Entropy Label Handling in MPLS LSP Stitching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in seamlessly handling entropy labels across stitched label-switched paths (LSPs) in MPLS networks, particularly in scenarios where segments have independent levels of entropy label support, leading to inefficiencies and data-plane issues during LSP stitching and hierarchy setups.
Innovation Solution
The implementation of techniques that determine the entropy label capabilities of segments within an end-to-end LSP, allowing for the appropriate addition or removal of entropy labels at stitching points to ensure efficient forwarding, including rules for propagating entropy label capabilities across stitching points and handling scenarios where segments differ in their entropy label support.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If entropy labels are uniformly applied across all segments of an end-to-end LSP, then load balancing efficiency is improved, but compatibility issues arise in segments that do not support entropy labels
Solution Approach 1:
The patent implements local quality by allowing entropy labels to be selectively applied only to segments that support them, rather than uniformly across the entire LSP. The system determines entropy label capability for each segment and applies labels locally where supported, while non-supporting segments simply forward packets without entropy label processing. This resolves the contradiction by maintaining load balancing efficiency in capable segments while ensuring compatibility across the entire end-to-end path.
2Productivity
If entropy labels are added at every stitching point where the next segment supports them, then load balancing is optimized, but device complexity and processing overhead increase
Solution Approach 1:
The patent applies preliminary action by having the notional ingress routing device determine the entropy label capabilities of all segments in advance during LSP setup. This capability information is stored and used to guide entropy label insertion decisions at stitching points, eliminating the need for complex real-time capability assessment. The stitching points simply follow pre-determined rules based on stored capability information, significantly reducing processing complexity while maintaining load balancing optimization.
3Reliability
If entropy labels are removed at notional egress points, then data-plane problems are avoided, but additional processing steps are required
Solution Approach 1:
The patent implements the extraction principle by removing entropy labels at notional egress routing devices before packets are forwarded to segments that do not support entropy labels. The system identifies the notional egress point where the LSP transitions from entropy label-capable to non-capable segments and extracts the entropy label at that specific location. This prevents data-plane problems in non-capable segments while confining the additional processing to only the necessary egress points, minimizing overall complexity.
Data Source
AI summary
In one example, a stitching point routing device, which stitches a previous segment of an end-to-end label-switched path (LSP) to a next segment of the end-to-end LSP, includes network interfaces configured to receive packets via the previous segment and send packets via the next segment, and one or more processors configured to determine whether the next segment supports entropy labels, determine whether a packet received from the previous segment is encapsulated by a label stack including an entropy label, when the next segment does not support entropy labels and when the packet is encapsulated by the label stack including the entropy label, remove the entropy label from the label stack, when the next segment supports entropy labels and when the packet is not encapsulated by the label stack including the entropy label, add an entropy label to the label stack, and forward the packet along the next segment.