Dynamic Load Balancing for Multi-Rooted Tree Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In Data Center Networks (DCNs) with multi-rooted tree topologies, existing load balancing strategies fail to balance traffic effectively during link failures, leading to network congestion due to asymmetry and improper utilization of link capacities, as traditional routing protocols like OSPF do not account for link contentions and partial failures in Link Aggregation Groups (LAGs).
Innovation Solution
The proposed solution involves calculating maximum flows through minimum cuts for subgraphs that incorporate effective link capacities, considering link failures and sharing among nodes, to balance traffic and fully utilize available link capacities, including those on partially failed links. This is achieved by determining effective link capacities and using them to split traffic loads across multiple interfaces, reducing computational complexity through subgraph conversion and differentiation between healthy and unhealthy pods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional routing protocols like OSPF are used to route traffic, then destination reachability is achieved, but traffic load balancing fails when link failures occur due to network asymmetry
Solution Approach 1:
The patent implements dynamic load balancing by continuously monitoring link status and adjusting routing decisions in real-time. When link failures occur, the system dynamically recalculates effective link capacities and redistributes traffic loads across available paths, transforming the static routing behavior into adaptive dynamic routing that responds to network conditions.
Solution Approach 2:
The system changes routing parameters by introducing effective link capacity as a dynamic parameter that reflects actual available bandwidth. Instead of using fixed metric values, the protocol adjusts routing decisions based on calculated effective capacities that account for partial link failures, transforming the routing parameter from static to state-dependent.
2Reliability
If the whole LAG is disabled to handle partial link failure, then routing protocol convergence is achieved, but link capacity is wasted and congestion occurs on larger scale
Solution Approach 1:
Instead of disabling the entire LAG (excessive action), the system applies partial action by selectively utilizing only the healthy physical links within the LAG. The patent calculates effective link capacities that reflect the partial availability of the LAG and routes appropriate portions of traffic through remaining functional links, avoiding the need to disable the entire logical link.
Solution Approach 2:
The patent segments the LAG into individual physical links and evaluates their status independently. Rather than treating the LAG as an atomic unit that must be entirely enabled or disabled, the system segments it into functional and non-functional components, allowing selective usage of healthy segments while isolating failed segments.
3Productivity
If additional links are disabled to assist routing protocol, then temporary congestion mitigation is achieved, but overall network capacity is reduced and larger scale congestion occurs
Solution Approach 1:
The system implements feedback mechanisms by continuously monitoring link status and capacity utilization, then using this information to adjust routing decisions. The effective link capacity calculation incorporates real-time feedback about link health and traffic patterns, enabling the routing protocol to adapt to changing conditions without arbitrarily disabling links.
Solution Approach 2:
The patent replaces the mechanical approach of physically disabling links with a software-based solution that uses calculation and routing optimization. Instead of mechanically removing links from the network (disabling them), the system substitutes this with algorithmic routing decisions that steer traffic away from congested or failed links while maintaining their physical availability for other uses.
Data Source
AI summary
Techniques are described for balancing traffic load for networks configured in multi-rooted tree topologies, in the presence of link failures. Maximum flows (through minimum cuts) are calculated for subgraphs that incorporate effective link capacities on links between source and destination nodes. Effective link capacities may be determined that take into account link failures, as well as sharing of current available link capacities by multiple nodes. Traffic is balanced while simultaneously fully utilizing available link capacities, even available link capacities on partially failed links (e.g., partially failed Link Aggregation Groups (LAGs)).