8 MINS READ
Highlights
To achieve superior checkout experience, the store infrastructure must be equipped for heavy processing and handling large volumes of data transfers to central databases. This paper provides insights on how to drive next-gen experiences with edge computing without spending a fortune.
Insights in this article
Insights in this articleinpage
INTRODUCTION
The evolved customers of today expect a far better in-store experience than before:
65% would like to scan items and see product information on phones, 50% want contactless payment options, and 20% are ready to use augmented reality magic mirrors to try on clothing.1 Such advanced and immersive technologies powered by the internet of things (IoT), sensors, and analytics can help retailers bridge the gap between in-store and online experience while providing troves of real-time data to drive business outcomes such as improving customer lifetime value (CLV) and increasing sales and conversions. But many retailers are unable to seize these opportunities as they struggle with the fundamentals of checkout experience such as performance and availability issues resulting in long wait times, and lack of online like digital capabilities such as product locator, rich product content, and in-store personalization.
To achieve a rich, high-performing, and uninterrupted in-store checkout experience, the store infrastructure must be equipped for heavy processing and handling large volumes of data transfers to the central instance.
This white paper discusses how edge computing can give retailers a strategic advantage in the highly competitive retail landscape in enabling a rich in-store experience. It also explains how retailers can realize edge computing for checkout with lightweight, open-source technologies.
CHALLENGES WITH LEGACY SYSTEMS AND TRANSFORMATION WITH CLOUD AND MICROSERVICES
For decades, retailers were relying on monolithic and store-centric approach for in-store checkout, which forced each store to act as a data center having primary and failover servers.
Besides impacting the time to market (agility) of new features, it resulted in a heavy and complex store hardware footprint, requiring high manual intervention and maintenance costs, and also special tools and software not just to distribute and deploy new releases but also to monitor infrastructure across the retail estate. Also, new releases required heavy change management and cutover activities at stores, impacting the time to market.
To simplify store infrastructure, reduce the high upfront infrastructure and maintenance costs, and be agile, several retailers have shifted to modern architecture with microservices and cloud technologies that can be scaled on demand and offer high availability with near-zero maintenance needs. But cloud architecture poses a fresh set of challenges, primarily around performance caused by network latencies between stores and the cloud, and availability. This has a direct bearing on customer experience, brand value, and profits of the retailer.
The key limitations of cloud architecture for in-store are:
Resiliency: Downtime due to service and network availability
Performance: Network latency, causing speed issues
Cost: High stress on network and bandwidth needs
Digital: Lack of digital capabilities at stores limiting online-like rich, personalized shopping experience at stores
NEXT-GEN CHECKOUT EXPERIENCE WITH EDGE COMPUTING
To overcome the above challenges and deliver on the customer expectations for performance, availability, capabilities, and flexibility, data and processing must be closer to the edge (stores).
Leveraging edge computing as a complementing technology to cloud computing not only saves time and money, but also solves latency, bandwidth, autonomy or compliance issues (see Figure 1).
With edge computing, the central cloud capabilities are extended to the edge by leveraging a central container-based orchestration platform that can seamlessly manage the edge nodes (network and storage equipment), and, thereby, software rollout/upgrades and associated configurations. Edge computing can handle applications and workloads for thousands of store locations by running computational power through nearby edge nodes rather than risking data transfer speed and bandwidth challenges by accessing all services and data directly from the cloud.
With edge computing, retailers can realize scalable checkout service that can support multiple channels and also deliver next-gen in-store experiences (see Figure 2) with key capabilities such as rich product content, visual search, voice assistants, contextual recommendations, fraud detection using computer vision, and cashier-less autonomous checkout.
EDGE COMPUTING WITH LIGHTWEIGHT, OPEN SOURCE FRAMEWORKS
The topology and architecture of edge computing provides a framework for IoT and enterprise architectures to converge.
This can further be extended to enable a high-performing, personalized digital checkout experience at stores with resiliency. Though there are a few edge solutions such as AWS Outposts and Snowcone (Amazon), and Azure Stack Edge and Azure Arc (Microsoft), these are all more focussed on the IoT edge architecture and require proprietary hardware and software platforms, resulting in high costs and vendor lock-in.
KubeEdge, which is a scalable, flexible, and lightweight open-source technology that can support heterogeneous platforms is a good alternative to the proprietary high-cost solutions for edge computing. It can extend native container orchestration (Kubernetes) and device management to the edge and provide core infrastructure support for networking, application deployment, and metadata synchronization between cloud and edge.
When evaluating edge architecture options, retailers must consider several factors:
Can the cloud capability be seamlessly extended to the edge?
Does it enable centralized app distribution and management?
Is the edge architecture scalable to thousands of store locations?
Can services be deployed, upgraded, and rolled back seamlessly?
Does it have built-in monitoring and self-healing capabilities?
Does it require proprietary hardware or software?
Can existing store hardware (virtual machines or the physical ones) be attached to the edge cluster to act as the edge gateway?
IMPLEMENTING EDGE COMPUTING USING KUBEEDGE
In a typical microservices and cloud architecture, a container orchestration platform such as Kubernetes (K8s) is used to centrally deploy, manage, and monitor the services as containerized apps.
The steps below show how the container orchestration capability (Kubernetes) can be extended to the edge using a lightweight, open-source technology such as KubeEdge. The checkout and digital services from the cloud can be deployed, managed, and monitored seamlessly at the stores without any specialized, proprietary hardware or software, ensuring a highly performant, always available and digitally enhanced checkout capability at the stores at minimal or no additional cost.
Setting up central container management platform (Kubernetes): This can either be cloud managed (for example: AWS EKS, Azure AKS, or Google GKE) or an on-premise Kubernetes cluster. Deploy the required checkout services in the central Kubernetes cluster.
Extending the central Kubernetes cluster to stores (edge) using KubeEdge: This involves two parts—the cloud component and an edge counterpart (called cloud and edge core). Both the KubeEdge cloud and edge components can be installed using an installer called Keadm. KubeEdge is a lightweight framework which requires minimal hardware footprint; for the cloud component, it is 2 core and 4 GB RAM, and for the edge core, it is only 0.5 core and 100 MB RAM. With keadm, KubeEdge cloud and edge core can be installed with simple commands. Once KubeEdge is installed (both the cloud and edge components), retailers can replicate the checkout services from the cloud to the stores (edge) seamlessly, making them highly performant and available.
Offline mode (in-store resilience): The edge nodes and the deployed checkout services can run offline. They are available even when there is no connectivity to the cloud, fulfilling the key resiliency requirement for an uninterrupted and personalized checkout experience.
Monitoring and self-healing: The entire edge infrastructure, including the edge nodes (virtual machines or the physical ones) can be monitored and managed through a centralized Kubernetes dashboard. This not only improves the end-to-end monitoring of the checkout services across the retail estate but also facilitates self-healing of the services, ensuring that the checkout services are able to recover from any failures or crash and, hence, are always available.
CONCLUSION
Delivering seamless next-gen retail experiences will be about bringing processing and data closer to the edge and also harnessing the power of analytics to be responsive, agile, and customercentric.
Edge computing gives a high level of redundancy to process all transactions at the store itself, enabling a highly available and performant checkout experience at the stores. Open-source frameworks such as Kubernetes and KubeEdge allow retailers to realize the edge architecture in a cost-effective, vendor-neutral manner, allowing a seamless extension of cloud capabilities to the stores.
