Storage System Reliability Design via Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods cannot straightforwardly determine the characteristics of a storage system that has a prescribed reliability, making it difficult to design systems with precise reliability specifications.
Innovation Solution
A computer-implemented method that computes reliability indicators for candidate storage systems, propagating indicators across hierarchical levels to identify precise system architectures and configurations matching a prescribed reliability, using stochastic analysis and non-invertible functions to compute mean time to data loss, expected annual fraction of data loss, and expected amount of data lost, allowing for accurate evaluation and configuration of storage systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reliability evaluation methods are used to assess storage systems, then reliability metrics can be precisely determined, but it is not possible to straightforwardly determine system characteristics for a prescribed reliability
Solution Approach 1:
The patent inverts the traditional reliability evaluation approach by creating a systematic method to work backwards from desired reliability metrics to system configuration. Instead of evaluating existing systems to determine their reliability characteristics, the invention provides a framework where target reliability parameters (MTTDL, EAFDL, E(H)) are specified first, and the system architecture is derived to achieve those targets through iterative candidate evaluation and comparison.
Solution Approach 2:
The patent applies preliminary action by pre-establishing comprehensive candidate storage system configurations with various architectures and redundancy schemes before actual deployment. Multiple candidate systems are designed in advance with different parameters (number of storage devices, data placement schemes, protection schemes), and their reliability metrics are pre-computed using stochastic analysis, allowing designers to select the optimal configuration that meets prescribed reliability requirements without trial-and-error experimentation.
2Reliability
If multiple reliability indicators are computed for candidate storage systems, then precise reliability assessment is achieved, but computational complexity increases
Solution Approach 1:
The patent segments the reliability assessment process into distinct hierarchical levels: individual storage device reliability, storage unit reliability (aggregating multiple devices), and overall storage system reliability. This segmentation allows computation of reliability indicators to be performed incrementally at each level using stochastic analysis, rather than computing all indicators simultaneously for the entire system. The three key indicators (MTTDL, EAFDL, E(H)) are computed separately for each candidate system, enabling modular and manageable computational complexity.
Solution Approach 2:
The patent systematically varies key parameters across candidate storage systems, including the number of storage devices, data placement schemes (clustered, declustered, symmetric), and protection schemes (replication, erasure coding). By changing these parameters across multiple candidates and computing reliability indicators for each, the invention identifies optimal parameter combinations that achieve prescribed reliability targets while managing computational complexity through focused parameter exploration rather than exhaustive search.
Data Source
AI summary
Technology for choosing a design for a computer data storage system having a prescribed reliability. The selection of a “matching storage system,” matching the prescribed reliability is based on computation of first and second reliability indicators.