Service Path Shortening via Backward Flow Offload Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current service chaining techniques in communications networks are limited in shortening service paths, resulting in longer latencies and reduced throughput, as they only allow bypassing of the closest service node in the infrastructure, failing to optimize the service path across multiple service nodes.
Innovation Solution
The method involves propagating flow offload decisions backward along the service path, caching these decisions at network elements, and extending service forwarding tables to include both next and previous hops, allowing for dynamic bypassing of service nodes and associated VEMs, thereby realizing the shortest possible service paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If flow offload is implemented only at the closest service node in the infrastructure, then the service path can be shortened for that specific node, but the overall service path remains suboptimal with longer latencies and reduced throughput
Solution Approach 1:
The patent combines multiple flow offload decisions from different service nodes into a unified infrastructure-level decision. By merging the offload requests from multiple service nodes along the service path, the infrastructure can create an optimized end-to-end service path that bypasses multiple service nodes simultaneously, rather than handling offload decisions individually at each node.
Solution Approach 2:
The patent introduces a new dimension of control by moving flow offload decision-making from the service node level to the infrastructure level. This dimensional shift allows the infrastructure to have a global view of the service path and make holistic offload decisions that optimize the entire path, rather than local decisions at each service node.
2Adaptability or versatility
If service paths are realized using overlays with service headers, then service metadata can be carried between classifier and service nodes, but the service path becomes longer with increased latency
Solution Approach 1:
The patent extracts the service header processing from the data path by implementing flow offload at the infrastructure level. Once a flow is offloaded, subsequent packets bypass the service nodes and their associated service header processing, eliminating the latency overhead of overlay encapsulation and service header handling while maintaining the ability to carry service metadata when needed.
Solution Approach 2:
The patent enables packets to skip through the overlay infrastructure and service nodes by implementing flow offload. Offloaded flows are rushed through the network core directly without stopping at intermediate service nodes, effectively skipping the latency-introducing overlay processing while still maintaining service chain functionality for flows that require it.
3Productivity
If flow offload decisions are propagated backward along the service path, then the service path can be optimized to its shortest length, but the infrastructure complexity increases
Solution Approach 1:
The patent implements preliminary action by propagating flow offload decisions backward along the service path before packets need to be forwarded. When a service node or infrastructure element determines that a flow can be offloaded, this decision is propagated upstream to all previous hops in the service path, allowing them to pre-configure their forwarding tables to bypass offloaded service nodes, thus optimizing the path before traffic flows.
Solution Approach 2:
The patent uses feedback by propagating flow offload decisions backward from the point of offload determination to the classifier and intermediate network elements. This feedback mechanism allows the infrastructure to learn about offload opportunities and adjust its forwarding behavior accordingly, creating an adaptive system that optimizes service paths based on actual flow characteristics and service node availability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method is provided in one example embodiment and includes receiving at a network element a flow offload decision for a first service node comprising a portion of a service chain for processing a flow; recording the flow offload decision against the first service node at the network element; and propagating the flow offload decision backward on a service path to which the flow belongs if the first service node is hosted at the network element. Embodiments may also include propagating the flow offload decision backward on a service path to which the flow belongs if the flow offload decision is a propagated flow offload decision and the network element hosts a second service node that immediately precedes the service node on behalf of which the propagated flow offload decision was received and a flow offload decision has already been received by the network element from the second service node.