Capacity Forecasting Using Historical KPI Predictive Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for tracking processing capacity in horizontally scalable, real-time applications are ineffective due to their reliance on static thresholds, failing to account for dynamic and rapidly changing processing demands, leading to inefficient monitoring and automation of system performance degradations.
Innovation Solution
A computerized method that monitors historical key performance indicators (KPIs) and uses a predictive model to forecast processing capacity, generating alerts when the capacity is forecasted to exceed defined thresholds, allowing for timely prediction and prevention of capacity issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If static thresholds based on performance benchmarks are used to track processing capacity, then the tracking system is simple to implement, but the tracking becomes ineffective at predicting when an application is about to exceed capacity limits due to volatile and rapidly changing processing demands
Solution Approach 1:
The patent replaces static thresholds with dynamic predictive modeling that continuously adapts to changing processing demands. The system uses historical KPI data to train models that forecast future capacity limits, allowing the tracking mechanism to respond to volatile transaction volumes and rapidly changing factors in real-time, thereby resolving the contradiction between system simplicity and tracking effectiveness.
Solution Approach 2:
The system transitions from fixed threshold parameters to dynamic predictive parameters that are continuously updated based on historical performance data. By changing from static benchmark-based thresholds to model-based forecasts that incorporate multiple historical KPIs, the system maintains effectiveness despite volatile processing demands while accepting increased computational complexity.
2Ease of operation
If conventional tracking systems use static historical data for capacity monitoring, then the system is easier to operate, but it cannot account for dynamic and rapidly changing processing demands leading to inefficient monitoring and automation responses
Solution Approach 1:
The patent implements feedback mechanisms where historical KPI data continuously feeds into predictive models to refine capacity forecasts. The system automatically adjusts its predictions based on actual performance deviations from forecasted values, enabling efficient adaptation to dynamic processing demands while maintaining ease of operation through automated model training and threshold adjustment.
Solution Approach 2:
The system performs preliminary actions by continuously training predictive models on historical data and pre-calculating forecasted capacity limits before actual capacity issues occur. This allows the monitoring system to anticipate future capacity constraints and trigger appropriate responses in advance, improving adaptability while maintaining operational simplicity through automation.
3Productivity
If the volume of transactions per second is allowed to increase without predictive monitoring, then system productivity is higher, but processing capacity limits are exceeded leading to operations being delayed or failed
Solution Approach 1:
The system performs preliminary forecasting of capacity limits using historical KPI data before actual capacity constraints are reached. By predicting future capacity thresholds in advance, the system can proactively adjust transaction processing volumes or trigger capacity expansion actions, thereby maintaining high productivity while preventing capacity exceedances that would cause delays or failures.
Solution Approach 2:
The patent uses feedback loops where actual transaction volumes and processing outcomes are continuously fed back into predictive models to refine capacity forecasts. This enables the system to adapt to changing productivity patterns and capacity constraints in real-time, maintaining reliable operation even as transaction volumes fluctuate and approach limiting thresholds.
Data Source
AI summary
Systems and methods forecast processing capacity by monitoring historical key performance indicators (KPIs). The KPIs include measured values relating to a number of transactions that can be processed in parallel, an average number of transactions currently being processed, and an average amount of time to process each current transaction. One or more query parameters are received to query stored KPI data corresponding to the monitored historical KPIs. A forecasted processing volume level based on predicted values determined from a predictive model of the stored KPI data. The method also comprises generating an alert in response to the processing volume level being forecast to exceed a defined threshold is generated, thereby providing more reliable capacity monitoring and prediction of problems with processing applications that may otherwise go undetected until performance impacts occur.