Directory Metadata Index Sharding for Scalable Distributed File Systems

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The inefficiency in reading and writing operations in distributed file systems due to the reliance on a single index shard for storing index information of directory metadata, leading to reduced scalability and low Queries-per-second (QPS) performance.

Innovation Solution

Implementing a distributed file system architecture with multiple index shards, where each shard stores a portion of the directory metadata, allowing for efficient routing and processing of directory metadata requests through hash-based path resolution and remote procedure calls, thereby optimizing storage capacity and QPS.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If a single index shard is used to store index information for all directory metadata, then the system structure is simple, but the reading/writing efficiency deteriorates due to loading the entire index shard

Engineering Contradiction:
Improvereading/writing efficiencyVSAvoidindex shard structure
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent divides the single index shard into multiple index shards (first index shard, second index shard, etc.), where each index shard stores index information for a specific portion of directory metadata. This segmentation allows the system to load only the required index shard instead of the entire index shard, thereby improving reading/writing efficiency while managing complexity through structured division

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If the entire index shard is loaded for every read/write operation, then data access is straightforward, but the system scalability deteriorates due to single-point bottlenecks

Engineering Contradiction:
Improvesystem scalabilityVSAvoiddata access operation
Core Design Contradiction:
Adaptability or versatilityVSEase of operation

Solution Approach 1:

By segmenting the index shard into multiple smaller index shards, the system enables selective loading of only the relevant index shard based on the processing request. This approach improves scalability by distributing the data access load across multiple shards, eliminating the single-point bottleneck while maintaining ease of operation through hash-based routing to the appropriate shard

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of loading the entire index shard (excessive action), the system loads only the specific index shard that contains the required directory metadata (partial action). This partial loading approach reduces memory consumption and improves scalability while maintaining efficient data access through precise shard identification

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If a single index shard stores all directory metadata, then storage management is simple, but the QPS performance deteriorates due to processing bottlenecks

Engineering Contradiction:
ImproveQPS performanceVSAvoidindex shard architecture
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements an architecture with multiple index shards, where each shard handles a specific portion of directory metadata. This segmentation enables parallel processing of multiple requests across different shards, thereby improving QPS performance while managing architectural complexity through systematic division and hash-based routing mechanisms

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4693061A2Method and apparatus for processing directory metadata in a distributed file system, and device
Publication Date: 2026.02.11 BEIJING BAIDU NETCOM SCI & TECH CO LTD
  • EP4693061A2 patent drawingFigure 1~2
  • EP4693061A2 patent drawingFigure 3~4
  • EP4693061A2 patent drawingFigure 5~6

AI summary

The present disclosure provides a method for processing directory metadata in a distributed file system, an apparatus and a device, which are related to the field of computer technologies, in particular to the field of big data and distributed file systems. The specific solution is as follows: when a processing request is acquired, a first index shard corresponding to directory metadata indicated by the processing request may be determined according to the processing request; the first index shard may be loaded, and a metadata shard corresponding to the directory metadata indicated by the processing request is determined according to the first index shard. Information in the determined metadata shard is adjusted. A second index shard related to the directory metadata indicated by the processing request is determined according to the processing request, and index information corresponding to the directory metadata in the second index shard is adjusted.