VLT LAG Flow Table MAC Learning for Switching Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network configurations with virtual link trunking (VLT) domains face inefficiencies in data traffic switching, leading to congestion and extra hops due to the handling of data traffic on inter-chassis links, which are not intended for heavy data traffic and can cause suboptimal performance.
Innovation Solution
Implementing a method where packets arriving on inter-chassis links are checked for Media Access Control (MAC) learning using an Access Control List (ACL) or sampling table, with sampled packets being processed by the CPU to determine if the MAC was learned on a VLT LAG, allowing for flow-based ingress sampling and re-routing through the VLT LAG to avoid the inter-chassis link, thereby preventing congestion and extra hops.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data traffic is switched on inter-chassis links in VLT domains, then connectivity between switches is maintained, but network congestion occurs and performance degrades due to extra hops
Solution Approach 1:
The patent introduces a intermediary mechanism (VLT LAG and flow table) that mediates between the source and destination switches, allowing direct traffic forwarding without routing through inter-chassis links. The flow table acts as an intermediary data structure that enables intelligent routing decisions to bypass congested paths.
Solution Approach 2:
The patent segments the network path into direct forwarding paths and control planes. By separating the data plane (direct L2 forwarding via VLT LAG) from the control plane (CPU-based flow table management), traffic can flow directly between switches without being routed through inter-chassis links, eliminating unnecessary hops and congestion.
2Measurement precision
If MAC learning is performed on all packets, then accurate forwarding paths are established, but processing overhead increases and switching efficiency decreases
Solution Approach 1:
The patent applies partial action by performing MAC learning only on sampled packets rather than all packets. The flow table is populated selectively based on sampling criteria, which is sufficient to establish accurate forwarding paths without the overhead of processing every packet for MAC learning.
Solution Approach 2:
The patent performs preliminary MAC learning on sampled packets to pre-populate the flow table with forwarding information. This preliminary action on a subset of packets enables subsequent packets to be forwarded efficiently using the pre-established flow table entries, avoiding the need to process every packet for MAC learning.
3Reliability
If inter-chassis links are used for data traffic, then network connectivity is maintained, but link congestion occurs and additional hops are introduced
Solution Approach 1:
The patent introduces dynamic routing capabilities through the flow table, which can be updated based on sampled traffic patterns. This dynamic approach allows the network to adaptively select optimal paths, directing traffic through VLT LAGs when available rather than statically using inter-chassis links, thereby reducing hops and latency.
Solution Approach 2:
The patent implements feedback mechanisms where sampled packets provide information about actual traffic flows, which is then used to update the flow table. This feedback loop enables the system to learn and optimize forwarding paths over time, ensuring traffic takes the most direct routes available rather than relying on fixed inter-chassis link routing.
Data Source
AI summary
Embodiments of the present invention include systems and methods for providing more efficient switching of data traffic in networks that include a link aggregation domain. In embodiments, when packets arrive at one virtual link trunking (VLT) unit via an inter-node link (INL) from a peer VLT unit in the VLT domain, the first VLT unit checks whether the Media Access Control source address (MAC SA) was learned on a VLT LAG. If the MAC SA was learned on a VLT LAG, then, in embodiments, a special packet is sent to a VLT termination unit to have it learn the MAC destination address of the original packet on a sub-LAG from which the special packet was received. Learning the MAC destination address on the sub-LAG rather than the parent LAG allows for more direct switching of future packets.