Enterprises are looking to adopt technologies in which they can move data and applications closer to distributed users, IoT devices and remote sites.
These technologies include a complex mix of terms: edge computing, cloud computing, fog computing, cloud edge and multi-access edge computing (MEC), among others. Each strategy differs in where it hosts resources and applications -- for example, in the cloud, at edge locations or somewhere in between. But they all essentially aim to place those resources closer to the users who need them.
Enterprises were adopting these strategies prior to the COVID-19 pandemic, largely to improve processing times for real-time and business-critical applications. But the pandemic exacerbated many of the problems found with traditional infrastructure.
For example, as remote work became the norm, networks quickly needed to support distributed users and locations. But traditional network designs that backhaul traffic to a central data center add latency that could be detrimental for business applications.
As a result, enterprise interest in edge computing and the cloud edge has grown steadily. According to IDC, the MEC market is expected to grow from $3.5 billion in 2020 to $16.7 billion in 2025, with a compound annual growth rate of 37%.
During a virtual session at the ONUG Fall 2021 conference, panelists discussed how cloud edge use cases have evolved. Use cases included robotics, autonomous processing, pharmaceuticals and supporting the developer experience.
Compare how edge cloud -- or cloud edge -- differs from cloud computing and edge computing.
Robotics and autonomous processing
Since the pandemic, FedEx has felt the effects of the worker shortage, said Preetha Vijayakumar, vice president of data center, network and collaboration services at FedEx. The shipping company delivers millions of packages each day, but it has struggled to hire people to process those packages.
Preetha Vijayakumar
Customer demands haven't decreased, though. In fact, more than ever, consumers increasingly expect free shipping. But, Vijayakumar said, "no shipping is free." Year after year, FedEx experienced a loss in profit per package as it tried to accommodate those demands. To try to reduce profit loss, FedEx implemented robotics, automated forklifts and drones to deliver packages more efficiently and reduce Opex.
Robotics require low latency and real-time data processing, and they can't backhaul data to a central data center whenever they need to make decisions. To support its autonomous initiatives, FedEx uses an agnostic multi-cloud edge strategy that doesn't lock the company into a single provider, Vijayakumar said.
"We don't want to be tied to any particular carrier and their MEC offerings," she said. "Our sites can't go down, so we need carrier diversity and path diversity."
Additionally, FedEx locations rely mainly on compute stacks. So, if the network goes down at a site, that location can still sort packages, she added.
Connectivity and security
But networks remain an important component of cloud edge environments, and connectivity shouldn't be an afterthought, said James Devine, vice president of product management at cloud networking vendor Aviatrix.
James Devine
One connectivity option he cited was 5G, which touts increased reliability and lower latency for connected devices. For example, 5G could enable IoT use cases in manufacturing, robotics, healthcare and transportation -- most of which also benefit from edge architecture. But 5G is still proving its relevance, Devine said.
"5G is an interesting, evolving space," he said. "The technology is certainly there, but at this point, it looks like a solution looking for a problem."
Guruprasad Ramamoorthy
While 5G offers the potential to provide better connectivity to edge environments, security stacks also need to move to the edge, said Guruprasad Ramamoorthy, global head of network architecture, engineering and operations for S&P Global India.
"Data is going to the edge, and customers are at the edge," he said. "One of the fundamental building blocks after all this change is building the network access around the edge."
Zero-trust network access (ZTNA) is one way enterprises can secure edge environments, he said. ZTNA assumes no user or device requesting access can be trusted. Secure Access Service Edge is another emerging architecture that has the potential to support distributed users and enable secure access.
The developer experience
Both Ramamoorthy and Vijayakumar said the developer experience was an important cloud edge use case for their companies.
When remote work became prevalent during the pandemic, S&P had to support its remote developers who continued to push daily code deployments. If the developers experienced interruptions in their continuous integration/continuous delivery (CI/CD) pipeline, it would affect S&P's customers who need that daily data to make market decisions.
"People need to keep the developer connected to the CI/CD pipeline so they can deploy the code at the edge with smaller internet bandwidth," Ramamoorthy said. "Prior to COVID, they were using big office internet with MPLS. That's a completely different experience."
With a cloud edge strategy, developers can more readily access the resources they require and deploy changes to distributed customers at their own edge locations. FedEx's Vijayakumar also stressed the importance of ensuring developers have a reliable experience as they deploy applications to the edge.
"The developers don't want to know whether you're running in an edge location or in an actual data center," she said. "They just want to be able to use the same methods to do it."
Network engineers don't want to be out configuring nodes at every edge location.
James DevineVice president of product management, Aviatrix
Automation and pharmaceuticals
An ideal developer experience requires a flexible network, Ramamoorthy said. Instead of spending time on manual network configuration, enterprises should look at how automation and infrastructure as code can manage and support configuration at scale.
Devine agreed, noting how many network teams are stuck in "old-school thinking," in which they deploy routers and manage them in the data center. But that's not feasible for most edge deployments.
"Network engineers don't want to be out configuring nodes at every edge location," he said. "You might have edge locations thousands of miles away, where you can't just go in and change things."
That means enterprises need ways to consistently operate, manage and secure their edge locations and the required connectivity. Automation is one option.
Jeff Gray
Jeff Gray, CEO of network automation company Gluware, has seen how automation and cloud edge work together for Gluware's pharmaceutical customers.
"Each pharma shop floor is its own mini edge or data center," Gray said. "But it's not just a matter of compute. It's a matter of the network, core and distribution, IoT and robotic capabilities."
With so many variables, these customers realized the importance of creating an end-to-end architecture that connects compute, routing, switching, security and patching. Automation could help "keep the environment clean," in case of a power surge or zero-day attack, he said.
Where to start with cloud edge
So, how do enterprises get started with their cloud edge strategies? S&P Global's Ramamoorthy said enterprises need to understand where their customers are located and how to best serve them. For example, evaluate the geographical regions in which the company should maintain an edge presence.
Devine noted that even the most well-funded enterprises might not have the budget to deliver hardware to each edge location. But, he said, the commoditization of edge computing and cloud computing can help enterprises achieve a uniform operational platform and consistent edge deployments.
Meanwhile, Vijayakumar emphasized the importance of starting now.
"If you wait for use cases and someone finds one, it will take you five years to get to where you're able to support it," she said. "Start now. Don't wait."
Preetha Vijayakumar
James Devine
Guruprasad Ramamoorthy
Jeff Gray
