Data Center Network Parameter Tuning for Congestion Fairness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network congestion leads to high delays, low throughput, and excessive resource consumption due to diverse network traffic models in data center networks, necessitating a flexible parameter adjustment method to improve network performance.
Innovation Solution
A parameter adjustment method involving exploration devices and management devices that dynamically adjust network parameters based on network status, using predefined reward functions and modification coefficients to optimize network conditions and ensure fairness among service nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If network parameters are adjusted dynamically based on network status, then network performance is improved, but parameter adjustment complexity increases
Solution Approach 1:
The exploration device autonomously performs parameter adjustment based on network status feedback without requiring manual intervention. The device automatically determines adjustment policies, modifies parameters, and evaluates results, enabling self-service operation that improves network performance while managing complexity through automation
Solution Approach 2:
The system implements a closed-loop feedback mechanism where the exploration device continuously monitors network status, compares actual performance against target values, and automatically adjusts parameters based on the deviation. This feedback-driven approach resolves the contradiction by making the system adaptive and self-regulating, improving performance while containing complexity through systematic control
2Speed
If parameter adjustment is performed quickly to meet preset conditions, then network congestion is reduced, but adjustment precision may be compromised
Solution Approach 1:
The parameter adjustment mechanism is designed to be dynamic, adapting its behavior based on network conditions. When congestion is severe, the system performs rapid adjustments to quickly alleviate the condition. As the network stabilizes, adjustments become more gradual and precise. This dynamic approach resolves the contradiction by making speed and precision context-dependent rather than fixed
Solution Approach 2:
The exploration device performs parameter adjustment in periodic cycles, with each cycle consisting of status monitoring, policy determination, parameter modification, and result evaluation. This periodic structure allows the system to balance speed and precision by conducting frequent but controlled adjustments, preventing both excessive sluggishness and hasty imprecise changes
3Productivity
If exploration devices adjust parameters autonomously, then service performance is improved, but network resource consumption increases
Solution Approach 1:
The exploration device performs partial parameter adjustments rather than exhaustive optimization at all times. It focuses adjustments only on parameters that significantly impact service performance, leaving less critical parameters unchanged. This selective approach improves service performance while avoiding the excessive resource consumption that would result from comprehensive parameter tuning
Solution Approach 2:
The system adjusts network parameters based on their relevance to current network conditions and service requirements. By dynamically selecting which parameters to modify and to what extent, the exploration device optimizes service performance while minimizing unnecessary parameter changes that would consume additional network resources. The parameter adjustment is proportional to the actual performance need
Data Source
AI summary
A parameter adjustment method and apparatuses are provided. The method includes: a first exploration device corresponding to a first service node receives a first parameter set from a management device, where the first parameter set includes one or more parameters used by the first exploration device to perform parameter adjustment, the first service node is any one of a plurality of service nodes included in a data center network, and the plurality of service nodes one-to-one correspond to a plurality of exploration devices. The first exploration device obtains a network status of the first service node based on each parameter adjustment until the network status reaches a preset condition, where each parameter adjustment includes: adjusting one parameter in the first parameter set. The parameter is flexibly adjusted based on the network status, to be applicable to different services.