Coordinated Gateways for TRILL and VXLAN Fabric Interworking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current approaches for interworking between TRILL fabric and VXLAN/IP fabric in data centers face limitations such as inability to perform per-flow load balancing, bandwidth wastage, and scalability issues, particularly in handling multi-destination traffic and network partitioning scenarios.
Innovation Solution
The implementation of new TLVs (Type Length Values) within the TRILL and VXLAN/IP fabrics, including Distribution Tree Allocation, Shared Tree Allocation, Virtual RBridge Nickname, and Virtual VTEP Address TLVs, allows for coordinated gateway management, preventing loopback issues and ensuring connectivity across fabric partitions by grafting IP multicast share trees with TRILL distribution trees and using split horizon techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If TRILL hello protocol is run over VXLAN overlay networks with gateways joining all VXLAN group addresses, then multi-destination traffic can be forwarded, but bandwidth is wasted and control plane load increases
Solution Approach 1:
The patent segments the VXLAN group address space by introducing a group address range field in the TRILL hello PDU. Instead of gateways joining all VXLAN group addresses, each gateway is assigned specific group address ranges it should join. This segmentation eliminates bandwidth waste by ensuring gateways only receive and process relevant multi-destination traffic for their assigned ranges.
Solution Approach 2:
The patent implements preliminary action by having gateways pre-announce their appointed forwarder status and assigned group address ranges through TRILL hello PDUs before actual multi-destination traffic flows. Remote RBridges can then directly determine which gateways to send traffic to based on this pre-established information, avoiding the need for gateways to join all VXLAN groups and reducing control plane load.
2Reliability
If vPC peer links are dedicated for TRILL and VXLAN fabric interworking, then connectivity is ensured, but the links may be underutilized or overloaded
Solution Approach 1:
The patent applies dynamics by enabling multiple active gateways to simultaneously handle multi-destination traffic between TRILL and VXLAN fabrics. Instead of a static vPC configuration with fixed peer links, the system dynamically load-balances traffic across multiple gateways based on their appointed forwarder status and assigned group address ranges. This dynamic distribution optimizes link utilization and prevents any single link from being underutilized or overloaded.
3Productivity
If load balancing is done on VLAN or segment basis in TRILL over VXLAN, then multi-destination traffic is handled, but per-flow load balancing is not possible
Solution Approach 1:
The patent implements local quality by allowing different load-balancing behaviors for different traffic types and flows. The system maintains per-flow load balancing for unicast traffic while using VLAN/segment-based load balancing for multi-destination traffic. This is achieved through the group address range assignment mechanism that enables fine-grained control over which gateway handles which traffic flow, providing adaptability for both unicast and multi-destination scenarios.
4Reliability
If vPC is used for gateway interconnection, then TRILL and VXLAN fabric interworking is achieved, but scalability is limited to two peer devices
Solution Approach 1:
The patent applies universality by using the standardized TRILL protocol as the interconnection mechanism between gateways, replacing the proprietary vPC solution. The TRILL hello PDU with group address range fields provides a universal interface that enables any number of gateways to interconnect and cooperate in handling multi-destination traffic. This universal approach removes the two-peer limitation of vPC and allows the system to scale to multiple gateways while maintaining reliability through the appointed forwarder model.
Data Source
AI summary
Coordinating gateways for multi-destination traffic across a TRILL fabric and a VXLAN/IP fabric with a plurality of TRILL IS-IS TLVs and a plurality of Layer 3 IS-IS TLVs is provided herein. The plurality of TRILL IS-IS TLVs and the plurality of Layer 3 IS-IS TLVs effectuate: grafting an IP multicast share tree with a plurality of TRILL distribution trees at only one of a plurality of gateways in a network interworking a TRILL fabric and a VXLAN/IP fabric; ensuring that multicast traffic traversing from the plurality of TRILL distribution trees is not looped back to the TRILL fabric through the VXLAN/IP fabric; restoring connectivity among a plurality of VXLAN/IP fabric partitions through the TRILL fabric if the VXLAN/IP fabric is partitioned; and restoring connectivity among a plurality of TRILL fabric partitions through the VXLAN/IP fabric if the TRILL fabric is partitioned.


