QoS Model Self-Correction for Cloud Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional adaptive resource allocation methods for cloud-based software services are costly and complex, with machine learning requiring extensive historical data and control theory methods involving high overhead for feedback iterations.
Innovation Solution
A cloud software service resource allocation method based on QoS model self-correction, which involves online self-learning, self-correction control, and an improved particle swarm optimization algorithm to adjust resources dynamically and improve QoS prediction model accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning technology is used to construct a QoS prediction model, then the resource allocation can be automated, but a large amount of historical data is required which leads to resource allocation errors
Solution Approach 1:
The patent applies preliminary action by pre-training the QoS prediction model offline using historical data to establish initial parameter values. This allows the system to have automated resource allocation capability from the start while avoiding the need to collect large amounts of historical data online, thus resolving the contradiction between automation and data requirement.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously monitors actual QoS performance and uses this information to correct and update the prediction model parameters online. This feedback loop enables the model to maintain accuracy without requiring large historical datasets, as it adapts using real-time performance data.
2Adaptability or versatility
If control theory methods are used for resource allocation, then the system can adapt dynamically, but a large amount of feedback iterations are required resulting in high overhead for stopping virtual machines
Solution Approach 1:
The patent performs preliminary resource allocation based on the pre-trained QoS prediction model before actual workload execution. This preliminary action provides a good initial resource configuration that reduces the need for extensive feedback iterations during runtime, thereby minimizing virtual machine stopping and restarting operations.
Solution Approach 2:
The patent implements dynamic resource adjustment through a hybrid approach that combines predictive allocation with limited corrective feedback. The system dynamically adjusts resources by making small incremental changes based on actual performance deviations, rather than requiring large-scale feedback iterations, thus reducing virtual machine operational overhead.
3Measurement precision
If the QoS prediction model is inaccurate, then resource allocation errors occur, but collecting more historical data increases system complexity
Solution Approach 1:
The patent implements self-service through an online correction mechanism where the QoS prediction model automatically adjusts its own parameters based on actual performance feedback. The system serves itself by identifying prediction errors and correcting model parameters without requiring external intervention or complex data collection infrastructure, thus improving accuracy while maintaining simplicity.
Solution Approach 2:
The patent improves QoS prediction accuracy by dynamically changing model parameters through online correction rather than collecting more historical data. The system adjusts parameters such as model coefficients and weights based on real-time performance feedback, achieving better accuracy through parameter optimization instead of data accumulation, thereby avoiding increased system complexity.
Data Source
AI summary
The present invention relates to a cloud software service resource allocation method based on QoS model self-correction, the method comprising: step S1: performing online self-learning to obtain a QoS prediction model; step S2: collecting runtime data under a certain workload, and improving the accuracy of the QoS prediction model under a current workload through self-correction control; step S3: constructing a fitness function in combination with the quality of service (QoS) and the cloud resource cost (Cost), and searching for a target resource allocation scheme by using an improved particle swarm optimization algorithm; step S4: comparing the current resource allocation situation with the searched target resource allocation scheme to obtain a difference therebetween, and then adjusting resources according to a certain proportion; and step S5: repeating steps S2 to S4 until the current resource allocation situation is the same as the target resource allocation scheme which means that resource adjustment is completed. The present invention can realize a best resource allocation when the QoS prediction model is inaccurate.