Cold Hot Data Identification Threshold Calculation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current hierarchical storage systems lack a uniform criterion for identifying cold and hot data across storage nodes, limiting the practicality and availability of hierarchical data storage and migration.

Innovation Solution

A method for calculating a uniform identification threshold for distinguishing cold and hot data by aggregating IO statistical information from multiple storage nodes, using IO count and sequentiality values to determine the threshold, and implementing data migration across nodes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If distributed cold and hot data identification function modules are deployed on storage nodes, then hierarchical storage can be implemented on individual nodes, but uniform identification criterion across the entire storage system cannot be achieved

Engineering Contradiction:
Improvehierarchical storage implementationVSAvoiduniform identification criterion
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

A storage management server is introduced as an intermediary component that collects IO statistical information from multiple storage nodes, performs centralized threshold calculation, and distributes unified identification criteria back to all nodes. This mediator enables system-wide uniformity while preserving distributed hierarchical storage capabilities.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The threshold calculation function is merged from individual storage nodes to a centralized storage management server. By combining IO statistical information from all nodes and performing unified threshold calculation, the system achieves consistent cold and hot data identification criteria across the entire storage system.

Inventive Principle:
Principle #5Merging (Combining)

2Productivity

If storage system capacity and performance are increased to meet requirements, then performance and capacity needs are satisfied, but system cost increases

Engineering Contradiction:
Improvestorage performance and capacityVSAvoidsystem cost
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

Different storage tiers with distinct performance characteristics are assigned to different data types. Hot data is stored on high-performance storage mediums while cold data is stored on lower-performance, lower-cost mediums. This local quality differentiation optimizes performance for frequently accessed data while reducing costs for less accessed data.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The storage system is segmented into multiple hierarchical tiers based on data access patterns. By dividing storage resources into hot data storage and cold data storage tiers, the system achieves cost-effective performance optimization without requiring all storage capacity to be high-performance.

Inventive Principle:
Principle #1Segmentation

3Adaptability or versatility

If data migration across storage nodes is enabled, then system-wide hierarchical storage is achieved, but system complexity increases

Engineering Contradiction:
Improvedata migration capabilityVSAvoidsystem coordination complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The storage management server acts as a mediator that handles the complexity of cross-node data migration. It calculates unified identification thresholds, determines which data blocks require migration, and coordinates the migration process, thereby simplifying the overall system architecture while enabling system-wide hierarchical storage.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements feedback mechanisms where storage nodes report IO statistical information to the management server, which then adjusts identification thresholds and triggers appropriate data migration actions. This feedback loop enables adaptive data migration across nodes while maintaining system-wide consistency.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3070591B1Cold and hot data identification threshold calculation method, apparatus and system
Publication Date: 2019.04.03 HUAWEI TECH CO LTD
  • EP3070591B1 patent drawingFigure 1
  • EP3070591B1 patent drawingFigure 2
  • EP3070591B1 patent drawingFigure 3

AI summary

A method, an apparatus, and a system for calculating a identification threshold to distinguish cold data and hot data. The method includes: sending an IO statistical information query command for a data block to each storage node; receiving IO statistical information sent by each storage node; performing an addition operation on IO count field values in a same count range in the received IO statistical information of all the storage nodes, to generate an IO count statistical value corresponding to each count range; and obtaining, according to a total capacity of hot data storage mediums, a size of data blocks, and the IO count statistical values on all the storage nodes and when the hot data storage mediums are filled with the data blocks corresponding to large IO count statistical values, a minimum IO count value in a count range corresponding to a minimum IO count statistical value in the large IO count statistical value, and using the minimum IO count value as a identification threshold to distinguish cold data and hot data. In the present invention, all storage nodes in a storage system, independently but cooperatively, implement hierarchical data storage in the entire storage system.