Building a High-Performance PPPoE Gateways with AsterNOS-VPP

Introduction

Core Highlights

• Farewell to Memory Exhaustion: A multi-threaded architecture replaces legacy scripts, breaking through concurrency bottlenecks.
• Unleashing Data Plane Potential: VPP vector acceleration technology achieves cross-platform line-rate forwarding for PPPoE.
• Rejecting Vendor Lock-in: Natively integrates standard RADIUS to achieve completely neutral, centralized AAA management.

With the popularization of broadband access in enterprise campuses and ISP (Internet Service Provider) networks, the industry is gradually transitioning to the vBNG (virtual Broadband Network Gateway) architecture, attempting to achieve higher network agility and optimize construction costs through hardware-software decoupling.

However, when actually deploying and scaling PPPoE services, network engineers often face two core challenges: long-term ecosystem lock-in caused by the vendor-specific attributes (VSAs) of traditional devices, and data plane concurrency performance bottlenecks in early virtual gateways due to kernel network stacks.

AsterNOS-VPP now fully supports the PPPoE Server capability. It aims to transform a standard Layer 3 gateway into a high-performance vBNG node. Through native vBNG RADIUS integration and underlying engineering architecture refactoring, AsterNOS provides a pragmatic solution for open networking that combines high forwarding performance with a low deployment TCO (Total Cost of Ownership).

1. Breaking Through the Performance and Architectural Bottlenecks of Traditional vBNG

When diving into the underlying operational mechanisms of traditional vBNGs, one typically finds structural pain points that restrict high-concurrency services.

• Memory Consumption Under High Concurrency: Traditional PPPoE services often rely on external ip-up Python scripts for state synchronization. Under a concurrency pressure of 100 connections per second, this legacy mode can instantly consume up to 3GB of system memory, easily leading to resource exhaustion or system anomalies.
• Proprietary Ecosystem Lock-in: To implement QoS policies or billing, traditional architectures usually require binding to specific vendor dictionaries. This solidifies network control and makes the adaptation costs of migrating to standard open-source systems prohibitively high.

2. AsterNOS-VPP Technical Refactoring: High-Performance Data Plane and Open Ecosystem

To build a more reliable vBNG RADIUS architecture, AsterNOS does not stop at conceptual decoupling; instead, it has deeply optimized the underlying data plane and process management.

• Multi-threaded Architecture Optimizes Memory Management: The system abandons the inefficient script mode and adopts a single-process, multi-threaded accel-ppp solution. Simultaneously, through the custom-developed extra/pppsync plugin, it efficiently updates APPL_DB data directly at the lower level when PPP interfaces are created or deleted, drastically reducing the system’s instantaneous memory overhead.
• VPP Vector Acceleration Data Plane: AsterNOS dynamically translates standard SAI commands into VPP API calls through the innovative libsaivpp.so shared library. This integration allows the data plane to process PPPoE encapsulation and forwarding in user space in a batch (vector) manner, breaking the performance ceiling of packet-by-packet processing.
• TCP MSS Auto-Clamping to Prevent Packet Loss: To address the “network black hole” or fragmented packet loss issues caused by unreliable PMTUD (Path MTU Discovery), the system integrates the mss_clamp_plugin at the base level. This ensures efficient and stable transmission for dial-in users in complex network environments.

💡 Ready to experience this performance firsthand? > [Download the AsterNOS-VPP x86 VM Image here] to deploy and test it in your own environment today!

3. vBNG RADIUS Integration: Unifying Flexibility and Reliability

At the business implementation level, AsterNOS provides highly standardized enterprise-grade integration capabilities to meet the deployment needs of networks of varying scales.

• Fully Standardized Centralized Management: As part of the open ecosystem, AsterNOS can hand over Authentication, Authorization, and Accounting (AAA) entirely to standard remote enterprise RADIUS servers (such as FreeRADIUS). It supports combining local IP allocation or fully centralized IP pool management driven by RADIUS, returning network control to the user.

• Cross-Platform Hardware Decoupling: Whether the network architecture is based on ARM physical appliances featuring Marvell OCTEON CN102 or CN103 chips (such as Asterfusion ET2500 / ET3600), or deployed in virtualized environments like KVM and GNS3, AsterNOS-VPP can provide consistent functionality and excellent performance.

4. Moving Towards a Standard Open Network

Transitioning from proprietary devices to an open vBNG RADIUS architecture is a critical step in building a modern agile network. By combining SONiC’s standard control plane with the deeply optimized VPP data plane, AsterNOS-VPP provides you with a standard, efficient, and cost-controllable PPPoE access platform.

To learn how to transform a standard gateway into a high-performance vBNG, please refer to the complete PPPoE Server Configuration Guide for detailed RADIUS integration and local deployment steps to kickstart your network architecture upgrade.