a-gartner-trend-insight-report
2. https://towardsdatascience.com/why-machine-learning-on-the-edge-92fac32105e6
3. https://cloud.google.com/blog/topics/inside-google-cloud/google-cloud-unveils-strategy-telecommunications-industry
4. https://news.microsoft.com/2019/11/26/att-integrating-5g-with-microsoft-cloud-to-enable-next-generation-solutions-on-the-edge/
5. https://www.forbes.com/sites/moorinsights/2019/12/10/aws-wavelength-could-supercharge-5g
6. https://www.information-age.com/edge-computing-game-changer-service-providers-123485783/