Virtual Failure Domains for Storage Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current networked storage systems rely solely on physical failure domains for redundancy, failing to account for correlated disk failures based on vendor, model, and age, which can lead to diminished reliability and increased data loss probability.
Innovation Solution
The introduction of virtual failure domains, which consider disk properties such as vendor, model, and age, allows for dynamic and adaptive data placement across different failure domains, reducing the likelihood of data loss by separating disks with similar failure rates and distributing data objects to minimize their inclusion in a single failure domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is distributed across physical failure domains, then physical redundancy is achieved, but correlated failures based on vendor, model, and age are not prevented
Solution Approach 1:
The system segments disks into virtual failure domains based on vendor, model, and age characteristics rather than relying solely on physical location. This creates finer-grained groups that capture correlated failure modes, allowing data to be distributed across truly independent failure domains and reducing the probability of simultaneous data loss.
Solution Approach 2:
The patent adds a logical dimension to failure domain classification by incorporating vendor, model, and age attributes beyond physical location. This multi-dimensional approach to domain segmentation enables the system to identify and separate disks with correlated failure characteristics that would otherwise be missed by physical-domain-only approaches.
2Reliability
If disks with similar failure rates are separated into different virtual failure domains, then reliability is improved, but data migration complexity increases
Solution Approach 1:
The system dynamically creates and manages virtual failure domains based on disk attributes (vendor, model, age) rather than using static physical domains. This allows the failure domain structure to adapt as disks are added, removed, or replaced, automatically maintaining proper data distribution and redundancy without manual reconfiguration.
Solution Approach 2:
The patent changes the parameters used to define failure domains from purely physical location to include logical attributes (vendor, model, age). This parameter expansion enables more accurate identification of correlated failure risks and allows automated data placement decisions based on these additional dimensions.
Data Source
AI summary
A method for storage systems improvement includes collecting information that indicates one or more failure correlations for disks in a storage system. The disks are then separated into a plurality of virtual failure domains based on the indicated one or more failure correlations. The method then determines that all data objects of a set of redundant data objects are included in a first virtual failure domain. Responsive to determining that all data objects of the set of redundant data objects are included in the first virtual failure domain, the method then migrates at least one data object of the set of redundant data objects from a first disk in the first virtual failure domain to a second disk in a second virtual failure domain.