MAC Address Learning Limiting on Access Network Switches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ethernet network switches face challenges in managing large MAC address tables due to the need to store addresses of all devices, leading to memory overflow and increased costs and power consumption, especially in data centers with numerous physical and virtual machines.
Innovation Solution
Implementing a MAC learning technique that limits address learning to only active host devices involved in data exchanges, using conversational MAC learning where switches learn MAC addresses only of hosts in the path between communicating devices, and configuring ports to enable unicast or all-frame MAC learning based on port types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If switches learn MAC addresses of all devices in the network using traditional MAC learning, then complete forwarding capability is achieved, but memory requirements and costs increase significantly
Solution Approach 1:
The patent applies local quality by differentiating between core switches and access switches in the network architecture. Access switches are configured to learn MAC addresses only for hosts in their immediate vicinity, while core switches maintain comprehensive MAC address tables. This localized learning approach reduces memory requirements at access switches while preserving overall network forwarding capability through the core switches that maintain complete addressing information.
Solution Approach 2:
The network is segmented into hierarchical levels (core switches and access switches) with different MAC learning functions. Access switches perform limited MAC learning for local hosts, while core switches perform comprehensive MAC learning for the entire network. This segmentation allows the system to achieve complete forwarding capability through the core layer while reducing memory requirements at the access layer.
2Reliability
If switches store MAC addresses of all physical and virtual machines, then data frame forwarding to any destination is enabled, but power consumption increases
Solution Approach 1:
The patent implements local quality by configuring access switches to learn MAC addresses only for hosts directly connected to them, rather than maintaining tables for all network devices. This localized approach enables data frame forwarding for local traffic while significantly reducing the power consumption associated with maintaining and searching large MAC address tables at access switches.
Solution Approach 2:
The system dynamically adjusts MAC learning behavior based on switch type and network position. Access switches use limited MAC learning for local hosts, while core switches use comprehensive MAC learning. This dynamic configuration allows the network to achieve complete forwarding capability while optimizing power consumption by avoiding redundant MAC address storage at access switches.
3Reliability
If traditional MAC learning is used to build forwarding tables, then all host MAC addresses are stored, but memory overflow occurs in data centers with numerous devices
Solution Approach 1:
The patent applies local quality by implementing different forwarding table strategies at different network levels. Access switches maintain forwarding tables containing only MAC addresses of hosts in their immediate network segment, while core switches maintain comprehensive forwarding tables for the entire network. This approach ensures forwarding table completeness for local traffic at access switches while avoiding memory overflow by limiting the scope of MAC address storage.
Solution Approach 2:
The forwarding functionality is segmented between access switches and core switches. Access switches handle forwarding decisions for local hosts using small, manageable forwarding tables, while core switches handle inter-segment forwarding using comprehensive forwarding tables. This segmentation prevents memory overflow at access switches while maintaining complete forwarding capability through the core switches.
Data Source
AI summary
A LAN includes a CORE switch, some number of TOR switches, each linked to the CORE switch, and each of the TOR switches are linked directly to some number of host devices. Each of the switches in the LAN operate to process and transmit data frames they receive from neighboring LAN devices. Each TOR switch in the LAN builds and maintains a layer-2 forwarding table that is comprised of MAC address information learned from frames they receive from neighboring LAN devices. Selected ports/VLAN s on some or all of the TOR devices are designated to be CORE/switch facing ports (CFP) or host facing ports (HFP). Each of the CFPs are configured to only learn the MAC address in unicast frames it receives and each of the HFPs can be configured to learn the MAC address of both unicast and multicast data frames provided the destination MAC address included in the unicast frame is known.