Weighted ECMP EVPN Overlay for Uneven TOR Capacity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network systems evenly distribute traffic to Top-of-Rack (TOR) switches, leading to inefficiencies where some TORs are overwhelmed while others have unused capacity due to unequal bandwidth and processing capabilities, resulting in packet delays, drops, and service disruptions.
Innovation Solution
Implement dynamic weighted Equal Cost Multi-Path (ECMP) routing by having TORs advertise their processing capabilities to border leaf routers, which then distribute traffic based on bandwidth and server instance counts to optimize load balancing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If equal cost multi-path (ECMP) routing is used to distribute traffic evenly between TORs, then traffic distribution simplicity is improved, but system efficiency deteriorates due to mismatch between traffic load and TOR capabilities
Solution Approach 1:
The patent implements dynamic weighted ECMP routing where the border leaf router continuously receives capability advertisements from TORs and adjusts traffic distribution weights in real-time based on current TOR capabilities (bandwidth, processing capacity, server instance counts). This transforms the static equal-weight ECMP into a dynamic system that adapts to changing conditions, resolving the contradiction between operational simplicity and system efficiency.
Solution Approach 2:
The patent changes the routing parameter from uniform weight to capability-based weight. Each TOR advertises its capabilities (bandwidth, processing capacity, number of server instances), and the border leaf router calculates weights proportional to these capabilities. This parameter change allows traffic to be distributed according to actual TOR capacities, improving system efficiency while maintaining the ECMP framework for simplicity.
2Ease of operation
If traffic is evenly distributed to all TORs, then load balancing simplicity is improved, but packet delivery reliability deteriorates when TORs are overwhelmed beyond their capacity
Solution Approach 1:
The system dynamically adjusts traffic distribution weights based on real-time TOR capability advertisements. When a TOR's capacity is exceeded, it advertises reduced capability, and the border leaf router automatically reduces the traffic weight for that TOR, preventing packet loss. This dynamic adjustment maintains reliability without compromising load balancing simplicity.
Solution Approach 2:
The patent implements a feedback mechanism where TORs continuously advertise their current capabilities (bandwidth availability, processing capacity, server instance counts) to the border leaf router. The border leaf router uses this feedback to adjust traffic distribution weights, ensuring that no TOR is overwhelmed. This feedback loop maintains packet delivery reliability while preserving the simplicity of automated load balancing.
3Device complexity
If border leaf router uses fixed traffic distribution ratios, then routing configuration simplicity is improved, but bandwidth utilization deteriorates when TOR capabilities vary
Solution Approach 1:
The patent changes the traffic distribution parameter from fixed ratio to capability-proportional weight. Each TOR advertises its capabilities, and the border leaf router calculates weights as a function of these capabilities (e.g., weight proportional to bandwidth or processing capacity). This parameter change enables full bandwidth utilization while maintaining automated configuration that is as simple as receiving capability advertisements.
Solution Approach 2:
The system implements self-service where each TOR automatically advertises its own capabilities to the border leaf router, which then autonomously calculates and applies the appropriate traffic distribution weights. No manual configuration is needed - the system self-adjusts to maximize bandwidth utilization based on actual TOR capacities, resolving the contradiction between configuration simplicity and utilization efficiency.
Data Source
AI summary
Systems, methods and products for distributing traffic to top-of-rack switches (TORs) that dynamically determine their capability to handle traffic and advertise this capability to border leaf router(s). The advertised information may include the bandwidth between the TOR and the instances of the service to which it is connected, and/or the number of instances of the service to which the TOR is connected. The border leaf router computes a distribution of received traffic to the TORs based on this information and forwards traffic to the TORs according to the calculated distribution. As parameters of the TORs and servers change, the TORs can advertise the updated information, and the border leaf router can update its computed distribution to account for the changed parameters.