In-Place Adjacency Table Updates for Network Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network elements face significant challenges in quickly converging forwarding information following topology changes, especially with large equal-cost multi-path routes, leading to increased CPU utilization and longer convergence times, which is undesirable for data centers requiring rapid maintenance and minimal CPU usage.
Innovation Solution
The method involves in-place updates of adjacencies within the network element's adjacency table in response to topology changes, such as prefix advertisements, peer up/down events, by determining next hop lists and updating or deleting entries in the adjacency table without rebuilding the entire forwarding information base, thus reducing churn and CPU usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the network element processes topology changes by rebuilding the entire forwarding information base, then the forwarding information can be updated, but the CPU utilization increases and convergence time increases
Solution Approach 1:
The patent segments the forwarding information base into two distinct components: the route table (containing prefix-to-adjacency mappings) and the adjacency table (containing adjacency details). When a topology change occurs, only the adjacency table is updated in-place rather than rebuilding the entire FIB. This segmentation allows selective updates to minimize processing overhead and reduce convergence time while maintaining forwarding information accuracy.
2Reliability
If the network element processes topology changes by rebuilding the entire forwarding information base, then the forwarding information can be updated, but the CPU utilization increases
Solution Approach 1:
The patent segments the forwarding information base into the route table and adjacency table, enabling selective updates. When a topology change occurs, only the adjacency table is modified in-place, avoiding the CPU-intensive operation of rebuilding the entire FIB. This significantly reduces CPU utilization during topology changes while maintaining the accuracy of forwarding information through targeted updates.
Solution Approach 2:
The patent recovers computational resources by avoiding redundant processing. Instead of discarding the existing FIB structure and rebuilding it entirely, the system recovers CPU resources by performing incremental in-place updates on the adjacency table only, thereby reducing energy consumption and CPU utilization during topology change events.
3Reliability
If the network element processes all routes during topology changes, then complete forwarding information can be updated, but the processing time increases
Solution Approach 1:
The patent segments the FIB into route table and adjacency table, allowing the system to process only adjacency changes rather than all routes. This segmentation enables the network element to maintain complete forwarding information while dramatically improving convergence speed by limiting processing to the smaller adjacency table during topology changes.
Solution Approach 2:
The patent extracts the adjacency table as a separate, manageable component from the larger FIB structure. By taking out the adjacency information into its own table, the system can process topology changes efficiently by updating only this extracted component, thereby improving convergence speed without compromising the completeness of forwarding information.
4Adaptability or versatility
If the network element uses large ECMP routes, then routing flexibility increases, but the convergence time increases
Solution Approach 1:
The patent segments the FIB into route table and adjacency table, where the route table maintains routing flexibility through ECMP routes with multiple equal-cost paths, while the adjacency table stores detailed adjacency information. During topology changes, only the adjacency table is updated in-place, allowing the system to maintain routing flexibility with large ECMP routes while achieving fast convergence by limiting processing to the smaller adjacency structure.
Data Source
AI summary
A method and apparatus of a network element that processes changes to forwarding information is described. In an exemplary embodiment, the network element receives an indication of a change to a network topology of the network. The network element forwarding information includes a first plurality of prefixes and a plurality of adjacencies and each of the first plurality of prefixes references one of the plurality of adjacencies. In addition, in response to receiving the indication of the change to the network topology, the network element identifies a change to the plurality of adjacencies. Furthermore, the network element updates, in-place, the plurality of adjacencies in-place with the change to the plurality adjacencies.