Weighted Routing Optimization for HPC Cluster Bandwidth Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing routing algorithms in high-performance computing clusters, such as those using Infiniband technology, do not optimize communication routes effectively, leading to suboptimal bandwidth allocation and performance due to static routing configurations that do not consider the specific needs of different types of nodes.
Innovation Solution
A method that assigns weights to communication links based on the types of nodes involved in the routing, allowing the routing algorithm to favor certain routes over others by incrementing load levels according to these weights, thereby optimizing bandwidth allocation and improving routing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If static routing configurations are used in HPC clusters, then routing simplicity and stability are maintained, but bandwidth allocation efficiency and performance are suboptimal
Solution Approach 1:
The patent changes the parameter of routing configuration from static to dynamic by introducing weight values that can be modified based on node types and communication patterns. The routing algorithm dynamically adjusts path selection by changing the weight parameters associated with different communication links, allowing optimal bandwidth allocation without complete reconfiguration of the routing infrastructure.
Solution Approach 2:
The patent introduces dynamic routing by allowing the routing algorithm to adaptively select paths based on current cluster conditions. The weight values assigned to communication links can be dynamically adjusted according to node types (compute, storage, administration) and communication frequencies, enabling the system to respond to changing workload requirements while maintaining routing stability through controlled parameter adjustments.
2Productivity
If uniform routing paths are used for all nodes, then routing algorithm simplicity is maintained, but specific data transfer needs of different node types are not met
Solution Approach 1:
The patent applies local quality by assigning different weight values to communication links based on the specific types of nodes involved (compute nodes, storage nodes, administration nodes). Instead of using a uniform routing approach, the system tailors the routing weights locally according to the communication needs of each node type pair, allowing optimized data transfer efficiency for each specific communication scenario while maintaining a unified routing algorithm framework.
3Productivity
If routing algorithms consider all possible paths, then routing optimization potential is maximized, but computational overhead and routing determination time increase
Solution Approach 1:
The patent applies partial action by having the routing algorithm consider only the most relevant paths based on pre-assigned weight values rather than evaluating all possible paths in the cluster. The weight-based scoring system allows the algorithm to efficiently identify and select optimal routes without exhaustive search, reducing computational overhead and routing determination time while maintaining effective routing optimization.
Data Source
AI summary
The invention relates in particular to the optimization of routing in a cluster comprising a plurality of nodes and static communication links connecting nodes of the plurality of nodes, said routing being based on load levels associated with the communication links. In order to establish a connection between at least two nodes of the cluster that have been identified (505), at least one route is determined (510) that connects the identified nodes according to the communication links, said route being determined according to the nodes identified, communication links and at least one load level associated with each communication link. A determined route is selected. Subsequently, a value of weight associated with the selected route is estimated (520) and a load level associated with each communication link of the selected route is incremented (525).