The connected edge and why hyperscalers can’t make the cut

Deployment was cumbersome, updates and patches meant extensive downtime, and security was hard to establish and even harder to maintain. The result? Implementing software updates and hardware upgrades was avoided like the plague, and when you finally had no option but to do it, it required meticulous planning and coordination, as well as truck rolls and travel to remote locations for your tech crew.

And then came the cloud

The cloud was a game-changer. Now you could update your software as often as needed, keeping your customers happy with new and improved features hot out of R&D ovens.

Managing your IT became easier, and security got a significant boost. However, all this goodness came with a major catch – the cloud had a hard time living up to its promise of low and predictable computing and data transfer costs – especially as your business scaled up and out.

Enter: The edge

The managed edge holds the promise of extending the convenience of cloud computing, to on-prem computing, so essentially you get the best of both worlds. Software deployment, updates, maintenance, and scaling become simpler and easier to handle, yet compute stays closer to your users and your data, which means you get real-time insights and smarter decision-making right where you need it.

But, for on-prem to become a managed edge and truly fulfill its potential, it must be reliably connected to the cloud (at least part time) – a vital link that adds a layer of management complexity and cost.

So how do hyperscalers fail at the edge?

Let’s look at the cost of managed hardware. When deploying to the cloud we’re usually unaware of the physical hardware costs, but deploying on the edge can easily translate to hundreds of sites and thousands of compute devices. When multiplying these numbers by the cost of managed hardware configurations offered by Azure, AWS, and GCP the numbers will make your CFO cry. Take, for instance, a typical VMware edge architecture – it includes multiple different boxes for compute and networking, along with multiple management and orchestration software to control them all. You could try using it for a few dozen sites, but what happens when you need to scale up? Now, imagine implementing that architecture across a network of thousands of sites.

Let’s take it even further with the need for reliable managed connectivity. For a truly managed edge, connectivity is not as straightforward as relying on public internet. You need reliable, guaranteed, and secure connectivity that connects your edge compute to your cloud backbone. Hyperscalers simply take that as someone else’s problem that needs to be handled separately.

A concrete use case

Now, let’s put theory into practice with a real-life example. Think about a massive ready-mix concrete manufacturer with global operations. Their challenge was that each customer or construction project has their own special formula for the mix, and the mix has to be ready in time for the cement truck pickup. Faulty mixers downtime, formula quality issues and loss of connectivity all translate to significant revenue loss and unsatisfied customers.

So, the manufacturer’s goal was simple: ensure that the right mix was ready at the right time, at the right plant, and at the required quality.

To achieve this goal, the company required a two-pronged approach: enhance connectivity throughout their entire network and introduce predictive maintenance, and automated mix inspection across all sites.

For the infrastructure, initially they turned to the usual suspects: Microsoft Azure managed stack edge servers to host the local AI models and VMware’s VeloCloud SD-WAN for secured reliable networking.

They soon realized that for the magnitude of their operation the cost of managed servers would be daunting and that they would need to use two separate, unintegrated management and orchestration platforms, one for the SD-WAN network and one for the compute servers.

As an alternative they decided to go for a virtualized approach and looked for a solution that would enable them to deploy the AI models on cheaper white- box servers and to manage both on-prem network and compute from a single hub.  The selected solution provided them with an edge operating system that seamlessly runs both AI models and a virtual VeloCloud SD-WAN, with a single management and orchestration hub to rule them all.  And to sweeten the deal the costs were fractional compared to the hyperscalers alternative.

The manufacturer gained control of their network and simplified the complexity of managing thousands of edge compute devices across multiple sites. They were able to easily deploy and manage network functions and AI models from multiple providers, running on white-box hardware and supported by robust management and orchestration software. This resulted in reduced operational costs, and enhanced efficiency

Now, in this real-life example, the manufacturer understood two crucial needs: having a connected edge and solving the mess of managing multiple assets at the edge. As a bonus, by using white-box hardware they also reduced their TCO and carbon footprint.

The connected edge brings together the ease of the cloud with on-prem control. While achieving smooth and secure connectivity is crucial, hyperscalers, while excelling in cloud deployments, fall short at the large-scale edge. The connected edge doesn’t have to be complex and costly. Opting for a unified solution, that combines secure, robust connectivity with white-box flexibility, and simplifies the mess of managing at scale, proves pivotal in achieving a cost-effective and efficient edge.

About the author

Ziv Koren is Chief Product Officer for Telco Systems – the makers of Edgility – a platform that brings the cloud experience to the edge, enabling ISVs, SIs and IT to easily deploy, update and monitor business apps on the edge, bringing them closer to consumers, data and sensors.

DISCLAIMER: Guest posts are submitted content. The views expressed in this post are that of the author, and don’t necessarily reflect the views of Edge Industry Review (EdgeIR.com).

Article Topics

AI models  |  cloud computing  |  edge compute  |  edge servers  |  hyperscalers  |  Telco Systems