Storage Maintenance Window Prediction via Time Series Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for scheduling maintenance activities in storage systems rely on administrator perception and guesswork, leading to potential incorrect scheduling decisions that can be costly and disruptive.
Innovation Solution
A processing platform that detects storage arrays, monitors performance data, generates time series analysis, and uses regression models to predict low-usage intervals for maintenance, providing users with optimized time intervals for maintenance activities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If maintenance activities are scheduled based on administrator perception and guesswork, then scheduling simplicity is maintained, but scheduling accuracy deteriorates leading to costly and disruptive decisions
Solution Approach 1:
The patent replaces the mechanical/manual system of administrator perception and guesswork with an automated computational system that collects performance data, generates time series analyses, and uses regression models to objectively determine optimal maintenance windows. This substitution transforms subjective human judgment into an objective data-driven automated process, resolving the contradiction between scheduling simplicity and accuracy.
Solution Approach 2:
The system enables the storage system itself to provide information about its own optimal maintenance timing through performance monitoring and analysis. By collecting and analyzing the system's own operational data, the system autonomously identifies when maintenance can be performed with minimal impact, eliminating the need for external administrator guesswork while maintaining scheduling simplicity for the user.
2Productivity
If maintenance is performed during high-usage periods, then maintenance availability is maintained, but system performance and productivity deteriorate
Solution Approach 1:
The system performs preliminary analysis of performance data and generates predictions about future low-usage periods before maintenance is actually scheduled. By anticipating optimal maintenance windows in advance through time series analysis and regression modeling, the system allows planners to schedule maintenance during predicted low-usage periods, thereby maintaining productivity while minimizing downtime loss.
Solution Approach 2:
The system dynamically identifies maintenance windows based on actual observed usage patterns rather than fixed schedules. By continuously monitoring performance data and adapting predictions to real-world usage variations, the system flexibly determines optimal maintenance timing that responds to changing system conditions, ensuring maintenance occurs during genuinely low-usage periods to maximize productivity.
Data Source
AI summary
A method in one embodiment comprises detecting one or more storage arrays in an information technology infrastructure, and receiving input-output (IO) operation performance data recorded over a given time period from the one or more storage arrays. The performance data comprises a plurality of IO operation counts, each IO operation count comprising a number of IO operations per time unit for a component of a given storage array. The method also includes analyzing metadata for the IO operation counts to generate a time series comprising the IO operation counts sorted over a plurality of ordered time intervals of the given time period, and identifying a plurality of time blocks within the time series, each of the time blocks comprising a subset of the ordered time intervals. A proposed time interval for performance of a planned maintenance activity is generated based on one or more of the time blocks.