Cache Retention Using Reference Count and LRU Hybrid Logic

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional data caching algorithms in multi-tiered computing environments fail to accurately manage frequently used data segments, leading to inefficient storage latency due to the lack of consideration for reference count information in deduplication systems.

Innovation Solution

A method that utilizes reference count information, in conjunction with LRU and statistical data, to determine the retention time of data segments in the cache, prioritizing segments with higher reference counts for longer retention and updating retention durations based on deduplication operations and time intervals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If conventional LRU caching algorithms are used to manage cache retention, then cache management is simple to implement, but storage latency increases because reference count information from deduplication systems is not considered

Engineering Contradiction:
Improvestorage latencyVSAvoidcache management complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent combines reference count information from the deduplication system with LRU caching algorithms to create a hybrid cache management approach. The cache management module integrates both LRU timestamps and reference counts to determine which data segments to retain, merging two previously separate mechanisms into a unified system that reduces storage latency without excessive complexity

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system uses reference count information as feedback from the deduplication operations to dynamically adjust cache retention decisions. By continuously monitoring reference counts and incorporating this feedback into cache management, the system optimizes storage latency by retaining data segments that are actively referenced by multiple virtual storage locations

Inventive Principle:
Principle #23Feedback

2Productivity

If data segments with high reference counts are retained longer in cache, then storage performance improves, but cache memory is consumed faster by high-reference-count segments

Engineering Contradiction:
Improvestorage performanceVSAvoidcache memory capacity
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

The patent implements dynamic cache retention where the retention duration of data segments is not fixed but adjusts based on their reference counts. High-reference-count segments are retained longer while low-reference-count segments are paged out sooner, creating a dynamic allocation strategy that optimizes storage performance while managing cache capacity constraints through adaptive decision-making

Inventive Principle:
Principle #15Dynamics

3Reliability

If reference count information is obtained and examined concurrent with deduplication operations, then cache optimization is achieved, but processing overhead increases

Engineering Contradiction:
Improvecache optimization accuracyVSAvoidprocessing overhead
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary actions by obtaining reference count information concurrent with deduplication operations rather than as a separate subsequent step. This timing allows the reference count data to be readily available when cache management decisions are made, achieving cache optimization without requiring additional processing passes or extra energy-intensive operations

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS9898419B2Enhancing data caching performance
Publication Date: 2018.02.20 INTERNATIONAL BUSINESS MACHINE CORPORATION
  • US9898419B2 patent drawing
  • US9898419B2 patent drawing
  • US9898419B2 patent drawing

AI summary

For a cache in which a plurality of frequently accessed data segments are temporarily stored, reference count information of the plurality of data segments, in conjunction with least recently used (LRU) information, is used to determine a length of time to retain the plurality of data segments in the cache according to a predetermined weight, where notwithstanding the LRU information, those of the plurality of data segments having a higher reference counts are retained longer than those having lower reference counts.