Multi-Stage SDN Packet Pipelines With Tenant Namespace Isolation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional packet processing pipelines in multi-tenant SDN platforms face challenges with increased latency and reduced throughput due to shared computing infrastructure, and lack adequate security measures to isolate and protect tenant data.

Innovation Solution

Implementing a programmable multi-stage packet processing pipeline using multiple separate namespaces in the SDN network, where each namespace encapsulates graph objects within a predefined scope, ensuring isolation and security by restricting access to authorized parts of the pipeline.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If multiple namespaces are implemented to isolate tenant data, then security is improved, but device complexity increases

Engineering Contradiction:
ImprovesecurityVSAvoiddevice complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent divides the packet processing pipeline into multiple isolated namespaces, where each namespace contains a subset of graph objects (processing stages) that are accessible only to authorized tenants. This segmentation isolates tenant data and operations, preventing unauthorized access while maintaining security. The namespace mechanism creates logical boundaries without requiring physical separation, thus improving security while managing complexity through virtualization.

Inventive Principle:
Principle #1Segmentation

2Reliability

If packet processing uses conventional chaining with packet queuing, then security isolation is achieved, but latency increases and throughput decreases

Engineering Contradiction:
Improvesecurity isolationVSAvoidthroughput
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent merges multiple namespace programming models into a single unified graph execution in the data plane. Instead of processing packets through separate chained services with queuing between them, the system combines all graph objects from different namespaces into one integrated directed graph that processes packets in a single pass. This eliminates intermediate packet queuing and reduces latency while maintaining security isolation through the namespace boundaries in the control plane.

Inventive Principle:
Principle #5Merging (Combining)

3Adaptability or versatility

If users have broad access to program the entire pipeline, then adaptability is improved, but security vulnerabilities increase

Engineering Contradiction:
ImproveadaptabilityVSAvoidsecurity vulnerabilities
Core Design Contradiction:
Adaptability or versatilityVSObject-affected harmful factors

Solution Approach 1:

The patent applies local quality by assigning different levels of access rights to different users and namespaces. Each namespace has a defined scope of graph objects that users can program, with access controlled by authorization mechanisms. This allows users to have full adaptability within their authorized namespace while preventing them from accessing or modifying objects in other namespaces, thus maintaining security while preserving adaptability within bounds.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS20250279950A1Namespaces for protected multi-tenant packet processing pipelines
Publication Date: 2025.09.04 MICROSOFT TECHNOLOGY LICENSING LLC
  • US20250279950A1 patent drawing
  • US20250279950A1 patent drawing
  • US20250279950A1 patent drawing

AI summary

A programmable multi-stage packet processing pipeline executing in a data plane of a software-defined networking (SDN) network handles data packets according to objects in a directed graph that are programmed using multiple separate namespaces. Each namespace provides isolation for programming in the processing stages to provide protection for the pipeline that may span across multiple different tenant domains. While programming is handled using separate namespaces, the programmed stages are converted into executable code that runs in a single data plane of the SDN network to optimize data packet flow while minimizing latency.