Failure Domain Tree Data Distribution for Storage Repair
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage systems face inefficiencies in data recovery due to high I/O bandwidth requirements and uneven distribution of data blocks across failure domains, leading to inferior repair performance and increased risk of data unavailability when multiple nodes fail.
Innovation Solution
The implementation of a failure domain tree system that uniformly distributes data blocks across multiple levels, allowing for flexible failure tolerance values and the generation of symmetric failure domain trees from asymmetric ones to ensure balanced node connectivity and improved resilience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data blocks are distributed across failure domains in existing storage systems, then data availability is improved, but I/O bandwidth requirements increase and repair performance deteriorates
Solution Approach 1:
The system segments data objects into multiple data blocks and distributes them across different failure domains in a hierarchical tree structure. Each failure domain level (storage device, enclosure, rack, data center) contains a subset of data blocks, enabling localized repair operations that don't require accessing all data blocks, thus improving repair performance while maintaining data availability
Solution Approach 2:
The patent implements local quality by allowing different failure domain levels to have different failure tolerance values tailored to their specific reliability requirements. Each level can be configured with appropriate redundancy and data block distribution strategies, enabling optimized repair performance at each hierarchical level while maintaining overall system reliability
2Productivity
If data blocks are uniformly distributed across failure domain levels, then repair performance is improved, but the system complexity increases
Solution Approach 1:
The system dynamically adapts the failure domain tree structure based on actual hardware configurations and failure patterns. The uniform distribution strategy can be adjusted according to the number of data blocks, failure tolerance requirements, and hierarchical level characteristics, allowing the system to maintain optimal repair performance without rigid complexity
Solution Approach 2:
The patent introduces a hierarchical dimensional structure with multiple failure domain levels (storage device, enclosure, rack, data center). Data blocks are distributed across these dimensions, transforming a single-level distribution problem into a multi-dimensional solution. This approach improves repair performance by enabling localized repairs at lower levels while maintaining overall system integrity through higher-level redundancy
3Ease of manufacture
If asymmetric failure domain trees are used to match hardware configurations, then ease of implementation is improved, but data distribution uniformity deteriorates
Solution Approach 1:
The system intentionally uses asymmetric failure domain trees that reflect the actual asymmetric hardware configuration of storage systems. Different failure domain levels can have different numbers of nodes and varying connectivity patterns, making the system easier to implement with real hardware while the uniform distribution algorithm ensures balanced data block placement across this asymmetric structure
Data Source
AI summary
Systems, apparatus, and methods for data distribution within a failure domain tree are described herein. Data objects may be distributed uniformly into symmetric failure domain trees. Asymmetric failure domain trees may be used to generate symmetric failure domain trees. The systems, apparatus, and methods may distribute a data object to attempt to satisfy the failure tolerance level of as many failure domain levels of the symmetric failure domain trees as possible.