Overview

Introduction

The Dataplane Stack provides users with solution reference implementations of high performance user space networking functionalities which would run on various Arm platforms.

End users interested in running these networking functions on Arm servers may include networking software developers, Arm partners and their customers, CPU performance analysts, CPU designers, and marketing team.

Target Platforms

Supported platforms

Current use cases have been validated in Ubuntu 20.04 distro on below machines:

  • Ampere Altra (Neoverse-N1)

Future Platforms

The solutions will be ported to more platforms in the future. The planned platform includes:

  • More silicon platforms - e.g., Neoverse-N2 server or embedded level CPUs, mainly focus on network functions and performance on Arm CPUs.

  • Fixed Virtual Platforms - FVP are complete simulations of an Arm system, including processor, memory and peripherals. These are set out in a “programmer’s view”, which gives a comprehensive model on which to build and test users’ software. FVP mainly focuses on verifying network functions and CPU behavior analysis in the early IP design stage.

Solution Architecture

_images/dataplane-stack-architecture.png

Dataplane Stack solution architecture

This software mainly focuses on the dataplane network, as performance is of the primary concerns in dataplane implementation on COTS servers, e.g., Ampere Altra, AWS Graviton3, Alibaba Yitian 710.

This project provides the networking functionalities based on some well-known user space network open source projects:

  • DPDK: acronym of Data Plane Development Kit, mostly runs in user space. It comprises of libraries to achieve high I/O performance and reach high packet processing rates, which are some of the most important features in the networking area. It supports multi vendors and architectures, and runs on variety of CPU architectures, e.g., Arm, Power, x86. DPDK was created for the telecom/datacom infrastructure, but today, it’s used almost everywhere, including the cloud, data centers, appliances, containers and more.

  • VPP: acronym of Vector Packet Processor, is a fast and scalable layer 2-4 multi-platform network stack that provides out-of-the-box production quality switch/router functionality. It runs in Linux user space on multiple architectures including Arm, Power and x86 architectures. The benefits of VPP are its high performance, proven technology, modularity, flexibility, and rich feature set.

  • Snort: is the foremost Open Source Intrusion Prevention System (IPS) in the world. Snort IPS uses a series of rules that help define malicious network activity to find packets that match against them and generates alerts for users.

Use Cases

Supported Use Cases

Currently, below use cases/features have been implemented and supported:

  • VPP based IPv4 forwarding

  • DPDK L3fwd based IPv4/IPv6 forwarding

  • VPP based L2 switching

  • VPP based TCP termination

  • VPP based SSL proxy

Future Use Cases

The solutions will gradually expand with more functions and features.

The functions below are classified from functionality perspective, instead of from software component or implementation perspective. For example, IPv4 routing function could be implemented using VPP L3 or DPDK L3 features, and the firewall function could be provided using VPP ACL/Classifier or Snort3. Certain use case has a large scope and may be implemented with multiple software components and instances.

The implementation mechanism of each network function below will be described in detail in later chapters in this documentation.

  • L2 Switching

    This is a typical L2 network function, including MAC learning, bridging, cross-connect, flooding, VLAN, etc.

  • IPv4 Routing

    Forward packets to the next-hop IP by looking up the destination IP in the packets.

  • IPv6 Routing

    IPv6 is almost identical to IPv4 routing under CIDR. The major difference is the addresses are 128-bit IPv6 addresses instead of 32-bit IPv4 addresses.

  • NAT

    Network Address Translation (NAT) is a method of mapping an IP address space into another by modifying network address information in the IP header of packets while they are in transit across a traffic routing device. This software covers the basic NAT functions, e.g., SNAT, DNAT.

  • VXLAN

    Virtual eXtensible Local Area Network (VXLAN) is an encapsulation protocol that provides data center connectivity. It uses tunneling to stretch Layer 2 connections over an underlying Layer 3 network. In data centers, VXLAN is the most commonly used protocol to create overlay networks that sit on top of the physical network, enabling the use of virtual networks.

  • IPSec

    Internet Protocol Security (IPSec) is a framework of techniques used to secure the connection over the network communication. The typical encryption/decryption algorithms are covered in the implementation.

  • Firewall

    A firewall is a network security device that monitors incoming and outgoing network traffic and permits or blocks data packets based on a set of security rules. The typical security applications are demonstrated based on some VPP and Snort3 security implementations.

  • TCP Termination

    This function is primarily based on VPP host-stack, and we investigate the implementations with some other open source TCP/IP user software.

  • SSL Proxy

    The SSL proxies control Secure Sockets Layer – SSL traffic -to ensure secure transmission of data between a client and a server. The SSL proxy is transparent, which means it performs SSL encryption and decryption between the client and the server.

  • Wireless Mid-haul or Back-haul

    Backhaul is better known than fronthaul. It refers to the connections between a mobile network and a wired network that backhaul traffic from disparate cell sites to a mobile switching telephone office.

Implement Consideration

This software aims to provide high-throughput packet processing software stack and solutions to solve customer and partner use case in networking applications. There are several considerations from technical perspective in the implementation process:

  • Integrate DPDK, VPP and SNORT/Hyperscan

    As mentioned above, this software provides network functions based on some open source projects, e.g., DPDK, VPP, Snort3 and scripts to combine them together for a complete solution.

  • Execute on Arm

    The software is mainly implemented, validated, optimized and deployed on Arm AArch64 architecture and platforms. The target platforms are listed in later section.

  • Optimize for Arm architectures

    Some of above open source projects are well-supported in community, including arch-specific compilation, community CI/CD, distro packages, arch-specific optimization, etc. While some of the projects are yet to be supported.

    That’s why the software has developed some scripts to resolve compilation issues on Arm platforms, compile project source code with arch-specific features, apply optimization patches done by Arm OSS Networking, deploy optimal parameters tuned for specific Arm platforms, etc.

  • Integrate security libraries, e.g., OpenSSL, IPSec

    A group of protocols are supported in established use cases to set up encrypted connections with the typical security libraries between devices. It helps keep data sent over public networks secure.

  • Support hardware offloads

    The solution supports use cases that allow to offload some router features onto the underneath hardware. This allows reaching wire speeds when routing packets, which simply would not be possible with the CPU.

  • Validate with multiple traffic generator

    Currently, quickstart and user guide use IXIA traffic generator to validate solution use cases. Some software packet generators are planned to be supported and provided along with the solution software in the future, e.g., TRex, DPDK Pktgen, Scapy.

Purposes

The network functions that the project provided serves multiple purposes of:

  • Showcase the integration of various components and act as proof of concept to all stakeholders who care about network function feasibility on Arm.

  • Allow for performance analysis/optimization with a solution that is close to customers’ production deployment.

  • Provide customers with a out-of-the-box reference design for rapid design modeling.

Repository Structure

The code repository are stored in Arm GitLab arm-reference-solutions group, with the link, https://gitlab.arm.com/arm-reference-solutions/dataplane-stack.

├── CHANGELOG.rst - Change notes for each releases
├── doc           - Sphinx based documents
├── LICENSE.rst   - License file of this project
├── MAINTAINERS   - Maintainer recording of specific module/use-cases
├── Makefile      - Compiling various components and documentation
├── setup.sh      - Script to set up use case environment
├── README.md
├── patches       - Patches for various components
├── usecase       - Scripts to help set-up various use cases quickly
└── tools         - Tools used to manage the project & traffic generators

License

The software is provided under an Apache 2.0 license (more details in Apache 2.0 ).

Contributions to the project should follow the same license.

Contributions

This project has not put in place a process for contributions currently. While proposals on new features, new use cases or any improvement are welcome always. You are welcome to create Merge-Requests in project’s Gitlab repository dataplane-stack to demonstrate your ideas and suggestions. We will sincerely investigate your proposals and provide feedback to you.

For bug reports, please submit an issue via GitLab repository dataplane-stack. Please provide as much information as possible:

  • Hardware information, including details on CPU, server, memory, disk, NIC, etc.

  • System configuration, like distro type and version, kernel version, command line parameters, NUMA, huge-page setting

  • Tools, e.g., compiler version, script tool version

  • Use case topology and testing method

Feedback and Support

To provide feedback or to request support, please contact project maintainers by email at dataplane-stack@arm.com, or check ‘MAINTAINERS’ file in top directory of repository for module/feature/use-case specific maintainers.

Arm licensees may also contact Arm via their partner managers to request support.