Storage Cache Page Allocation for Hit Ratio and Delay Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing memory resource allocation methods in storage systems result in reduced hit ratios and increased service processing delays due to uniform eviction of cache objects, particularly those with large page resource requirements, leading to inefficient utilization of memory resources.

Innovation Solution

A resource allocation method and apparatus that adjusts page resource allocation based on a first model indicating the resource needed to meet a specific hit ratio, using weight values, historical information, and statistical data to optimize resource distribution among cache objects, thereby improving memory pool utilization and reducing service processing delays.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If resource allocation is based on historical data and simple algorithms, then implementation is simple, but resource allocation accuracy is low

Engineering Contradiction:
Improveimplementation complexityVSAvoidresource allocation accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The system implements feedback mechanisms by collecting real-time resource usage data and performance metrics, then using this feedback to dynamically adjust resource allocation decisions through the reinforcement learning model, improving accuracy while maintaining manageable complexity

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent replaces traditional mechanical/algebraic resource allocation algorithms with a neural network-based reinforcement learning system that learns optimal allocation strategies through interaction with the environment, achieving higher accuracy without proportionally increasing implementation complexity

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If resource allocation is based on manual configuration, then control precision is high, but labor cost is high

Engineering Contradiction:
Improvecontrol precisionVSAvoidlabor cost
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The system enables self-service by implementing automated resource allocation through reinforcement learning, where the system independently learns and executes optimal allocation decisions without requiring manual configuration, thereby maintaining high control precision while eliminating labor costs

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent substitutes manual configuration processes with an intelligent reinforcement learning system that automatically learns optimal resource allocation policies, achieving both high precision and automation

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If resource allocation is dynamic and adaptive, then resource utilization efficiency is improved, but system complexity increases

Engineering Contradiction:
Improveresource utilization efficiencyVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system implements dynamic resource allocation by using reinforcement learning to continuously adapt allocation decisions based on changing conditions, improving resource utilization efficiency while the learned policies help manage the inherent complexity

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system uses real-time feedback from resource usage monitoring to dynamically adjust allocations, improving utilization efficiency through adaptive decision-making that responds to actual system state

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4365749B1Resource allocation method and device
Publication Date: 2026.05.06 HUAWEI TECH CO LTD
  • EP4365749B1 patent drawingFigure 1
  • EP4365749B1 patent drawingFigure 2
  • EP4365749B1 patent drawingFigure 3~4

AI summary

This application discloses a resource allocation method and an apparatus, and relates to the field of data storage. The method includes: A processor outputs a first model based on a page resource allocated to each cache object and a data read hit ratio of the cache object, where the page resource allocated to each cache object is from a memory pool provided by a storage system. The first model indicates a page resource that needs to be allocated to the cache object when a specific hit ratio is met. The processor outputs a second model based on the first model corresponding to each cache object, where the second model indicates a target page resource to be allocated to each cache object. The processor adjusts, based on the second model, the page resource allocated to each cache object. The processor adjusts the page resource of the cache object based on the second model. In this way, that the processor evenly allocates or evicts the page resource of the cache object is avoided, utilization and allocation rationality of the page resource in the memory pool are improved, and a service processing delay is reduced.