Predictive Container Quantity Adjustment for Application Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing container resource allocation methods for applications fail to adjust container quantities in a timely manner, leading to potential application running faults and limitations due to post-adjustment based on traffic changes.
Innovation Solution
A method and apparatus for predicting future traffic and container quantity distribution using pre-trained models, allowing for timely adjustment of container quantities based on historical data, target utilization, and reinforcement learning to ensure optimal resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If post-adjustment is performed based on traffic change, then container quantity is adjusted according to actual traffic, but adjustment timing is delayed causing application running limitations or faults
Solution Approach 1:
The patent applies preliminary action by predicting future traffic trends using historical data and machine learning models before traffic changes actually occur. The system forecasts traffic patterns and proactively adjusts container quantities in advance, transforming the reactive post-adjustment approach into a proactive pre-adjustment strategy that prevents application running limitations and faults before they happen.
2Productivity
If container quantity is increased to handle traffic increase, then application service capability is improved, but resource waste occurs when traffic decreases
Solution Approach 1:
The patent implements dynamics by establishing a dynamic container quantity adjustment mechanism that continuously adapts to changing traffic conditions. Using real-time traffic monitoring and predictive modeling, the system dynamically optimizes container quantities to match actual service demands, ensuring high productivity when traffic increases while minimizing resource waste when traffic decreases, thus achieving flexible and efficient resource allocation.
Solution Approach 2:
The patent applies feedback by implementing a closed-loop control system that continuously monitors actual traffic, compares it with predicted traffic, and adjusts container quantities accordingly. The system uses feedback from traffic patterns and application performance metrics to refine predictions and optimize container allocation, ensuring that resource allocation remains aligned with actual service needs and preventing both over-provisioning and under-provisioning.
3Device complexity
If fixed container quantity is allocated, then resource allocation is simple, but application running faults occur when traffic fluctuates
Solution Approach 1:
The patent applies parameter changes by transitioning from a static container quantity parameter to a dynamic parameter that changes based on traffic conditions. The system uses machine learning models to predict traffic parameters and adjusts container quantities accordingly, allowing the resource allocation system to adapt to traffic fluctuations while maintaining manageable complexity through automated prediction and adjustment mechanisms.
Data Source
AI summary
Implementations of methods and apparatuses for container quantity adjustment are disclosed. In an implementation, a method includes: determining historical data of an application, predicting traffic distribution of the application within a predetermined duration based on the historical data by using a pre-trained traffic prediction model, predicting container quantity distribution of the application within the predetermined duration based on the traffic distribution and a predetermined target utilization of a container by using a pre-trained quantity prediction model, and adjusting a container quantity of the application at each of a plurality of predetermined moments within the predetermined duration based on the container quantity distribution.