Automated Capacity Manager Optimizes Exponential Smoothing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud computing systems face service interruptions due to unpredictable user demands, as current predictive methods like exponential smoothing may lead to significant variations in forecasted resource requirements, causing capacity exhaustion before proactive measures can be taken.
Innovation Solution
An automated capacity manager optimizes exponential smoothing models by incorporating a 'forecast churn' metric alongside error measurements to stabilize predictions, ensuring consistent resource allocation and provisioning, thereby reducing the risk of service interruptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exponential smoothing is used to predict future resource usage based on historical data, then forecast accuracy is improved, but significant variations in forecasted values occur leading to service interruptions
Solution Approach 1:
The patent applies preliminary action by proactively provisioning resources before capacity exhaustion occurs. The system uses forecasted resource usage data to trigger resource allocation in advance, rather than reacting after problems occur. This is implemented through monitoring forecasted usage against thresholds and automatically provisioning resources before the shared pool is exhausted, preventing service interruptions before they happen.
Solution Approach 2:
The patent implements feedback mechanisms by continuously monitoring both actual resource usage and forecasted usage, then using this information to adjust resource provisioning decisions. The system compares forecasted usage patterns against available capacity and dynamically adjusts resource allocation based on this feedback loop, allowing the system to adapt to changing demand patterns while maintaining service reliability.
2Productivity
If reactive remedial actions are taken when capacity exhaustion occurs, then service recovery is achieved, but service interruptions have already happened
Solution Approach 1:
The system performs preliminary actions by provisioning additional resources in advance based on forecasted usage patterns before capacity exhaustion occurs. The monitoring component compares forecasted resource usage against available capacity thresholds and triggers resource allocation proactively, preventing service interruptions rather than reacting after they occur. This shifts the system from reactive to proactive resource management.
Solution Approach 2:
The patent applies beforehand cushioning by maintaining a buffer of available capacity through proactive resource provisioning. The system forecasts future usage and ensures sufficient resources are allocated in advance to cushion against demand spikes, creating a safety margin that prevents service interruptions. This cushioning approach ensures service continuity even when usage patterns deviate from historical averages.
3Reliability
If resource provisioning is increased to meet peak demands, then service availability is improved, but resource waste occurs during low usage periods
Solution Approach 1:
The patent implements dynamics by making resource provisioning adaptive and flexible rather than static. The system continuously monitors forecasted usage patterns and dynamically adjusts resource allocation levels accordingly. Resources are provisioned based on predicted实际需求 rather than fixed over-provisioning, allowing the system to optimize between service availability and resource efficiency by adjusting capacity in response to changing demand forecasts.
Solution Approach 2:
The system applies parameter changes by adjusting resource allocation parameters based on forecasted usage patterns. Rather than maintaining fixed resource levels, the system modifies provisioning parameters dynamically according to predicted demand, using historical data and forecasting models to optimize resource levels for different usage scenarios, thereby reducing waste during low-demand periods while maintaining availability during peak periods.
Data Source
AI summary
Techniques for automated capacity management in computing systems are disclosed herein. In one embodiment, a method includes generating multiple time series models each representing predicted usage levels of the computing resource based on historical usage levels of the computing resource. The method can then include selecting, from the generated multiple time series models, one of the time series models that has a combined value of a forecast error and a forecast churn smaller than the other generated time series models. The method can further includes determining a future usage level of the computing resource in the computing system at the future time point using the selected one of the time series models and allocating and provisioning an amount of the computing resource in the computing system in accordance with the predicted future usage level of the computing resource at the future time point.