Logical Router Pod Linking Across Nodes for Distributed L7 Services
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Kubernetes orchestration platforms face challenges in effectively leveraging networking functions for micro-service-based cloud-native applications, particularly with network function virtualization data plane functions, as they struggle to handle layer 5-7 services efficiently due to limitations in current edge node architectures like the DPDK model.
Innovation Solution
Implementing logical routers in a Kubernetes cluster with a first Pod (L4 Pod) performing data forwarding operations and separate Pods (L7 Pods) handling layer 7 services, where the L4 Pod is affinitized to a node and L7 Pods are distributed based on resource usage, using a DPDK datapath with run-to-completion threads and control threads, and configuring L7 Pods through a network management system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a monolithic edge appliance model with DPDK is used, then networking performance is maximized for layer 4 services, but the architecture shows limitations when networking and security functions move to the application layer (L5-L7)
Solution Approach 1:
The patent segments the logical router functionality into separate Pods: an L4 Pod handling data forwarding operations and multiple L7 Pods handling specific layer 7 services. This segmentation allows the system to maintain high-performance DPDK processing for L4 while adding flexible L5-L7 service capabilities through independent, movable L7 Pods that can be distributed across different nodes based on resource usage.
2Adaptability or versatility
If L7 Pods are distributed across multiple nodes based on resource usage, then scalability and resilience are enhanced, but connectivity between L4 Pod and L7 Pods on different nodes becomes complex
Solution Approach 1:
The patent introduces Kubernetes networking components (service mesh, service discovery mechanisms, and virtual networking infrastructure) as intermediaries that automatically manage connectivity between the L4 Pod and distributed L7 Pods. These intermediaries handle service registration, load balancing, and network routing, eliminating the need for manual connectivity configuration and enabling seamless communication across nodes.
3Reliability
If L4 Pod is affinitized to a specific node, then data forwarding operations are stable, but the system loses flexibility in resource utilization
Solution Approach 1:
The patent segments the router functionality so that only the L4 Pod (handling stable, performance-critical data forwarding) is affinitized to a specific node, while L7 Pods (handling flexible, resource-intensive services) are allowed to migrate across nodes based on resource availability. This segmentation maintains forwarding stability while enabling flexible resource utilization for service functions.
Data Source
AI summary
Some embodiments provide a method for implementing a logical router of a logical network at a first Pod executing on a first node of a Kubernetes cluster to implement data message forwarding for the logical router. The method receives a data message for processing by the logical router. The method determines that the data message requires layer 7 (L7) service processing at the logical router. The method selects a second Pod from multiple Pods that perform L7 service for the logical router. Each of the Pods executes on a different node of the cluster. The method forwards the data message to the second Pod via a layer 2 (L2) construct that connects the first and second Pods.