Predicting Upgrade Success in Hyper-Converged Infrastructure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In hyper-converged infrastructure (HCI) clusters, especially in remote office branch office (ROBO) deployments, the predictability of lifecycle management (LCM) events is low due to factors like varying network conditions, time zones, and lack of correlation between ROBO sites, making it difficult to determine the likelihood of success or failure of upgrades across multiple sites.
Innovation Solution
An information handling system that collects and analyzes data from multiple remote sites to determine a ranking of metrics criticality for upgrade events, allowing for a prediction of the likelihood of success based on historical data, enabling proactive remedial actions before attempting upgrades.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If upgrade bundles are pushed from centralized management to multiple ROBO sites simultaneously, then upgrade deployment efficiency is improved, but network bandwidth constraints and varying network conditions cause unpredictable success rates and failures at different sites
Solution Approach 1:
The system collects LCM event data from multiple ROBO sites, including success/failure outcomes and associated metrics. This feedback loop enables the system to learn from actual upgrade results across different network conditions and site configurations, improving prediction accuracy for future upgrades while maintaining efficient simultaneous deployment.
Solution Approach 2:
Before executing upgrades at ROBO sites, the system performs preliminary analysis by comparing site-specific metrics against historical data from similar successful and failed upgrades. This preliminary action identifies potential failure risks in advance, allowing administrators to take remedial actions before the actual upgrade attempt, thus maintaining high deployment efficiency while improving success rate predictability.
2Adaptability or versatility
If LCM events are executed independently at each ROBO site without correlation, then site autonomy and operational flexibility are improved, but the ability to predict success rates across the organization deteriorates due to lack of shared learning
Solution Approach 1:
The system merges data from independent LCM events across multiple ROBO sites by collecting and analyzing success/failure outcomes and associated metrics from all sites. This combination creates a comprehensive organizational knowledge base that maintains site autonomy while enabling predictive analytics across the entire organization, allowing each site to learn from collective experiences.
Solution Approach 2:
Independent LCM events at each ROBO site generate feedback data that is collected and analyzed centrally. This feedback mechanism enables the system to identify patterns, correlations, and critical success factors across different sites and conditions, improving organizational-wide predictability without interfering with individual site operational flexibility.
3Reliability
If comprehensive data collection from multiple ROBO sites is performed to improve prediction accuracy, then LCM success rate predictability is improved, but data processing complexity and system resource requirements increase
Solution Approach 1:
The system extracts only the most relevant and critical metrics from comprehensive LCM event data, such as upgrade bundle version, site hardware configuration, network conditions, and success/failure outcomes. By focusing on key discriminating factors rather than processing all available data, the system achieves high prediction accuracy while maintaining manageable data processing complexity.
Solution Approach 2:
The system transforms raw LCM event data into standardized parameters and features that are suitable for machine learning analysis. This parameter transformation includes normalizing metrics, encoding categorical variables, and selecting feature subsets, which reduces data processing complexity while preserving the information necessary for accurate success rate prediction.
Data Source
AI summary
An information handling system may be configured for: receiving first information from a plurality of other remote information handling systems indicative of a success or a failure of a corresponding upgrade event that was performed at such other remote information handling systems; receiving second information from the plurality of other remote information handling systems indicative of scores for such other remote information handling systems in a plurality of metrics; determining, based on the first and second information, a ranking of the metrics based on their criticality to the upgrade event; receiving third information from the particular remote information handling system indicative of scores for the particular remote information handling system in the plurality of metrics; and determining a likelihood of success for the upgrade event based on the determined ranking of the metrics and the scores for the particular remote information handling system in the plurality of metrics.