Storage System Inline Deduplication Load Adaptation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Inline deduplication in storage systems incurs significant resource overheads due to frequent fingerprint queries, which can impact system performance.

Innovation Solution

The method determines the load of the storage system and decides whether to perform inline deduplication on a data block. When the load is high, the block is directly stored without inline deduplication, and a mapping entry between the fingerprint and the storage address is recorded. This approach reduces resource overheads and improves storage performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of substance

If inline deduplication operation is performed by querying fingerprint table, then data redundancy is reduced, but resource overheads increase and system performance deteriorates

Engineering Contradiction:
Improvedata redundancyVSAvoidsystem performance
Core Design Contradiction:
Loss of substanceVSProductivity

Solution Approach 1:

The patent dynamically adjusts the deduplication strategy based on system load conditions. When load is high, it switches to deferred deduplication mode to avoid performance degradation; when load is low, it performs inline deduplication to maximize data reduction. This dynamic adaptation resolves the contradiction by making the system flexible rather than fixed in its approach.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent implements partial deduplication by recording mapping entries in logs for high-load scenarios instead of performing full inline deduplication. This partial action reduces resource overhead while still maintaining the ability to perform deduplication later, accepting that not all deduplication occurs immediately but achieving the goal under constrained conditions.

Inventive Principle:
Principle #16Partial or excessive action

2Volume of stationary object

If inline deduplication operation is performed on all data blocks, then storage space is optimized, but resource consumption increases

Engineering Contradiction:
Improvestorage spaceVSAvoidresource consumption
Core Design Contradiction:
Volume of stationary objectVSUse of energy by moving object

Solution Approach 1:

The patent uses periodic batch processing to perform deduplication on data blocks that were deferred during high-load periods. Instead of continuous inline deduplication, it periodically processes accumulated data blocks in batches when system load decreases, reducing instantaneous resource consumption while achieving storage optimization over time.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The patent extracts the deduplication operation from the immediate write path during high-load conditions, separating it into a deferred batch processing stage. This extraction removes the resource-intensive fingerprint querying from the critical write path, reducing resource consumption during peak periods while still achieving deduplication in the extracted batch processing phase.

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If fingerprint query is performed for every data block, then deduplication accuracy is maintained, but resource overheads increase significantly

Engineering Contradiction:
Improvededuplication accuracyVSAvoidresource overheads
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary recording of data block mappings to logs during high-load periods without immediate fingerprint verification. This preliminary action captures necessary information for future deduplication while avoiding the resource-intensive fingerprint query step at the time of writing, reducing overhead while preserving the ability to achieve accurate deduplication later when resources are available.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS12242747B2Data deduplication method and apparatus, and computer program product
Publication Date: 2025.03.04 HUAWEI TECH CO LTD
  • US12242747B2 patent drawing
  • US12242747B2 patent drawing
  • US12242747B2 patent drawing

AI summary

A data processing method in a storage system is provided. The method includes: when the storage system is under a first load, performing an inline deduplication operation; and when the storage system is under a second load, directly storing a received second data block without performing the inline deduplication operation, where the first load is less than the second load.