CN106933649B - Virtual machine load prediction method and system based on moving average and neural network - Google Patents

Virtual machine load prediction method and system based on moving average and neural network Download PDF

Info

Publication number
CN106933649B
CN106933649B CN201611191490.3A CN201611191490A CN106933649B CN 106933649 B CN106933649 B CN 106933649B CN 201611191490 A CN201611191490 A CN 201611191490A CN 106933649 B CN106933649 B CN 106933649B
Authority
CN
China
Prior art keywords
load
predicted
time
virtual machine
neural network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201611191490.3A
Other languages
Chinese (zh)
Other versions
CN106933649A (en
Inventor
赵淦森
林成创
张海明
刘创辉
王欣明
林嘉洺
唐华
聂瑞华
汤庸
吴杰超
李振宇
孔祥明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong Guangye Kaiyuan Technology Co ltd
South China Normal University
Original Assignee
Guangdong Guangye Kaiyuan Technology Co ltd
South China Normal University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong Guangye Kaiyuan Technology Co ltd, South China Normal University filed Critical Guangdong Guangye Kaiyuan Technology Co ltd
Priority to CN201611191490.3A priority Critical patent/CN106933649B/en
Publication of CN106933649A publication Critical patent/CN106933649A/en
Application granted granted Critical
Publication of CN106933649B publication Critical patent/CN106933649B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Control Of Metal Rolling (AREA)

Abstract

The invention discloses a virtual machine load prediction method and a virtual machine load prediction system based on a moving average and a neural network, wherein the method comprises the following steps: s1, collecting time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted; s2, obtaining time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to obtain a first load inertia predicted value of the next time interval; s3, acquiring continuous load data before a time period to be predicted, and obtaining a second load inertia predicted value by adopting RBF neural network prediction in combination with the first load inertia predicted value; and S4, outputting the second load inertia predicted value as a final virtual machine load inertia predicted value. The method can reduce the hysteresis of continuous load prediction, improve the timeliness of load prediction, improve the prediction accuracy, has strong adaptability to abnormal conditions, and can be widely applied to the field of load prediction of virtual machines.

Description

Virtual machine load prediction method and system based on moving average and neural network
Technical Field
The invention relates to the field of data processing, in particular to a virtual machine load prediction method and system based on moving average and a neural network.
Background
The noun explains:
infrastructure as a service: an English abbreviation IaaS, which takes the virtual machine, the storage space, the database and other basic facilities as services and provides the services for users in the form of the virtual machine;
virtual machine: the system is a complete computer system which has complete hardware system functions and runs in a completely isolated environment through software simulation;
loading: memory usage, or CPU usage, or network bandwidth usage, etc. of the virtual machine at a certain time;
second moving average method: carrying out primary moving average on the data in the observation period from far to near according to a certain crossing period, and then carrying out secondary moving average calculation on the basis of the primary average value to obtain a final predicted value;
RBF neural network: the neural network using the radial basis function as the hidden node activation function has the characteristics of short training time, good mathematical interpretability and capability of approximating any function.
In an infrastructure as a service (IaaS) cloud, in order to keep the IaaS cloud in an ideal operating state all the time under the condition of saving energy consumption, an IaaS cloud platform needs to automatically and spontaneously adjust virtual machine deployment according to an actual operating condition, and autonomously perform virtual machine scheduling. For the prediction of the future operation load condition of the cloud platform, a powerful quantitative basis can be provided for the scheduling of the virtual machine, and higher activity and flexibility are provided for the scheduling of the virtual machine. Therefore, the load condition of the cloud environment virtual machine in a future period of time is predicted, and the method has the necessity.
In the prior art, most of the load prediction methods adopted by the virtual machines construct a time sequence according to the continuous load conditions of a plurality of moments before the moment to be predicted, and calculate the load prediction condition of the moment to be predicted by adopting a certain mathematical calculation method (such as an exponential smoothing method, a template matching method and the like). However, there is a case where there is a sudden change in the load in the cloud environment, i.e., its change is not a smooth increase or decrease in a gentle step; however, in a private IaaS environment, the load of the virtual machine has a certain periodic regularity, and has a higher or lower load level inertia at a specific time. In the prior art, load level inertia at each specific moment is ignored, so that a prediction module cannot predict sudden change of the moment to be predicted in time, and prediction has certain hysteresis. In addition, if a time-division prediction mode is only adopted, for a certain abnormal day, the prediction method cannot acquire the abnormal condition of the day, prediction calculation is still performed according to the original rule, and the obtained prediction result can be separated from the actual virtual machine running condition of the day, so that serious over-high or over-low prediction is caused, resources are seriously wasted, or the cloud platform is caused to be fluctuated due to insufficient prediction.
Generally, the existing virtual machine load prediction method has certain hysteresis, poor timeliness, and low prediction accuracy and flexibility.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a virtual machine load prediction method based on a moving average and a neural network, and the invention also aims to provide a virtual machine load prediction system based on a moving average and a neural network.
The technical scheme adopted by the invention for solving the technical problems is as follows:
the virtual machine load prediction method based on the moving average and the neural network comprises the following steps:
s1, collecting time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted;
s2, obtaining time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to obtain a first load inertia predicted value of the next time interval;
s3, acquiring continuous load data before a time period to be predicted, and obtaining a second load inertia predicted value by adopting RBF neural network prediction in combination with the first load inertia predicted value;
and S4, outputting the second load inertia predicted value as a final virtual machine load inertia predicted value.
Further, step S1 specifically includes:
s11, acquiring load peak data of the virtual machine in the period of time in the past several days as time-period historical load data from the historical load data of the virtual machine after obtaining the time information of the period of time to be predicted;
and S12, collecting load peak data of a plurality of continuous time intervals before the time interval to be predicted as continuous load data.
Further, step S2 specifically includes:
s21, acquiring time-interval historical load data of a time interval to be predicted;
s22, according to the time-interval historical load data, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method;
and S23, taking the prediction calculation result as a first load inertia prediction value of the next time interval.
Further, in step S22, it specifically includes:
according to the time-interval historical load data, the load inertia of the virtual machine is predicted and calculated by using a quadratic moving average method according to the following formula:
Figure GDA0001408670170000031
in the above formula, xt、xt-1、…xt-(n-1)Sequentially represents load peak data of the past several days in a period to be predicted, t represents a period, n represents a step length,
Figure GDA0001408670170000032
representing a moving average over a period t,
Figure GDA0001408670170000033
the quadratic moving average over the period t is represented, and the linear trend prediction model of the prediction calculation process is as follows:
Figure GDA0001408670170000034
wherein the content of the first and second substances,
Figure GDA0001408670170000035
indicates a trend prediction value, atRepresenting an estimate of intercept, btRepresents an estimated value of the slope, and T represents a coefficient.
Further, step S3 specifically includes:
s31, acquiring a load peak sequence of a plurality of continuous time periods before a time period to be predicted, and acquiring a first load inertia predicted value;
s32, constructing an RBF neural network, and taking the load peak value sequence and the first load inertia predicted value as an input sequence of the RBF neural network;
and S33, obtaining a second load inertia predicted value after calculation by adopting an RBF neural network.
The other technical scheme adopted by the invention for solving the technical problem is as follows:
the virtual machine load prediction system based on the moving average and the neural network comprises:
the acquisition module is used for acquiring time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted;
the load level inertia prediction module is used for acquiring time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to acquire a first load inertia prediction value of the next time interval;
the continuous load prediction module is used for acquiring continuous load data before a time period to be predicted, and acquiring a second load inertia prediction value by adopting RBF neural network prediction in combination with the first load inertia prediction value;
and the result output module is used for taking the second load inertia predicted value as a final virtual machine load inertia predicted value and outputting the final virtual machine load inertia predicted value.
Further, the acquisition module specifically includes:
the first acquisition submodule is used for acquiring load peak data of a period of time in the past several days from historical load data of the virtual machine as time-interval historical load data after time information of the period of time to be predicted is acquired;
and the second acquisition submodule is used for acquiring load peak data of a plurality of continuous time periods before the time period to be predicted as continuous load data.
Further, the load level inertia prediction module specifically includes:
the historical data acquisition submodule is used for acquiring time-interval historical load data of a time interval to be predicted;
the moving average calculation submodule is used for carrying out prediction calculation on the load inertia of the virtual machine by using a secondary moving average method according to the time-interval historical load data;
and the result obtaining submodule is used for taking the prediction calculation result as a first load inertia prediction value of the next time interval.
Further, the moving average calculation sub-module is specifically configured to:
according to the time-interval historical load data, the load inertia of the virtual machine is predicted and calculated by using a quadratic moving average method according to the following formula:
Figure GDA0001408670170000041
in the above formula, xt、xt-1、…xt-(n-1)Sequentially represents load peak data of the past several days in a period to be predicted, t represents a period, n represents a step length,
Figure GDA0001408670170000042
representing a moving average over a period t,
Figure GDA0001408670170000043
the quadratic moving average over the period t is represented, and the linear trend prediction model of the prediction calculation process is as follows:
Figure GDA0001408670170000044
wherein the content of the first and second substances,
Figure GDA0001408670170000051
indicates a trend prediction value, atRepresenting an estimate of intercept, btRepresents an estimated value of the slope, and T represents a coefficient.
Further, the result output module specifically includes:
the continuous data acquisition submodule is used for acquiring a load peak value sequence of a plurality of continuous time periods before a time period to be predicted and acquiring a first load inertia predicted value;
the neural network construction sub-module is used for constructing the RBF neural network and taking the load peak value sequence and the first load inertia predicted value as an input sequence of the RBF neural network;
and the neural network calculation submodule is used for obtaining a second load inertia predicted value after calculation is carried out by adopting the RBF neural network.
The invention has the beneficial effects that: the virtual machine load prediction method based on the moving average and the neural network comprises the following steps: s1, collecting time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted; s2, obtaining time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to obtain a first load inertia predicted value of the next time interval; s3, acquiring continuous load data before a time period to be predicted, and obtaining a second load inertia predicted value by adopting RBF neural network prediction in combination with the first load inertia predicted value; and S4, outputting the second load inertia predicted value as a final virtual machine load inertia predicted value. The method can reduce the hysteresis of continuous load prediction, improve the timeliness of load prediction and improve the prediction accuracy, and has strong adaptability to abnormal conditions.
The invention has the following beneficial effects: the invention discloses a virtual machine load prediction system based on a moving average and a neural network, which comprises the following components: the acquisition module is used for acquiring time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted; the load level inertia prediction module is used for acquiring time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to acquire a first load inertia prediction value of the next time interval; the continuous load prediction module is used for acquiring continuous load data before a time period to be predicted, and acquiring a second load inertia prediction value by adopting RBF neural network prediction in combination with the first load inertia prediction value; and the result output module is used for taking the second load inertia predicted value as a final virtual machine load inertia predicted value and outputting the final virtual machine load inertia predicted value. The system can reduce the hysteresis of continuous load prediction, improve the timeliness of load prediction and improve the prediction accuracy, and has strong adaptability flexibility to abnormal conditions.
Drawings
The invention is further illustrated by the following figures and examples.
FIG. 1 is a flow chart of a method for virtual machine load prediction based on moving averages and neural networks of the present invention;
FIG. 2 is a system block diagram of the moving average and neural network based virtual machine load prediction system of the present invention;
FIG. 3 is a diagram of an exemplary architecture of a neural network used in the method for predicting the load of a virtual machine based on moving averages and the neural network of the present invention;
fig. 4 is a schematic diagram of a neuron model of the neural network structure of fig. 3.
Detailed Description
Example one
Referring to fig. 1, a virtual machine load prediction method based on a moving average and a neural network includes the steps of:
s1, collecting time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted;
s2, acquiring time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to obtain a first load inertia predicted value P1 of the next time interval;
s3, acquiring continuous load data before a time period to be predicted, and obtaining a second load inertia predicted value P2 by adopting RBF neural network prediction in combination with the first load inertia predicted value P1;
and S4, outputting the second load inertia predicted value P2 as a final virtual machine load inertia predicted value P.
Further, step S1 specifically includes:
s11, acquiring load peak data of the virtual machine in the period of time in the past several days as time-period historical load data from the historical load data of the virtual machine after obtaining the time information of the period of time to be predicted; for example, when load inertia in a period of 5: 00-5: 59 needs to be predicted, load peak data of 5: 00-5: 59 in past days is read as time-period historical load data.
S12, collecting load peak value data of a plurality of continuous moments or continuous time periods before a time period to be predicted as continuous load data, for example, when loads in the time period of 5: 00-5: 59 need to be predicted, reading load peak values of a plurality of continuous time periods of 4: 00-4: 59, 3: 00-3: 59 … … and the like as the continuous load data.
Further, step S2 specifically includes:
s21, acquiring time-interval historical load data of a time interval to be predicted;
s22, according to the time-interval historical load data, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method;
and S23, taking the prediction calculation result as a first load inertia prediction value P1 of the next time interval.
Further, in step S22, it specifically includes:
according to the time-interval historical load data, the load inertia of the virtual machine is predicted and calculated by using a quadratic moving average method according to the following formula:
Figure GDA0001408670170000071
in the above formula, xt、xt-1、…xt-(n-1)Sequentially represents load peak data of the past several days in a period to be predicted, t represents a period, n represents a step length,
Figure GDA0001408670170000072
representing a moving average over a period t,
Figure GDA0001408670170000073
the quadratic moving average over the period t is represented, and the linear trend prediction model of the prediction calculation process is as follows:
Figure GDA0001408670170000074
wherein the content of the first and second substances,
Figure GDA0001408670170000075
indicates a trend prediction value, atRepresenting an estimate of intercept, btRepresents an estimated value of the slope, and T represents a coefficient.
Further, step S3 specifically includes:
s31, acquiring a load peak value sequence { W1, W2, … …, Wn } of a plurality of continuous time intervals before a time interval to be predicted, and acquiring a first load inertia predicted value P1;
s32, constructing an RBF neural network, and taking the load peak value sequence and the first load inertia predicted value P1 as input sequences { W1, W2, … …, Wn, P1} of the RBF neural network;
and S33, calculating by adopting an RBF neural network to obtain a second load inertia predicted value P2.
The structure of the neural network is shown in fig. 3, and includes an input layer, a hidden layer, and an output layer, and in general, the neural network is a massively parallel distributed processor composed of simple processing units. Neurons are the basic units of information processing of neural networks. The neural network explores and stores the rule of the load sequence through the neurons. Each neuron is connected to other neurons, and when a neuron reaches an excitation threshold, the neuron is activated to generate a signal that is transmitted to the next layer of neurons to which it is connected. The basic structural model of a neuron is shown in fig. 4. The model shown in the figure receives input signals from n other neurons, passes the input signals through weighted connections, and then produces neuron outputs through processing of an activation function phi within the neuron.
Neural networks are composed of multiple layers of neurons. A single hidden layer neural network refers to a neural network with only one hidden layer. In a neural network of a single hidden layer, the mathematical expression of the relationship between the network output yt and the network input is:
Figure GDA0001408670170000081
wherein aj represents a network connection weight from the hidden layer to the output layer, bij is a connection weight from the input layer to the hidden layer, p is the number of neurons of the input layer, q is the number of neurons of the hidden layer, and phi is an activation function of a neural network node.
In this embodiment, the RBF neural network is a special neural network in which the hidden layer activation function phi is a radial basis function. The radial basis function adopted in the invention is a Gaussian function, and the formula is as follows:
Figure GDA0001408670170000082
r denotes the distance to a fixed point and σ is the spreading constant of the radial basis function.
The method can amplify the signal that the load suddenly rises or falls at the moment to be predicted, reduce the hysteresis of continuous load prediction, improve the timeliness of load prediction and improve the prediction accuracy. And the load level inertia prediction of the step S2 and the continuous load prediction of the step S3 are combined for prediction, the continuous load prediction can adjust the result of the load level inertia prediction module according to the actual load condition of the day, the influence of the prediction module on the relative difference between the predicted load and the actual load (namely the actual load value-the predicted load value) of the load can be reduced on the day with abnormal load, and the adaptive flexibility of the prediction method on the day with abnormal load is improved.
Example two
Referring to fig. 2, the present invention further provides a virtual machine load prediction system based on a moving average and a neural network, including:
the acquisition module is used for acquiring time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted;
the load level inertia prediction module is used for acquiring time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to acquire a first load inertia predicted value P1 of the next time interval;
the continuous load prediction module is used for acquiring continuous load data before a time period to be predicted, and obtaining a second load inertia prediction value P2 by adopting RBF neural network prediction in combination with the first load inertia prediction value P1;
and the result output module is used for outputting the second load inertia predicted value P2 as a final virtual machine load inertia predicted value.
Further, the acquisition module specifically includes:
the first acquisition submodule is used for acquiring load peak data of a period of time in the past several days from historical load data of the virtual machine as time-interval historical load data after time information of the period of time to be predicted is acquired; for example, when load inertia in a period of 5: 00-5: 59 needs to be predicted, load peak data of 5: 00-5: 59 in past days is read as time-period historical load data.
The second acquisition submodule is used for acquiring load peak value data of a plurality of continuous time periods before the time period to be predicted as continuous load data, for example, when loads in the time period of 5: 00-5: 59 need to be predicted, load peak values of a plurality of continuous time periods of 4: 00-4: 59, 3: 00-3: 59 … … and the like are read as the continuous load data.
Further, the load level inertia prediction module specifically includes:
the historical data acquisition submodule is used for acquiring time-interval historical load data of a time interval to be predicted;
the moving average calculation submodule is used for carrying out prediction calculation on the load inertia of the virtual machine by using a secondary moving average method according to the time-interval historical load data;
and the result acquisition submodule is used for taking the prediction calculation result as the first load inertia prediction value P1 of the next time interval.
Further, the moving average calculation sub-module is specifically configured to:
according to the time-interval historical load data, the load inertia of the virtual machine is predicted and calculated by using a quadratic moving average method according to the following formula:
Figure GDA0001408670170000091
in the above formula, xt、xt-1、…xt-(n-1)Sequentially represents load peak data of the past several days in a period to be predicted, t represents a period, n represents a step length,
Figure GDA0001408670170000092
representing a moving average over a period t,
Figure GDA0001408670170000093
representing a second movement within the period tMean, and the linear trend prediction model of the prediction calculation process is as follows:
Figure GDA0001408670170000101
wherein the content of the first and second substances,
Figure GDA0001408670170000102
indicates a trend prediction value, atRepresenting an estimate of intercept, btRepresents an estimated value of the slope, and T represents a coefficient.
Further, the result output module specifically includes:
the continuous data acquisition submodule is used for acquiring load peak value sequences { W1, W2, … … and Wn } of a plurality of continuous time intervals before a time interval to be predicted and acquiring a first load inertia predicted value P1;
the neural network construction sub-module is used for constructing the RBF neural network, and the load peak value sequence and the first load inertia predicted value are used as input sequences { W1, W2, … …, Wn, P1} of the RBF neural network;
and the neural network calculation submodule is used for obtaining a second load inertia predicted value P2 after calculation is carried out by adopting the RBF neural network.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (8)

1. The virtual machine load prediction method based on the moving average and the neural network is characterized by comprising the following steps of:
s1, collecting time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted;
s2, obtaining time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to obtain a first load inertia predicted value of the next time interval;
s31, acquiring a load peak sequence of a plurality of continuous time periods before a time period to be predicted, and acquiring a first load inertia predicted value;
s32, constructing an RBF neural network, and taking the load peak value sequence and the first load inertia predicted value as an input sequence of the RBF neural network;
s33, calculating by using an RBF neural network to obtain a second load inertia predicted value;
s4, taking the second load inertia predicted value as a final virtual machine load inertia predicted value and outputting the final virtual machine load inertia predicted value;
the RBF neural network is a special neural network with hidden layer activation function as radial basis function
Figure FDA0002549645810000011
The formula of (1) is as follows:
Figure FDA0002549645810000012
where r denotes the distance to a fixed point and σ is the spreading constant of the radial basis function.
2. The method for predicting the load of the virtual machine based on the moving average and the neural network according to claim 1, wherein the step S1 specifically includes:
s11, acquiring load peak data of the virtual machine in the period of time in the past several days as time-period historical load data from the historical load data of the virtual machine after obtaining the time information of the period of time to be predicted;
and S12, collecting load peak data of a plurality of continuous time intervals before the time interval to be predicted as continuous load data.
3. The method for predicting the load of the virtual machine based on the moving average and the neural network according to claim 1, wherein the step S2 specifically includes:
s21, acquiring time-interval historical load data of a time interval to be predicted;
s22, according to the time-interval historical load data, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method;
and S23, taking the prediction calculation result as a first load inertia prediction value of the next time interval.
4. The method for predicting the load of the virtual machine based on the moving average and the neural network according to claim 3, wherein the step S22 is specifically as follows:
according to the time-interval historical load data, the load inertia of the virtual machine is predicted and calculated by using a quadratic moving average method according to the following formula:
Figure FDA0002549645810000013
Figure FDA0002549645810000021
in the above formula, xt、xt-1、…xt-(n-1)Sequentially represents load peak data of the past several days in a period to be predicted, t represents a period, n represents a step length,
Figure FDA0002549645810000022
representing a moving average over a period t,
Figure FDA0002549645810000023
the quadratic moving average over the period t is represented, and the linear trend prediction model of the prediction calculation process is as follows:
Figure FDA0002549645810000024
Figure FDA0002549645810000025
Figure FDA0002549645810000026
wherein the content of the first and second substances,
Figure FDA0002549645810000027
indicates a trend prediction value, atRepresenting an estimate of intercept, btRepresents an estimated value of the slope, and T represents a coefficient.
5. The virtual machine load prediction system based on the moving average and the neural network is characterized by comprising the following steps:
the acquisition module is used for acquiring time-interval historical load data of a time interval to be predicted and continuous load data before the time interval to be predicted;
the load level inertia prediction module is used for acquiring time-interval historical load data of a time interval to be predicted, predicting and calculating the load inertia of the virtual machine by using a quadratic moving average method, and predicting to acquire a first load inertia prediction value of the next time interval;
the continuous data acquisition submodule is used for acquiring a load peak value sequence of a plurality of continuous time periods before a time period to be predicted and acquiring a first load inertia predicted value;
the neural network construction sub-module is used for constructing the RBF neural network and taking the load peak value sequence and the first load inertia predicted value as an input sequence of the RBF neural network;
the neural network computing submodule is used for obtaining a second load inertia predicted value after computing by adopting an RBF neural network;
the result output module is used for taking the second load inertia predicted value as a final virtual machine load inertia predicted value and outputting the final virtual machine load inertia predicted value;
the RBF neural network is a special neural network with hidden layer activation function as radial basis function
Figure FDA0002549645810000028
The formula of (1) is as follows:
Figure FDA0002549645810000029
where r denotes the distance to a fixed point and σ is the spreading constant of the radial basis function.
6. The virtual machine load prediction system based on the moving average and the neural network according to claim 5, wherein the collection module specifically includes:
the first acquisition submodule is used for acquiring load peak data of a period of time in the past several days from historical load data of the virtual machine as time-interval historical load data after time information of the period of time to be predicted is acquired;
and the second acquisition submodule is used for acquiring load peak data of a plurality of continuous time periods before the time period to be predicted as continuous load data.
7. The virtual machine load prediction system based on the moving average and the neural network according to claim 5, wherein the load level inertia prediction module specifically comprises:
the historical data acquisition submodule is used for acquiring time-interval historical load data of a time interval to be predicted;
the moving average calculation submodule is used for carrying out prediction calculation on the load inertia of the virtual machine by using a secondary moving average method according to the time-interval historical load data;
and the result obtaining submodule is used for taking the prediction calculation result as a first load inertia prediction value of the next time interval.
8. The moving average and neural network based virtual machine load prediction system of claim 7, wherein the moving average computation sub-module is specifically configured to:
according to the time-interval historical load data, the load inertia of the virtual machine is predicted and calculated by using a quadratic moving average method according to the following formula:
Figure FDA0002549645810000031
Figure FDA0002549645810000032
in the above formula, xt、xt-1、…xt-(n-1)Sequentially represents load peak data of the past several days in a period to be predicted, t represents a period, n represents a step length,
Figure FDA0002549645810000033
representing a moving average over a period t,
Figure FDA0002549645810000034
the quadratic moving average over the period t is represented, and the linear trend prediction model of the prediction calculation process is as follows:
Figure FDA0002549645810000035
Figure FDA0002549645810000036
Figure FDA0002549645810000037
wherein the content of the first and second substances,
Figure FDA0002549645810000038
indicates a trend prediction value, atRepresenting an estimate of intercept, btRepresents an estimated value of the slope, and T represents a coefficient.
CN201611191490.3A 2016-12-21 2016-12-21 Virtual machine load prediction method and system based on moving average and neural network Active CN106933649B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611191490.3A CN106933649B (en) 2016-12-21 2016-12-21 Virtual machine load prediction method and system based on moving average and neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611191490.3A CN106933649B (en) 2016-12-21 2016-12-21 Virtual machine load prediction method and system based on moving average and neural network

Publications (2)

Publication Number Publication Date
CN106933649A CN106933649A (en) 2017-07-07
CN106933649B true CN106933649B (en) 2020-11-06

Family

ID=59444027

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611191490.3A Active CN106933649B (en) 2016-12-21 2016-12-21 Virtual machine load prediction method and system based on moving average and neural network

Country Status (1)

Country Link
CN (1) CN106933649B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107391230B (en) * 2017-07-27 2021-03-09 苏州浪潮智能科技有限公司 Implementation method and device for determining load of virtual machine
CN107689891A (en) * 2017-09-01 2018-02-13 广西大学 Container cloud elastic supply time point Forecasting Methodology and system under extensive prominent visit situation
CN107404409B (en) * 2017-09-01 2020-01-31 广西大学 Method and system for predicting quantity of container cloud flexible supply containers for sudden load
CN108255581A (en) * 2018-01-15 2018-07-06 郑州云海信息技术有限公司 A kind of load based on neural network model determines method, apparatus and system
CN110362377B (en) * 2018-04-09 2023-05-30 阿里巴巴集团控股有限公司 Scheduling method and device of virtual machine
CN109117269A (en) * 2018-07-26 2019-01-01 郑州云海信息技术有限公司 A kind of distributed system dispatching method of virtual machine, device and readable storage medium storing program for executing
CN109522093A (en) * 2018-11-16 2019-03-26 国家电网有限公司 Electric power cloud virtual machine load predicting method
CN110321810A (en) * 2019-06-14 2019-10-11 华南师范大学 Single channel signal two-way separation method, device, storage medium and processor
CN111913780B (en) * 2020-07-15 2022-01-11 中国人民解放军陆军工程大学 Resource prediction and scheduling method in cloud computing
CN113705840A (en) * 2021-09-23 2021-11-26 重庆允成互联网科技有限公司 Equipment predictive maintenance method and device, computer equipment and storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103748993B (en) * 2009-05-31 2011-02-16 北京理工大学 A kind of host load prediction method based on multisequencing combination
CN102479339A (en) * 2010-11-24 2012-05-30 香港理工大学 Method and system for forecasting short-term wind speed of wind farm based on hybrid neural network
CN103577268A (en) * 2012-08-07 2014-02-12 复旦大学 Self-adaptive resource supply method based on application load
CN104239968A (en) * 2014-09-02 2014-12-24 浙江大学 Short-term load predicting method based on quick fuzzy rough set

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103748993B (en) * 2009-05-31 2011-02-16 北京理工大学 A kind of host load prediction method based on multisequencing combination
CN102479339A (en) * 2010-11-24 2012-05-30 香港理工大学 Method and system for forecasting short-term wind speed of wind farm based on hybrid neural network
CN103577268A (en) * 2012-08-07 2014-02-12 复旦大学 Self-adaptive resource supply method based on application load
CN104239968A (en) * 2014-09-02 2014-12-24 浙江大学 Short-term load predicting method based on quick fuzzy rough set

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
云计算平台的自适应资源供给;赵淦森等;《电信科学》;20120131;第31-37页 *

Also Published As

Publication number Publication date
CN106933649A (en) 2017-07-07

Similar Documents

Publication Publication Date Title
CN106933649B (en) Virtual machine load prediction method and system based on moving average and neural network
CN110223517B (en) Short-term traffic flow prediction method based on space-time correlation
Kuchuk et al. Improving big data centers energy efficiency: Traffic based model and method
CN112000459B (en) Method for expanding and shrinking capacity of service and related equipment
CN113242568A (en) Task unloading and resource allocation method in uncertain network environment
CN105760213B (en) The early warning system and method for resources of virtual machine utilization rate under cloud environment
CN114915630B (en) Task allocation method, network training method and device based on Internet of Things equipment
CN113778691B (en) Task migration decision method, device and system
Mehdi et al. Cloud traffic prediction based on fuzzy ARIMA model with low dependence on historical data
CN113746696A (en) Network flow prediction method, equipment, storage medium and device
CN113852432A (en) RCS-GRU model-based spectrum prediction sensing method
CN113051130B (en) Mobile cloud load prediction method and system of LSTM network combined with attention mechanism
CN105740431A (en) Time series data prediction method based on multi-layer period exponent hierarchy
CN115065992A (en) Edge calculation-based cooperative sensing and resource allocation method
CN115714820A (en) Distributed micro-service scheduling optimization method
CN116541106A (en) Computing task unloading method, computing device and storage medium
CN113837383B (en) Model training method and device, electronic equipment and storage medium
CN117436485A (en) Multi-exit point end-edge-cloud cooperative system and method based on trade-off time delay and precision
CN107608781A (en) A kind of load predicting method, device and network element
Cui et al. Multi-Agent Reinforcement Learning Based Cooperative Multitype Task Offloading Strategy for Internet of Vehicles in B5G/6G Network
Xu et al. A fusion model for CPU load prediction in cloud computing
Jeon et al. Intelligent resource scaling for container based digital twin simulation of consumer electronics
CA3117168A1 (en) System and method for neural time series preprocessing
Tang et al. Forecasting Network Traffic at Large Time Scales by Using Dual-Related Method
CN113297540A (en) APP resource demand prediction method, device and system under edge Internet of things agent service

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant