As more enterprises expand their markets across broader regions, the increasing number of edge sites presents significant challenges to their IT infrastructure strategies — both in terms of capabilities and management.

To address this, Arcfra offers a high-performance, low-latency, highly reliable, and easily scalable edge cloud solution built on the Arcfra Enterprise Cloud Platform (AECP). This solution enables enterprises to support business systems at edge locations while achieving unified infrastructure management.

Below, we will provide more insights on the edge applications’ requirements for IT infrastructure, disclose more features of Arcfra Edge Cloud solution, and share Foxconn’s story in using AECP to integrate production lines on their edge sites.

Analysis of the Technical Requirements for IT Infrastructure Supporting Edge Applications

The IT infrastructure supporting edge computing is typically categorized as “cloud-edge-terminal”. “Cloud” refers to clouds or enterprise data centers. “Edge” refers to branch offices, branch data centers, or distributed clouds, also known as “near edge”. “Terminal” refers to device terminals or gateway edges, also known as “far edge”. Each layer assumes distinct operational responsibilities within this architecture.

1. Data Storage and Processing

• The far edge handles collection, staging, and preprocessing of raw data. For example, raw data captured by sensors, cameras, or smart terminals is filtered and screened locally before being transmitted to the near edge. Most edge applications and devices at this layer perform similar tasks.
• The near edge preliminarily stores, backs up, synchronizes, analyzes, and processes all data transmitted from the far edge.
• The central cloud provides unified storage, backup, and management for all data transmitted from all edges.

2. Application Latency Sensitivity

• Latency-sensitive applications (e.g., autonomous driving demanding millisecond-level responses) must operate at the far edge to ensure proximity to terminal devices and end users.
• Latency-insensitive applications (e.g., data anonymization and long-term training) can reside in the central cloud or data centers.
• Moderate latency-sensitive applications can be deployed at the near edge.

3. Application Management and Launching

• The near edge and far edge primarily execute the computation of algorithms and programs. Since algorithms and programs cannot originate from the far edge, they must be distributed from the central cloud to the near edge or the far edge.
• Executing algorithms and programs in the edge environments requires unified management and scheduling at the central site (typically data centers or the central cloud). Even when region-specific algorithms and program variations are needed, unified management and distribution remain the most efficient approach.

In summary, the far edge primarily collects and transmits the data, but without the capability to store it for the long term. From the near edge to the central cloud, these sites progressively undertake more data storage, analysis, and processing tasks. This necessitates an elastic and highly compatible infrastructure architecture to flexibly adapt to applications of varying scales and types. Additionally, it must also enable unified control of multi-site IT infrastructure, unified management, and on-demand distribution of hybrid workloads (algorithms and programs), as well as security policy management at each node to safeguard internal data and external access.

Arcfra Edge Cloud Solution

Arcfra Edge Cloud Solution integrates Arcfra Enterprise Cloud Platform for both VM and container environments, providing high-performance, low-latency, high reliability, easily scalable, and easy-to-maintain hybrid application deployment support for edge cloud scenarios.

Scalable and Flexible Computing

• Cluster size configured on demand
• Flexible expansion of nodes or resources

Production-Grade Storage with High Availability

• Full-stack optimization to unleash hardware potential
• Battle-tested in production with guaranteed stability

High-Performance and Low-Latency Computing

• Minimizes data transfer and access latency
• Supports compute offloading to hardware accelerators such as GPUs

Centralized Management

• Unified management of virtualization resource pools and physical servers across sites
• Unified resource monitoring, analysis, and alerting

Hybrid Application Environment

• Simultaneously deliver virtualization and container environments
• Uniformly manage resources for different applications

Intrinsic Security

• Enhanced security for the platform
• Embedded multiple security capabilities, such as network micro-segmentation and security auditing

Customer Story

Foxconn: Leverages AECP to Integrate Production Lines on Edge Sites

Foxconn, the world’s largest electronics manufacturing services (EMS) provider, initiated a plan years ago to replace traditional VMware virtualization infrastructure with a private cloud, supporting IT systems across its global branch companies. To ensure production security, each branch’s manufacturing line systems require strict quarantine from the intranet. Therefore, a dedicated independent infrastructure should be deployed for the intranet production line system, hosting critical production systems such as MES, SAP, and production line management platforms.

The company’s original private cloud solution — based on a self-maintained OpenStack + Ceph architecture — was so complex that it required more than 20 nodes for initial deployment. This not only led to high construction costs but also significantly increased operations and maintenance (O&M) complexity.

Against this backdrop, Foxconn conducted a POC test of AECP. Using the same single-node configuration, the company compared the performance of the original 7-node private cloud cluster with a 3-node AECP cluster. The results showed that despite reducing the number of server nodes by over half, AECP still outperformed the original solution in both IOPS and latency, far exceeding the company’s expectations.

Ultimately, Foxconn chose AECP to consolidate the IT infrastructure of its branch factories across Mainland China, Taiwan, Vietnam, and North America. At each edge site, the user deployed AECP clusters to support both DMZ and intranet production systems, including MES, ERP, and production line management systems, as well as development and testing(DevTest) environments and VDI workloads.

In Foxconn’s DevTest environment, Arcfra Kubernetes Engine (AKE) is used to support GPU-based workloads, enabling unified management of VMs and containers while improving service quality and efficiency. Additionally, Arcfra Network Service (ANS) has been deployed at multiple branch factories to provide microsegmentation-based network security for the DMZ environment. AECP clusters across multiple regions are centrally managed via Arcfra Operation Center (AOC), significantly improving O&M efficiency.

In addition, Foxconn has also built Active-Active stretched clusters across multiple branch factories, providing high availability protection for critical systems such as the MES system, its associated Oracle RAC databases, the SFC (Shop Floor Control) system, and the Production Line Management System. With AECP, Foxconn achieves data availability at a cross–data-center level with “zero RPO and RTO in minutes.”

Benefits

• Unified management of production lines across branch factories reduced infrastructure costs and streamlined operations.
• Active-Active clusters with ANS network security ensured multi-layered protection for business and critical data.
• The legacy VMware virtualization + SAN storage architecture was replaced to accommodate future business scalability.
• AKE supports GPU-accelerated applications for agile innovation while simplifying Kubernetes management.