Data center server power consumption management and optimization method based on reinforcement learning
A data center and reinforcement learning technology, applied in electrical digital data processing, digital data processing components, instruments, etc., can solve the problem of inability to optimize data center load distribution and power consumption distribution, power consumption changes, and inability to adapt to system architecture service quality and reliability requirements, to achieve the effect of easy engineering deployment, reducing overall power consumption, and improving classification accuracy
Active Publication Date: 2019-02-12
HANGZHOU DIANZI UNIV
View PDF5 Cites 19 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
In addition, due to the heterogeneity of the hardware configuration and performance of different servers, the power consumption of different servers with different types and different intensities of loads also has a large change.
Therefore, the traditional empirical, semi-automatic and semi-manual data center power consumption management methods can no longer adapt to the dynamically scalable, self-adaptive system architecture and strict service quality and reliability requirements of the data center in the elastic cloud computing environment. It is also impossible to dynamically optimize the load distribution and power consumption distribution of the data center according to the load change, thereby reducing the energy consumption of the entire data center
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
[0032] The present invention will be further described below in conjunction with accompanying drawing, please refer to figure 1 . Such as figure 1 As shown, SR is the collection of task loads reaching the data center, SQ is the existing task queue in the data center, and SP is the server node. The present invention predicts the arrival interval of the request at the next moment according to the historical data of the task load request of the data center, divides the interval time into three categories: "long", "short" and "unknown", and calculates the possible distribution of future task load under known conditions Probability, select the category with the highest probability as the prediction result, and predict the arrival interval time of the next task on the current server node, so as to judge whether the current server node can be put into sleep state. The "unknown" classification result only indicates a conservative estimate when the probability difference between the ...
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 discloses a data center server power consumption management and optimization method based on reinforcement learning. The invention solves the problem of power consumption management andoptimization of a data center by using a reinforcement learning method, and sequentially makes a decision by continuously observing the load arrival, load distribution and power consumption use information of the random system of the data center. namely by selecting one action from the set of available actions to make a decision according to the observed state at each time. Decision makers make new decisions based on newly observed states and repeat them over and over again. The invention can directly optimize the load distribution strategy of the data center on line without any prior knowledge, thereby reducing the overall operation power consumption of the data center.
Description
technical field [0001] The invention relates to an automatic method for managing and distributing system resources in a data center, in particular to a method for distributing multiple virtual machines with perceivable power consumption on a data center server. Background technique [0002] With the development of cloud computing, big data, machine learning and other technologies, in order to meet the needs of a large number of users for data storage, processing and intelligent analysis, the scale of data centers is becoming larger and larger, which also leads to high energy consumption costs. No more. Energy consumption has become a key issue restricting the scalability, reliability and service quality of data centers. In recent years, power consumption management and optimization in data centers has become a widely concerned issue in both industry and academia. While ensuring the normal business and safe and reliable operation of the data center, it is extremely importan...
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): G06F9/455G06F9/48G06F1/329
CPCG06F1/329G06F9/45558G06F9/4856G06F2009/4557Y02D10/00
Inventor 蒋从锋崔中江樊甜甜仇烨亮万健张纪林殷昱煜任祖杰
Owner HANGZHOU DIANZI UNIV