Sub-LAG Routing for VLT Next Hop Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In Virtual Link Trunking (VLT) networks, existing routing methods often result in sub-optimal paths due to lack of control over packet hashing, leading to unnecessary hops and oversubscription of interconnect links, as they prioritize Link Aggregation Group hashing over equal-cost multi-path decisions.
Innovation Solution
The implementation of a routing scheme that generates sub-LAG egress tables using Link Layer Discovery Protocol (LLDP) and Address Resolution Protocol (ARP) tables to identify and program sub-LAGs, ensuring packets are routed through optimal paths that honor ECMP decisions and avoid unnecessary hops.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If LAG hashing is performed over the entire VLT LAG, then load distribution across VLT peers is achieved, but packets may reach incorrect next hops requiring additional hops over ICL
Solution Approach 1:
The patent segments the VLT LAG into multiple sub-LAGs, where each sub-LAG is associated with a specific next hop. This segmentation allows the routing system to control which sub-LAG (and thus which VLT peer) receives packets destined for a particular next hop, eliminating unnecessary ICL hops while maintaining load distribution across peers.
Solution Approach 2:
The patent performs preliminary routing decisions by programming sub-LAG egress tables with next hop information before packet forwarding. This preliminary action ensures that packets are directed to the correct sub-LAG and VLT peer in advance, preventing the need for additional hops over ICL links.
2Device complexity
If LAG hashing determines the actual next hop, then simple routing is maintained, but ECMP decisions are ignored leading to sub-optimal paths
Solution Approach 1:
By segmenting the VLT LAG into sub-LAGs and associating each sub-LAG with specific next hops, the patent enables ECMP-aware routing without significantly increasing complexity. The sub-LAG egress tables provide a structured way to map next hops to specific sub-LAGs, maintaining manageable complexity while achieving path optimization.
Solution Approach 2:
The patent introduces sub-LAG egress tables as an intermediary between the routing decision and the actual packet forwarding. These tables act as a mediator that translates ECMP next hop decisions into specific sub-LAG selections, allowing ECMP optimization while keeping the forwarding mechanism relatively simple.
3Adaptability or versatility
If packets are hashed to VLT peers not multi-homed on all peers, then routing flexibility is maintained, but ICL links become oversubscribed
Solution Approach 1:
The patent segments traffic into different sub-LAGs based on next hop requirements, ensuring that packets are directed only to VLT peers that are actually connected to the destination. This segmentation prevents ICL links from being oversubscribed by routing traffic through peers that lack direct connections to the destination.
Solution Approach 2:
The patent applies local quality by making routing decisions specific to each next hop and its associated sub-LAG. Instead of applying a uniform hashing approach across all VLT peers, the system tailors the routing path to the local connectivity requirements of each destination, preventing ICL oversubscription while maintaining overall routing flexibility.
Data Source
AI summary
Various embodiments of the invention allow for rapid communication in virtual link trunking (VLT) networks in which network traffic flows over not all-homed VLT peer devices, while honoring Equal Cost Multi Path (ECMP) decisions and normal route decisions about next hops. Traffic flow is made deterministic and free of sub-optimal paths that otherwise cause unnecessary traffic over inter-node links in the VLT domain. In embodiments, this is accomplished by using receiving VLAN interface-IP addresses from VLT devices in order to create and use a sub-LAG egress table from which sets of ports that lead to intended VLT devices are derived. In embodiments, instead of a VLAN interface-IP addresses a routing MAC address is used when forming the sub-LAG.