DPU Data Rate Feedback for Storage Cluster Congestion Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing TCP flow control methods in storage clusters can only reduce data traffic output rates but fail to alleviate pressure on the storage cluster, leading to suboptimal performance.
Innovation Solution
A method and apparatus that dynamically adjust data transmission rates based on reliability indicator data, such as packet loss rates, to optimize data traffic to and from storage clusters, using a DPU system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If TCP flow control method is used to reduce data traffic output rate, then network congestion is avoided, but storage cluster pressure is not relieved and performance does not meet expectation
Solution Approach 1:
The patent implements a feedback mechanism where the DPU detects reliability indicator data (such as packet loss rates) from the storage cluster and uses this feedback to dynamically adjust the data transmission rate. This closed-loop control allows the system to respond to actual storage cluster conditions, relieving pressure on the storage cluster while maintaining stable performance, thereby resolving the contradiction between avoiding network congestion and maintaining storage cluster productivity.
Solution Approach 2:
The system enables the DPU to autonomously monitor reliability indicators and self-adjust transmission rates without external intervention. This self-service capability allows the system to automatically balance network congestion avoidance with storage cluster performance optimization, eliminating the need for manual tuning and ensuring continuous optimal operation.
2Productivity
If data transmission rate is increased to improve efficiency, then data transmission efficiency is improved, but network congestion occurs and storage cluster stability deteriorates
Solution Approach 1:
The patent employs dynamic adjustment of the data transmission rate based on real-time reliability indicator data detected from the storage cluster. Instead of using a fixed transmission rate, the system continuously adapts the transmission rate to match current storage cluster conditions, thereby maintaining high data transmission efficiency while preventing network congestion and ensuring storage cluster stability.
Solution Approach 2:
The system changes the transmission rate parameter dynamically based on detected reliability indicators such as packet loss rates. By adjusting this key parameter in response to storage cluster conditions, the system optimizes data transmission efficiency while maintaining stability, resolving the contradiction between speed and reliability.
3Device complexity
If fixed transmission rate is used to simplify control, then system complexity is reduced, but inability to adapt to changing conditions reduces performance
Solution Approach 1:
The DPU autonomously monitors reliability indicators and adjusts transmission rates without requiring complex external control systems. This self-service approach maintains relatively simple system architecture while achieving adaptive performance optimization, resolving the contradiction between control simplicity and performance adaptability.
Data Source
AI summary
The present disclosure relates to a method and apparatus for controlling a data transmission rate based on a DPU, and a data processing unit. The method includes: receiving data traffic fed back by a storage cluster on the basis of an initial data read request, where the initial data read request is transmitted by a processor to the storage cluster at a first moment at an initial transmission rate; detecting reliability indicator data corresponding to the data traffic; analyzing the reliability indicator data to obtain a target adjustment strategy corresponding to the initial transmission rate; and adjusting the initial transmission rate based on the target adjustment strategy to obtain a target transmission rate, so that the processor transmits a data read request to the storage cluster at the target transmission rate at a second moment, where the second moment is a next moment of the first moment.