Edge & 5G
Gain crucial understandings of Edge software and 5G concepts with Intel® industry experts
91 Discussions

Design a Scalable Edge Architecture for Consumer Industries

Vicki_Cantrell
Employee
0 0 2,646

Consumer industries, comprising sectors including retail, hotels, restaurants, retail banking, and entertainment, operate in an environment characterized by rapid change and demanding customer expectations. This blog explains the benefits of, and design considerations for, adopting a scalable edge architecture to meet current challenges and flexibly adapt to the unpredictable demands of tomorrow.

Terminology

The consumer industry’s technology architecture includes hardware and software installed in various places, starting at the storefront and extending into the Cloud.

Terms are defined here for clarity:

 

Application

 

 

It is used interchangeably with “monolithic application,” a computer program that includes all its services in one entity.
ContainerA stand-alone executable software package including the application code and everything needed to run the piece of software, including runtime libraries and system tools.
MicroserviceAn independent part of an application, typically , executes a single function and can communicate with other microservices through APIs.
WorkloadThis term represents applications and microservices (any software that utilizes hardware resources).
Cloud

A collection of servers provided by a CSP (Cloud Service Provider), accessible over the network and providing hardware and software resources.

Enterprise data center

Like the Cloud, except the enterprise owns and operates it.

Edge

Physically, the storefront or near-customer warehouse. This includes, for example, the retail store, restaurant, hotel, and near-consumer fulfillment center.

Edge deviceDevices include digital signage media players, in-kiosk PCs, or PCs driving self-checkout stations. These devices are generally at the point of customer interaction.

Gateway

A device focused on a single workload, providing data consolidation and communication from multiple edge devices. For example, a gateway focused on digital signage provides content caching of new media pushed to the store and transfers consolidated audit/performance data to the enterprise data center or the cloud.
On-premise edge server
(or edge server)

 

A server in the store acting as the central computing hub for store applications. The edge server may also consolidate applications typically run on gateways.

 

Scalable Software Methodologies Are Key

Focusing on scalable software development methodologies is key, including microservices and containerization, as these practices have emerged as game-changers in how businesses leverage their hardware investments. By breaking down monolithic applications into modular and scalable components, industries can harness the full potential of their existing hardware infrastructure. This approach optimizes resource utilization and enhances flexibility.

Existing Technology Deployments

An edge installation with many technology solutions may look like the diagram below.

sea-before.jpg

 

Deploying and maintaining software solutions from different vendors can be complex and time-consuming:

  • Many solutions are deployed using an edge device and sensors (including cameras); some may also require a separate gateway, resulting in a highly fragmented set of hardware.
  • Maintaining and supporting solutions from various vendors increases IT resources and requires a technician per solution for on-site repairs.
  • As the amount of hardware increases, the infrastructure design and supporting costs increase. This includes space, thermal considerations, network cabling, and power.
  • Infrastructure modifications increase deployment timelines. Plus, replacing a solution may incur these costs again.

 

Simplifying deployment, maintenance, and upgrades

We propose adopting a scalable edge architecture utilizing an on-premise edge server with containerized and virtualized applications to address these challenges. The edge server consolidates applications that do not need to be installed in customer-facing devices including applications that typically run on the gateway.

 

 sea-after.jpg

 

sea-server-sw.jpg

This architecture provides the following benefits:

  • Scalability: The architecture supports stores of all sizes. Stores with a single digital sign or interactive kiosk may have only one edge device. In contrast, large enterprise-scale retailers may have many edge devices plus an on-premise edge server. On-premise compute can be increased or decreased to meet the needs of the applications running at the edge.
  • Simplified onboarding: The solution enables easy integration and deployment of new applications, reducing the time and effort required for implementation.
  • Support for diverse applications: The architecture accommodates modern microservices-based and legacy monolithic applications, allowing retailers to leverage existing investments while embracing newer technologies.
  • Support for diverse operating systems: The architecture enables a mix of operating systems to allow the retailer to utilize the best application for each function in the store, regardless of that application’s OS requirement.

Integrating an Enterprise Data Center (EDC)

The architecture may include an enterprise data center (EDC). An EDC is owned and operated by the retailer. The retailer is responsible for the infrastructure design, procurement, setup, maintenance, and management of the data center's hardware, networking equipment, and software.

This expanded architecture allows the retailer to run applications and store data at the edge, in the EDC, or the Cloud.

sea-edc.jpg

 

Hybrid Architecture with Edge, EDC, and Cloud Microservices

While an edge-centric architecture brings significant benefits, it is often complemented by a hybrid approach combining edge, EDC, and cloud-based microservices.

In this scenario, retailers can strike a balance between the localized edge environment for store-specific requirements and the EDC and cloud for centralized, enterprise-level functions:

  • Edge Microservices: Retail stores host critical microservices at the edge to ensure low latency, enhanced offline capabilities, and localized data processing. These microservices handle store-specific functionalities like point-of-sale systems, inventory management, and customer engagement applications.
  • EDC Microservices: Retailers utilize the EDC for microservices that aren’t required to provide real-time results if the hardware resources at the edge are fully utilized. The EDC is also used for applications that generate insights from data generated at multiple stores. The retailer can choose to store sensitive data in the EDC to keep direct control over the security measures for that data or meet privacy regulations.
  • Cloud Microservices: Certain microservices that benefit from centralized management, extensive computing resources, and data analysis capabilities are deployed in the cloud. These may include enterprise-wide applications, data warehousing, business intelligence, and advanced analytics tools. Cloud microservices enable retailers to leverage cloud platforms' scalability, global reach, and cost-efficiency.

Sensor Events

The scalable edge architecture isn’t just about deploying applications; it’s about orchestrating a symphony of sensor events. Sensor event processing revolutionizes how retailers develop insights and respond to dynamic retail demands by bringing data processing closer to the source.

Creating a unified event processing system can reduce the purchasing of various sensors and streamline the development and scale of retail use cases. Placing sensor processing and insight-generating compute on-premise helps eliminate latency and allows for fast response to critical events

  • Diversity of Sensors: Retail relies on a vast array of sensors – RFID, cameras, motion, weight, and temperature sensors – each providing a unique perspective on customer behavior, inventory status, and environmental conditions. Leveraging the full range of sensors across use cases is critical to experimenting, testing, and deploying new use cases at scale.
  • Real-time Responsiveness: Real-time insights from emerging sensor data are critical, enabling immediate responses such as personalized customer engagements, dynamic inventory management, and robust security measures.
  • Intelligent Event Filtering: Efficiently filtering and prioritizing insights is imperative. Categorizing events based on urgency, relevance, or specific business needs and directing them to the correct employees ensures that critical events receive prompt attention.
  • Data Quality and Accuracy: Ensuring the quality and accuracy of sensor data is paramount; validating and cleaning up sensor data prevents the propagation of ghost insights. The reliability of insights hinges on the integrity of the processed data. Creating trust between alerts, retail employees, and customers allows for more effective collaboration.
  • Security: The retail landscape contains strict privacy and non-negotiable compliance measures. Adding encryption, access controls, and other security measures to sensor event processing helps protect sensitive data.

Workload Placement

A complete scalable edge architecture (edge devices, on-premise edge computing, enterprise data center, and Cloud) enables the placement of applications and microservices on the right hardware at the right time. Placement is based on the requirements discussed in previous sections, such as latency, sensitive data handling, and cost targets. In addition, a microservice can transition from running on-premise to the Cloud if demand exceeds the on-premise compute available.

 

sea-workloads-placement.jpg

 

Edge devices are still required for some workloads. However, there are a variety of benefits to this implementation concerning hardware utilization in the store:

  • Resource utilization: Efficient use of hardware resources since many workloads are placed on a single on-premise edge server instead of multiple edge devices. The latter often leaves much of an edge device’s hardware resources unused.
  • Reduced hardware complexity: Less power and network infrastructure are needed because of the reduced number of edge devices and gateways.
  • Simplified remote management: A reduced set of diverse hardware devices requires less training for IT personnel and potentially faster triage.
  • Reduced maintenance costs: Hardware failures are reduced by lowering the number of devices and migrating applications away from aging hardware, which can degrade over time.
  • Lower noise levels: Less hardware may generate less noise, a concern in customer-populated areas. An edge server with a potentially higher noise profile can be placed in a discreet location in the back room to avoid impacting employees and customers.

Choosing the Right Hardware

The Intel® hardware product selection supports everything from basic CPU-based applications to heavy CPU and GPU workloads. The move toward GPU-heavy applications is often driven by adding high-quality graphics in Digital Signage or integrating CV and AI into the retail store.

Benchmarking is an important stage in selecting hardware. The Intel® Developer Cloud for the Edge offers access to CPUs, GPUs, and accelerators to help identify the best hardware platform for your software.

For a one-hour overview of Intel® Developer Cloud for the Edge, including a demo showing how to choose hardware and access performance data, check out this episode of the OpenCV YouTube channel.

Partner with Intel®

The scalable edge architecture creates a store infrastructure as flexible as Cloud infrastructure. Stores can implement this architecture to create a solution that streamlines operations, increases efficiency, reduces complexity, and reduces costs.

Reach out to your Intel® representative for more information. We understand the unique challenges businesses face in consumer industries. We are committed to providing the right blend of hardware solutions that empower your organization to thrive in this dynamic environment.

Resources

Visit these sites to learn more about Intel® products:

About the Author
Degree in Computer Systems Engineering, expertise in architecture, systems management, security accelerators, and retail deployments.