Distributed File System Metadata Update via Row Group Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing Append-Only file systems with columnar storage lack an effective method for dynamically updating metadata, resulting in high computing resource overhead and time consumption due to the need for regenerating entire files.

Innovation Solution

A metadata updating method that splits data records into multiple row groups, reorganizes the file format to include actual data blocks, data index blocks, local metadata blocks, metadata index blocks, and file footers, allowing for incremental updates of metadata without regenerating entire files, using the Append-Only file system's appending feature.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If metadata updating is performed in existing Append-Only file systems with columnar storage, then the file system maintains high write performance and consistency, but the metadata update operation requires regeneration and rewriting of all corresponding storage files, resulting in huge computing resource overhead and time consumption

Engineering Contradiction:
ImproveconsistencyVSAvoidmetadata update efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent segments the file structure into distinct components: data blocks, metadata blocks, and footers. This segmentation allows metadata to be updated independently without regenerating entire files. The metadata blocks are further divided into column-specific metadata, enabling granular updates of only the affected metadata portions while preserving the Append-Only property of the file system.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts metadata from the main file structure and stores it in separate metadata blocks. This extraction allows metadata to be manipulated independently from the data blocks. When metadata needs updating, only the extracted metadata blocks are modified and reappended, rather than regenerating the entire file, thus improving update efficiency while maintaining consistency.

Inventive Principle:
Principle #2Taking out (Extraction)

2Stability of the object's composition

If metadata updating is performed in existing Append-Only file systems, then the file system structure is maintained, but the operation requires rewriting of all corresponding storage files, leading to huge time consumption

Engineering Contradiction:
Improvefile structureVSAvoidmetadata update time
Core Design Contradiction:
Stability of the object's compositionVSLoss of time

Solution Approach 1:

The patent performs preliminary organization of metadata into separate metadata blocks with clear delimiters and structure markers. This preliminary action establishes a file structure that anticipates future update operations, allowing metadata to be modified by simple append operations rather than requiring complete file rewrites, thus reducing update time while maintaining structural stability.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces dynamic metadata blocks that can be independently modified and reappended to the file. Unlike static file structures that require complete regeneration for any modification, this dynamic approach allows the metadata portion to be updated flexibly by appending only the changed portions, significantly reducing update time while preserving the overall file structure integrity.

Inventive Principle:
Principle #15Dynamics

3Adaptability or versatility

If metadata updating is performed in existing Append-Only file systems, then the columnar storage structure is preserved, but the operation requires regeneration of entire files, resulting in extremely huge computing resource overhead

Engineering Contradiction:
Improvecolumnar storage capabilityVSAvoidcomputing resource overhead
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the columnar storage structure into independent data blocks and metadata blocks. Each column's data remains in its dedicated data block, preserving the columnar storage capability. The metadata blocks contain column-specific metadata that can be updated independently. This segmentation enables adaptability for columnar operations while reducing computing overhead by avoiding regeneration of unchanged data blocks.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by making only the metadata blocks subject to update operations, while the data blocks remain unchanged. This localized approach preserves the columnar storage structure's integrity and capabilities where needed, while minimizing computing resource overhead by avoiding unnecessary regeneration of the entire file. The update operation is confined to the specific metadata portions that require changes.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3113043B1Method, device and host for updating metadata stored in columns in distributed file system
Publication Date: 2018.04.25 HUAWEI TECH CO LTD
  • EP3113043B1 patent drawingFigure 1
  • EP3113043B1 patent drawingFigure 2
  • EP3113043B1 patent drawingFigure 3~4

AI summary

The present invention discloses a metadata updating method and apparatus based on columnar storage in a distributed file system, and a host, where the method includes: acquiring to-be-updated metadata in a data table of the distributed file system, splitting data records of the data table into multiple row groups on a row basis, and converting the data table into global file metadata and multiple row group files, where the row group file includes an actual data block, a data index block, a local metadata block, a metadata index block, and a file footer; determining whether the to-be-updated metadata belongs to the global file metadata, and if the to-be-updated metadata does not belong to the global file metadata, then: updating local metadata; and adding an updated local metadata block, an updated metadata index block, and an updated file footer to the multiple row group files according to updated local metadata. According to the method in the present invention, dynamic updating of metadata is implemented, which dramatically saves a time of executing an updating operation of this type and needed computing resources.