Load-time window-oriented DQN-based cloud software resource adaptive allocation method
A technology of software resources and time windows, applied in neural learning methods, software simulation/interpretation/simulation, computer components, etc., can solve the problem that it is difficult to collect training data and is not suitable for the reality with variable workload and service requests World cloud environment and other issues
Pending Publication Date: 2022-03-04
FUZHOU UNIV
View PDF0 Cites 1 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, in real-world cloud environments, it is difficult to collect enough training data to support ML-based methods
Therefore, these classical methods of resource allocation for cloud-based software services may not be well suited for real-world cloud environments with variable workloads and service requests
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0027] The technical solution of the present invention will be specifically described below in conjunction with the accompanying drawings.
[0028] A load-time window-based adaptive allocation method for DQN cloud software resources of the present invention comprises the following steps:
[0029] Step S1. Construct the training set of the DQN model through the historical operation data. The training set includes the load time window at a certain moment, the virtual machine resource configuration, the system QoS value and the target resource allocation plan at that moment. According to the training set, use the DQN algorithm to train and manage Operation Q value prediction model, management operation Q value prediction model can evaluate the Q value of management operation under different system states;
[0030] Step S2. During operation, use the management operation Q value prediction model obtained in step S1 to predict the Q value of different management operations according...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention relates to a load-time window-oriented DQN-based cloud software resource adaptive allocation method. Most of the existing methods only consider the current condition of the workload, so that the real cloud environment influenced by the fluctuation of the workload cannot be well adapted. According to the method, the current and future workloads in the resource allocation process are considered. Specifically, an original Deep Q-Network (DQN) management operation prediction model based on a workload time window is trained, and the model can be used for predicting proper management operation in different system states. Next, a new feedback control mechanism is designed, and an objective resource allocation plan in a current system state is constructed by iteratively executing management operation; a large number of simulation results show that the prediction precision of the management operation generated by the DRAW method can reach 90.69%. In addition, the DRAW can achieve optimal / nearly optimal performance, and the performance is 3-13% higher than that of other classic methods under different conditions.
Description
technical field [0001] The invention relates to a load-time window-based adaptive allocation method of DQN cloud software resources. Background technique [0002] Over the years, cloud-based software services for a wide range of applications have grown rapidly. However, due to the complex and changeable system state in cloud environment, it is very challenging to achieve a good trade-off between QoS and resource cost when doing resource allocation for cloud-based software services. To meet this challenge, it is crucial to design an adaptive resource allocation method for cloud-based software services. There are some classical methods, such as rule-based, control theory, and machine learning (ML)-based methods, which can solve the problem of cloud resource allocation to a certain extent. The rule-based approach needs to define various rules for software services, which leads to huge cost of rule setting and limits its application in dynamic cloud environments. Furthermore,...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06F9/455G06K9/62G06N3/04G06N3/08
CPCG06F9/45558G06N3/04G06N3/08G06F2009/45591G06F2009/4557G06F18/214
Inventor 陈星张铭豪杨立坚陈佳雯
Owner FUZHOU UNIV