Coarse-Fine Reference List Management for Scalable Deduplication

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Deduplication-based storage systems face inefficiencies in managing large numbers of segment objects due to limitations in reference counting and traditional reference lists, which can lead to incorrect resource reclamation and performance issues as the system scales.

Innovation Solution

Implementing a reference list management system with coarse and fine level entries, where coarse level entries are updated until a threshold is reached, then switching to fine level management, and associating entries with backup transactions to reduce processing time and overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If reference counting is used to track segment object usage, then resource reclamation can be automated, but the system becomes vulnerable to incorrect updates that may lead to premature deletion or failure to reclaim segments

Engineering Contradiction:
Improveautomated resource reclamationVSAvoidaccuracy of segment usage tracking
Core Design Contradiction:
Extent of automationVSReliability

Solution Approach 1:

The patent introduces reference list entries as an intermediary data structure between files and segment objects. Instead of directly counting references, the system maintains explicit list entries that record which files reference which segments. This intermediary approach allows for more reliable tracking because each reference is explicitly recorded rather than implicitly counted, reducing the risk of incorrect updates while still enabling automated reclamation when segments are no longer referenced.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If traditional fine-level reference lists are maintained for each segment object, then accurate tracking of file usage is possible, but processing time and system overhead increase significantly as the number of segment objects grows to billions

Engineering Contradiction:
Improveaccuracy of segment usage trackingVSAvoidprocessing time for reference list updates
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the reference list management into two levels: container-level (coarse) and segment-level (fine). At the container level, the system maintains a reference list of files that reference segments within each container. At the segment level, it maintains references from containers to segments. This segmentation allows the system to track usage accurately while reducing processing overhead by operating primarily at the coarser container level, only diving into fine-grained segment management when necessary.

Inventive Principle:
Principle #1Segmentation

3Quantity of substance

If the system scales to store hundreds of terabytes of data with billions of segment objects, then storage capacity increases, but the management overhead and indexing complexity become unacceptable

Engineering Contradiction:
Improvestorage capacityVSAvoidmanagement overhead
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent changes the dimension of reference list management by organizing references through containers as an intermediate layer. Instead of directly managing file-to-segment references across the entire system, the system introduces container-level abstraction. Files reference containers, and containers reference segments, creating a hierarchical structure that reduces the complexity of direct file-segment management while enabling the system to scale to billions of segment objects.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

4Reliability

If reference lists are updated at every file deletion to ensure accurate reclamation, then resource reclamation accuracy is maintained, but system performance deteriorates due to frequent updates on large datasets

Engineering Contradiction:
Improveaccuracy of resource reclamationVSAvoidsystem performance
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies partial action by updating reference lists selectively rather than comprehensively at every file deletion. The system updates container-level reference lists when files are deleted, and only updates segment-level references when necessary (e.g., when a container is fully emptied or when segment-level tracking is required). This partial updating approach maintains sufficient accuracy for resource reclamation while significantly reducing the performance impact compared to updating every reference at every deletion.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2612242B1System and method for scalable reference management in a deduplication based storage system
Publication Date: 2014.10.08 GEN DIGITAL INC
  • EP2612242B1 patent drawingFigure 1
  • EP2612242B1 patent drawingFigure 2
  • EP2612242B1 patent drawingFigure 3

AI summary

A system and method for managing a resource reclamation reference list at a coarse level. A storage device is configured to store a plurality of storage objects in a plurality of storage containers, each of said storage containers being configured to store a plurality of said storage objects. A storage container reference list is maintained, wherein for each of the storage containers the storage container reference list identifies which files of a plurality of files reference a storage object within a given storage container. In response to detecting deletion of a given file that references an object within a particular storage container of the storage containers, a server is configured to update the storage container reference list by removing from the storage container reference list an identification of the given file. A reference list associating segment objects with files that reference those segment objects may not be updated response to the deletion.