CN109710376A - The dynamic dispatching method and device of container cluster management system - Google Patents
The dynamic dispatching method and device of container cluster management system Download PDFInfo
- Publication number
- CN109710376A CN109710376A CN201811519798.5A CN201811519798A CN109710376A CN 109710376 A CN109710376 A CN 109710376A CN 201811519798 A CN201811519798 A CN 201811519798A CN 109710376 A CN109710376 A CN 109710376A
- Authority
- CN
- China
- Prior art keywords
- node
- pod
- score
- container cluster
- parameter information
- 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.)
- Pending
Links
Abstract
This application discloses the dynamic dispatching methods and device of a kind of container cluster management system, are related to field of cloud computer technology, for balanced node load, improve cluster stability.This method comprises: obtaining the parameter information of all node and the parameter information of pod in container cluster；All node score and pod score are calculated separately according to the parameter information；The first prediction model and the second prediction model are constructed respectively；The pod score in future time period is predicted by the node score in first prediction model prediction future time period and by second prediction model；If the node score of node is less than predetermined threshold value, pod score under the node is moved into node score higher than under the node of second threshold lower than the pod of first threshold.
Description
Technical field
This application involves field of cloud computer technology more particularly to a kind of dynamic dispatching method of container cluster management system and
Device.
Background technique
The realization of Docker virtualization is Cgroup the and Namespace technology based on linux kernel, is eliminated
Hypervisor layers of expense, therefore it is the virtualization technology of lightweight.Docker can be created and be managed simplerly
Container, therefore Docker container technique can be used, application is packaged.After container reaches certain scale, container is constituted
Cluster meets the resource request of the data service of differentiation and the task of multiplicity to rationally and effectively use the resource of cluster,
A cluster resource scheduling system is needed to be managed scheduling to container cluster.Kubernetes relies on its powerful container layout
The characteristics of ability and light weight are increased income is most widely used in numerous container Trunked Radio Systems.Kubernetes container cluster management
The major function of system includes: to be packaged, instantiated and run to application program using Docker；It is run in a manner of cluster
And container of the management across host；Solve the communication issue etc. between the container run between different hosts.
However the dispatching algorithm that the resource dispatching strategy of Kubernetes and system carry is all more single, can not be suitable for
Complicated application scenarios.Existing scheduling strategy only considered scheduling strategy of the container in creation, when container behaves,
Due in container using occupied resource with time change, will result in that node load is unbalanced, and cluster is unstable.
Summary of the invention
Embodiments herein provides the dynamic dispatching method and device of a kind of container cluster management system, existing for solving
There is in technology scheduling strategy single and the technical issues of scheduling occasion is not easy control.
To achieve the purpose that solve above-mentioned technical problem, embodiments herein is adopted the following technical scheme that
In a first aspect, embodiments herein provides a kind of dynamic dispatching method of container cluster management system, it is described
It include multiple nodes in container cluster, the node is computer equipment, and each node includes multiple containers, the container
For the application program operated in the computer equipment, this method comprises:
The parameter information of all node and the parameter information of pod in container cluster are obtained, the parameter information includes CPU
At least one of in utilization rate, memory usage, image network storage transfer rate and persistence network storage transmission rate；Root
All node score and pod score are calculated separately according to the parameter information, the node score is for indicating the node
Load state, the pod score is used to indicate the computational resource requirements situation of the pod；The node score is constructed respectively
With the corresponding relationship of the time value for obtaining the parameter information as train the first sample data of the first prediction model with
And the building pod score and the corresponding relationship for the time value for obtaining the parameter information are used as training the second prediction model
The second sample data；The first prediction model is obtained by first sample data training and by second sample number
The second prediction model is obtained according to training；By the node score in first prediction model prediction future time period and pass through institute
State the pod score in the second prediction model prediction future time period；It, will be described if the node score of node is less than predetermined threshold value
Pod score moves to node score higher than under the node of second threshold lower than the pod of first threshold under node.
Second aspect, embodiments herein provides a kind of dynamic scheduler of container cluster management system, described
It include multiple nodes in container cluster, the node is computer equipment, and each node includes multiple containers, the container
For the application program operated in the computer equipment, which is characterized in that it is applied in container cluster during container operation, it should
Device includes:
Acquiring unit, for obtaining the parameter information of the parameter information of all node and pod in container cluster, the ginseng
Number information includes in CPU usage, memory usage, image network storage transfer rate and persistence network storage transmission rate
At least one of；Computing unit, for calculating separately all node score and pod score, institute according to the parameter information
Node score is stated for indicating the load state of the node, the pod score is used to indicate the computational resource requirements of the pod
Situation；Construction unit is made for constructing the corresponding relationship of time value of the node score and the acquisition parameter information respectively
For for train the first prediction model first sample data and the building pod score and obtain the parameter information when
Between the corresponding relationship that is worth as training the second sample data of the second prediction model；Training unit, for passing through described the
The training of one sample data obtains the first prediction model and obtains the second prediction model by second sample data training；In advance
Unit is surveyed, for predicting the node score in future time period by first prediction model and passing through the second prediction mould
Type predicts the pod score in future time period；Scheduling unit, if the node score for node is less than predetermined threshold value, by institute
It states the pod that pod score is lower than first threshold under node and moves to node score higher than under the node of second threshold.
The third aspect, provides a kind of computer readable storage medium for storing one or more programs, it is one or
Multiple programs include instruction, and described instruction makes the container of the computer execution as described in relation to the first aspect when executed by a computer
The dynamic dispatching method of cluster management system.
Fourth aspect provides a kind of computer program product comprising instruction, when described instruction is run on computers
When, so that computer executes the dynamic dispatching method of container cluster management system as described in relation to the first aspect.
5th aspect, a kind of dynamic scheduler of container cluster management system is provided, comprising: processor, memory and
Communication interface；Wherein, communication interface is for the dynamic scheduler and other equipment or network communication；The memory is used for
One or more programs are stored, which includes computer executed instructions, when dynamic scheduler operation,
Processor executes the computer executed instructions of memory storage, to execute container cluster management described in above-mentioned first aspect
The dynamic dispatching method of system.
Embodiments herein provides the dynamic dispatching method and device of a kind of container cluster management system, using in container
The parameter information of all node and pod that get during operation calculate node score and pod score, and obtain and when
Between the sample data of the corresponding relationship that is worth as training prediction model, to predict the node score and pod of future time section
Score solves node load imbalance, container cluster finally according to the pod in the dynamic scheduling container cluster of prediction result
Unstable problem.
Detailed description of the invention
Fig. 1 is a kind of system architecture schematic diagram for container cluster management system that embodiments herein provides；
Fig. 2 is a kind of dynamic dispatching method process signal for container cluster management system that embodiments herein provides
Figure；
Fig. 3 is a kind of dynamic scheduler schematic diagram one for container cluster management system that embodiments herein provides；
Fig. 4 is a kind of dynamic scheduler schematic diagram two for container cluster management system that embodiments herein provides.
Specific embodiment
With reference to the accompanying drawing, to a kind of dynamic dispatching method of container cluster management system provided by the embodiments of the present application and
Device is described in detail.
Term " first " and " second " in the description of the present application and attached drawing etc. be for distinguishing different objects, or
Person is used to distinguish the different disposal to same target, rather than is used for the particular order of description object.
In addition, the term " includes " being previously mentioned in the description of the present application and " having " and their any deformation, it is intended that
It is to cover and non-exclusive includes.Such as the process, method, system, product or equipment for containing a series of steps or units do not have
It is defined in listed step or unit, but optionally further comprising the step of other are not listed or unit, or optionally
It further include the other step or units intrinsic for these process, methods, product or equipment.
It should be noted that in the embodiment of the present application, " illustrative " or " such as " etc. words make example, example for indicating
Card or explanation.Be described as in the embodiment of the present application " illustrative " or " such as " any embodiment or design scheme do not answer
It is interpreted than other embodiments or design scheme more preferably or more advantage.Specifically, " illustrative " or " example are used
Such as " word is intended to that related notion is presented in specific ways.
In the description of the present application, unless otherwise indicated, the meaning of " plurality " is refer to two or more
Embodiments herein provides a kind of dynamic dispatching method of container cluster management system, can be applied to a variety of appearances
In device cluster, such as using Docker container technique to using the container cluster for being packaged and forming.It is wrapped in the container cluster
Multiple physical node node are included, which is computer equipment, such as mobile phone, computer, smartwatch, and each node includes more
A pod, pod are the minimum scheduling unit in container cluster management system, include no less than one container in each pod, should
Dynamic dispatching method is applied in container cluster management system during container operation.
Refering to what is shown in Fig. 1, the embodiment of the present application is illustrated by taking Kubernetes container cluster management system as an example, the appearance
Device cluster management system 10 includes interface server (API Server) 11, memory (Etcd) 12, controller (Controller
Manager) 13 and scheduler (Scheduler) 14, interface server 11 is stored for carrying out data transmission with other equipment
Device 12 is used for memory system data, and controller 13 is run for control system, and scheduler 14 is in scheduling container cluster
pod.Kubelet component 15, monitor (Cadvisor) 16 and log is integrated on each node in container cluster to receive
Collection tool (Fluentd) 17.
Refering to what is shown in Fig. 2, the method comprising the steps of S101-S106:
S101, the parameter information of all node and the parameter information of pod in container cluster are obtained.
In Kubernetes container cluster management system, the monitoring function of monitor 16 has been already integrated into each node
Kubelet component 15 in, kubelet component 15 can be started on each node, be used for process container cluster management system 10
It is issued to the task of this node and manages container, the meeting of kubelet component 15 register node information on interface server 11, and
The resource usage amount of pod and node is periodically reported to monitor 16.However the pod resource occupation of monitor 16 offer single nodes
Situation needs to carry out performance monitoring to all node and whole pod in extensive container cluster.Therefore the present embodiment is adopted
The acquisition of container cluster performance data is realized with Heapster tool.
Container in container cluster can generate many kinds of parameters information during operation, and the parameter information of pod is by the pod packet
The parameter information that the container contained generates during operation calculates to obtain, and the parameter information of node is by available pod all on the node
Parameter information composition.The parameter information include CPU usage, memory usage, image network storage transfer rate and persistently
Change at least one in network storage transmission rate.During container operation, the parameter information of node and pod can constantly occur
The restAPI of monitoring nodes device 16 is called in variation by Heapster, obtains all node and pod in container cluster respectively
Parameter information, the monitoring resource of whole container cluster is provided.
S102, all node score and pod score are calculated separately according to the parameter information.
After obtaining the parameter information of all node and the parameter information of pod by step S101, according to these parameters
Information calculates separately all node score and pod score, which is used to indicate the load state of the node, should
Pod score is used to indicate the computational resource requirements situation of the pod.
Calculation method includes: that calculate separately the parameters information of all node and pod average with corresponding parameter information
The ratio of value, such as the image network storage transfer rate of node and the ratio of image network storage average transmission rate are calculated,
Log system is needed to acquire data, the installation log collection kit 17 on each node is mainly used to collection vessel log letter
Breath, the network storage average transmission rate of each available node and the net of all available node are acquired by log collection tool 17
Network stores average transmission rate summation, calculates the ratio of the two and is recorded in system in the form of timestamp and data.Similarly may be used
Obtain the ratio of the ratio of the other parameters information of the node and the parameters information of pod.It is recorded in acquisition system respectively
The ratio of all parameter informations of node and pod, and according to default weighted value and the ratio calculation each node and pod
Score.
Application program during operation, in case of such as access number of users increase suddenly external cause, most directly
Performance is exactly that CPU usage and memory usage increase, therefore usual CPU usage and memory usage both parameter informations
The weight of setting is higher.Such as it is respectively 0.3 that CPU usage and the weight of memory usage, which can be set, image network storage
The weighted value of transmission rate and persistence network storage transmission rate is respectively 0.2.Therefore node score or pod score are equal to
Ratio+the 0.2* of the ratio+0.2* image network storage transfer rate of the ratio+0.3* memory usage of 0.3*CPU utilization rate
The ratio of persistence network storage transmission rate.
S103, the corresponding relationship for constructing the node score and the time value for obtaining the parameter information respectively are used as training
The corresponding relationship of the time value of the first sample data and building pod score and acquisition of first the prediction model parameter information is made
For for training the second sample data of the second prediction model.
Obtained node score and pod score by step S102, obtained while being recorded in the timestamp in system, this when
Between stamp be to obtain the time value of the parameter information, the corresponding relationship for constructing node score and timestamp is used as training first
The corresponding relationship of the first sample data of prediction model, building pod score and the timestamp is used as training the second prediction mould
Second sample data of type.In the present embodiment as unit of 12 hours, the node score and pod in following 12 hours are predicted
Score, therefore sample data is all the matrix of 12*1.
S104, the first prediction model is obtained by first sample data training and by second sample data training
Obtain the second prediction model.
By the first sample data and the second sample data that obtain step S103 as time series data, time sequence
Column data has the characteristics that noise, unstable, randomness, mainly has automatic return for the prediction technique prior art of this kind of data
Return sliding average ARMA and neural network etc..But these methods have some the disadvantage is that very formidable, and what ARMA included is linear
Behavior do not include for nonlinear factor；And the structure of neural network needs specified in advance or exists using heuritic approach
It is corrected in training process, while the solution that neural network obtains is local optimum rather than global optimum.These deficiencies greatly limit
The applications of these methods in practice.And the above problem is then not present in support vector machines, therefore applies for that the present embodiment uses
Support vector machines are as training pattern.
The sample data in a period of time is collected, if to predict the node score and pod score in following 12 hours,
Then using 12 hours as dimension, all input variables are all the matrixes of 12*1.By first sample data and the second sample data
As training set, model parameter is adjusted using the thought of cross validation.In this training, examined using mean square error MSE
The fitting effect and prediction effect of first prediction model and the second prediction model, it is determined whether be effectively fitted former time series number
According to after testing result is met the requirements, the first prediction model and the second prediction model be can be used in 12 hours futures of prediction
Node score and pod score.
S105, by first prediction model predict future time period in node score and pass through second prediction model
Predict the pod score in future time period.
After having obtained the first prediction model and the second prediction model by step S104, then pass through first prediction model
It predicts the node score in 12 hours following and passes through the pod score in second prediction model, 12 hours futures of prediction.
If the node score of S106, node are less than predetermined threshold value, by pod score under the node lower than first threshold
Pod moves to node score higher than under the node of second threshold.
A predetermined threshold value is set up, if the node score of node is less than the predetermined threshold value, pod under the node is obtained
The pod lower than first threshold is divided to move to node score higher than under the node of second threshold, the first threshold and second threshold can
To be adjusted according to demand.
Optionally, if node score is less than the predetermined threshold value, the minimum pod of pod score under the node is moved to
Under the node of node highest scoring.
Optionally, if running container such as web is applied, since a certain period amount of access increases suddenly, system detection
CPU, internal storage access amount, network transmission speed etc. to some application are uprushed, if going to dispatch at this time, just will appear service delay increasing
Greatly, the case where or even interrupting.If therefore the node score of node is less than predetermined threshold value, can the node score to it is corresponding when
Between preset time before value pod score under the node is moved into node score higher than the second threshold lower than the pod of first threshold
Under the node of value.Such as prediction obtains the node in 12 node scores at following certain day 0-12 o'clock, the node at 8 o'clock is obtained
Point be less than predetermined threshold value, and entire scheduling process needs 30 seconds, then the preset time is 30 seconds, therefore need to 7 points 59 minutes 30
The time point of second executes the scheduling thread.
Embodiments herein provides a kind of dynamic dispatching method of container cluster management system, using in the container runtime
Between the parameter information of the parameter information of all node and pod that gets, calculate node score and pod score, and obtain
Sample data with the corresponding relationship of time value as training prediction model, thus predict future time section node score and
Pod score, finally according to the container in the dynamic scheduling container cluster of prediction result, it is unbalanced to solve node load, container
The unstable problem of cluster.
The embodiment of the present application can carry out the division of functional module or functional unit according to above method example to device,
For example, each functional module of each function division or functional unit can be corresponded to, it can also be by two or more function
It can be integrated in a processing module.Above-mentioned integrated module both can take the form of hardware realization, can also use software
Functional module or the form of functional unit are realized.It wherein, is signal to the division of module or unit in the embodiment of the present application
Property, only a kind of logical function partition, there may be another division manner in actual implementation.
Referring to fig. 3, the embodiment of the present application provides a kind of dynamic scheduler of container cluster management system, can
To be applied to the dynamic dispatching method of container cluster management system as shown above.The dynamic dispatching of the container cluster management system
Device 100 includes:
Acquiring unit 101 should for obtaining the parameter information of the parameter information of all node and pod in container cluster
Parameter information includes CPU usage, memory usage, image network storage transfer rate and persistence network storage transmission rate
At least one of in.
Computing unit 102 should for calculating separately all node score and pod score according to the parameter information
Node score is used to indicate the load state of the node, which is used to indicate the computational resource requirements situation of the pod.
Construction unit 103, for constructing the corresponding relationship of the node score with the time value for obtaining the parameter information respectively
As for training the first sample data of the first prediction model and constructing the time of the pod score and acquisition parameter information
The corresponding relationship of value is as training the second sample data of the second prediction model.
Training unit 104, for by the first sample data training obtain the first prediction model and by this second
Sample data training obtains the second prediction model.
Predicting unit 105 is somebody's turn to do for predicting the node score in future time period by first prediction model and passing through
Second prediction model predicts the pod score in future time period.
Scheduling unit 106 is low by pod score under the node if the node score for node is less than predetermined threshold value
It is moved under node of the node score higher than second threshold in the pod of first threshold.
Optionally, the computing unit 102, is specifically used for:
Calculate separately the parameters information of all node and pod and the ratio of corresponding parameter information average value；According to
The pod score of the node score and pod of default weighted value and each node of the ratio calculation.
Optionally, the scheduling unit 106, is also used to:
If the node score of node is less than predetermined threshold value, the minimum pod of pod score under the node is moved into node
Under the node of highest scoring.
Optionally, the scheduling unit 106, is also used to:
If the node score of node is less than predetermined threshold value, the preset time before corresponding time value will be under the node
Pod score moves to node score higher than under the node of second threshold lower than the pod of first threshold.
Embodiments herein provides a kind of computer readable storage medium for storing one or more programs, one
Or multiple programs include instruction, described instruction makes computer execute container cluster as shown in Figure 2 when executed by a computer
The dynamic dispatching method of management system.
Embodiments herein provides a kind of computer program product comprising instruction, when instruction is run on computers
When, so that computer executes the dynamic dispatching method of container cluster management system as shown in Figure 2.
Fig. 4 shows another possibility of the dynamic scheduler of involved container cluster management system in above-described embodiment
Structural schematic diagram.The device includes: processor 202 and communication interface 203.Processor 202 is used for the movement to device and carries out
Control management, for example, executing above-mentioned acquiring unit 101, computing unit 102, construction unit 103, training unit 104, prediction list
The step of member 105 and scheduling unit 106 execute, and/or other processes for executing techniques described herein.Communication connects
Mouth 203 is for supporting the communication of the device Yu other network entities.Terminal can also include memory 201 and bus 204, storage
Device 201 is used for the program code and data of storage device.
Wherein, above-mentioned processor 202 may be implemented or execute various exemplary in conjunction with described in present disclosure
Logic block, unit and circuit.The processor can be central processing unit, general processor, and digital signal processor is dedicated
Integrated circuit, field programmable gate array or other programmable logic device, transistor logic, hardware component or its
Any combination.It, which may be implemented or executes, combines various illustrative logic blocks described in present disclosure, unit
And circuit.The processor is also possible to realize the combination of computing function, such as combines comprising one or more microprocessors,
DSP and the combination of microprocessor etc..
Memory 201 may include volatile memory, such as random access memory；The memory also may include non-
Volatile memory, such as read-only memory, flash memory, hard disk or solid state hard disk；The memory can also include above-mentioned
The combination of the memory of type.
Bus 204 can be expanding the industrial standard structure (Extended Industry Standard
Architecture, EISA) bus etc..Bus 404 can be divided into address bus, data/address bus, control bus etc..For convenient for table
Show, only indicated with a thick line in Fig. 4, it is not intended that an only bus or a type of bus.
Through the above description of the embodiments, it is apparent to those skilled in the art that, for description
It is convenienct and succinct, only with the division progress of above-mentioned each functional unit for example, in practical application, can according to need and will be upper
It states function distribution to be completed by different functional units, i.e., the internal structure of device is divided into different functional units, to complete
All or part of function described above.The specific work process of the system, apparatus, and unit of foregoing description, before can referring to
The corresponding process in embodiment of the method is stated, details are not described herein.
By dynamic scheduler, the computer-readable storage medium of container cluster management system in the case of this application
Matter, computer program product can be applied to the above method, therefore, can be obtained technical effect see also the above method
Embodiment, details are not described herein for the embodiment of the present application.
It should be noted that above-mentioned each unit can be the processor individually set up, also can integrate controller certain
It is realized in one processor, in addition it is also possible to be stored in the form of program code in the memory of controller, by controller
Some processor calls and executes the function of the above each unit.Processor described here can be a central processing unit
(Central Processing Unit, CPU) or specific integrated circuit (Application Specific
Integrated Circuit, ASIC), or be arranged to implement one or more integrated circuits of the embodiment of the present application.
It should be understood that magnitude of the sequence numbers of the above procedures are not meant to execute suitable in the various embodiments of the application
Sequence it is successive, the execution of each process sequence should be determined by its function and internal logic, the implementation without coping with the embodiment of the present application
Process constitutes any restriction.
Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure
Member and algorithm steps can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions are actually
It is implemented in hardware or software, the specific application and design constraint depending on technical solution.Professional technician
Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed
Scope of the present application.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided herein, it should be understood that disclosed system, apparatus and method, it can be with
It realizes by another way.For example, apparatus embodiments described above are merely indicative, for example, the unit
It divides, only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components
It can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, it is shown or
The mutual coupling, direct-coupling or communication connection discussed can be through some interfaces, the indirect coupling of equipment or unit
It closes or communicates to connect, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, each functional unit in each embodiment of the application can integrate in one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.
Claims (11)
- It include multiple physical node node in the container cluster 1. a kind of dynamic dispatching method of container cluster management system, The node is computer equipment, includes multiple pod in each node, and the pod is in container cluster management system Minimum scheduling unit includes no less than one container in each pod, which is characterized in that be applied to container fortune in container cluster Between the departure date, which comprisesThe parameter information of all node and the parameter information of pod in container cluster are obtained, the parameter information includes that CPU is used At least one of in rate, memory usage, image network storage transfer rate and persistence network storage transmission rate；All node score and pod score are calculated separately according to the parameter information, the node score is for indicating institute The load state of node is stated, the pod score is used to indicate the computational resource requirements situation of the pod；The corresponding relationship for constructing the node score respectively and obtaining the time value of the parameter information is used as training first The first sample data of prediction model and the corresponding relationship of the building pod score and the time value of the acquisition parameter information As for training the second sample data of the second prediction model；The first prediction model is obtained by first sample data training and is obtained by second sample data training Second prediction model；It is predicted by the node score in first prediction model prediction future time period and by second prediction model Pod score in future time period；If the node score of node is less than predetermined threshold value, the pod by pod score under the node lower than first threshold is migrated It is higher than under the node of second threshold to node score.
- 2. the dynamic dispatching method of container cluster management system according to claim 1, which is characterized in that described according to institute It states parameter information and calculates separately all node score and pod score and include:Calculate separately the parameters information of all node and pod and the ratio of corresponding parameter information average value；According to the pod score of the node score and pod of default weighted value and each node of the ratio calculation.
- 3. the dynamic dispatching method of container cluster management system according to claim 1, which is characterized in that if the node Node score be less than predetermined threshold value, then pod score under the node is moved into node score lower than the pod of first threshold Higher than including: under the node of second thresholdIf the node score of node is less than predetermined threshold value, the minimum pod of pod score under the node is moved into node and is obtained Divide under highest node.
- 4. the dynamic dispatching method of container cluster management system according to claim 1, which is characterized in that if the node Node score be less than predetermined threshold value, then pod score under the node is moved into node score lower than the pod of first threshold Higher than under the node of second threshold further include:If the node score of node is less than predetermined threshold value, the preset time before corresponding time value will be under the node Pod score moves to node score higher than under the node of second threshold lower than the pod of first threshold.
- It include multiple physical node node in the container cluster 5. a kind of dynamic scheduler of container cluster management system, The node is computer equipment, includes multiple pod in each node, and the pod is in container cluster management system Minimum scheduling unit includes no less than one container in each pod, which is characterized in that be applied to container fortune in container cluster Between the departure date, described device includes:Acquiring unit, for obtaining the parameter information of the parameter information of all node and pod in container cluster, the parameter letter Breath include CPU usage, memory usage, image network storage transfer rate and persistence network storage transmission rate in extremely One item missing；Computing unit, for calculating separately all node score and pod score according to the parameter information, the node is obtained Divide the load state for indicating the node, the pod score is used to indicate the computational resource requirements situation of the pod；Construction unit, the corresponding relationship conduct of the time value for constructing the node score and the acquisition parameter information respectively For training the first sample data of the first prediction model and the time of the building pod score and the acquisition parameter information The corresponding relationship of value is as training the second sample data of the second prediction model；Training unit, for obtaining the first prediction model by first sample data training and by second sample Data training obtains the second prediction model；Predicting unit, for by first prediction model predict future time period in node score and pass through described second Prediction model predicts the pod score in future time period；Pod score under the node is lower than first if the node score for node is less than predetermined threshold value by scheduling unit The pod of threshold value moves to node score higher than under the node of second threshold.
- 6. the dynamic scheduler of container cluster management system according to claim 5, which is characterized in that the calculating is single Member is specifically used for:Calculate separately the parameters information of all node and pod and the ratio of corresponding parameter information average value；According to the pod score of the node score and pod of default weighted value and each node of the ratio calculation.
- 7. the dynamic scheduler of container cluster management system according to claim 5, which is characterized in that the scheduling is single Member is also used to:If the node score of node is less than predetermined threshold value, the minimum pod of pod score under the node is moved into node and is obtained Divide under highest node.
- 8. the dynamic scheduler of container cluster management system according to claim 5, which is characterized in that the scheduling is single Member is also used to:If the node score of node is less than predetermined threshold value, the preset time before corresponding time value will be under the node Pod score moves to node score higher than under the node of second threshold lower than the pod of first threshold.
- 9. a kind of computer readable storage medium for storing one or more programs, which is characterized in that one or more of journeys Sequence includes instruction, and it is according to any one of claims 1-4 that described instruction when executed by a computer executes the computer The dynamic dispatching method of container cluster management system.
- 10. a kind of computer program product comprising instruction, which is characterized in that when described instruction is run on computers, make Obtain the dynamic dispatching method that the computer executes container cluster management system according to any one of claims 1-4.
- 11. a kind of dynamic scheduler of container cluster management system characterized by comprising processor, memory and communication Interface；Wherein, communication interface is for the dynamic scheduler and other equipment or network communication；The memory is for storing One or more programs, which includes computer executed instructions, when dynamic scheduler operation, processing Device executes the computer executed instructions of memory storage, to execute container cluster according to any one of claims 1-4 The dynamic dispatching method of management system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811519798.5A CN109710376A (en) | 2018-12-12 | 2018-12-12 | The dynamic dispatching method and device of container cluster management system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811519798.5A CN109710376A (en) | 2018-12-12 | 2018-12-12 | The dynamic dispatching method and device of container cluster management system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109710376A true CN109710376A (en) | 2019-05-03 |
Family
ID=66256256
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811519798.5A Pending CN109710376A (en) | 2018-12-12 | 2018-12-12 | The dynamic dispatching method and device of container cluster management system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109710376A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110262899A (en) * | 2019-06-20 | 2019-09-20 | 无锡华云数据技术服务有限公司 | Monitor component elastic telescopic method, apparatus and controlled terminal based on Kubernetes cluster |
| CN110377395A (en) * | 2019-07-03 | 2019-10-25 | 无锡华云数据技术服务有限公司 | A kind of Pod moving method in Kubernetes cluster |
| CN110413380A (en) * | 2019-08-02 | 2019-11-05 | 北京智芯微电子科技有限公司 | The dispatching method of container cluster |
| CN111090523A (en) * | 2019-12-18 | 2020-05-01 | 浪潮云信息技术有限公司 | Resource scheduling method based on dynamic priority under kubernets environment |
| CN111787349A (en) * | 2020-05-19 | 2020-10-16 | 视联动力信息技术股份有限公司 | Data caching method, device, equipment and medium |
| CN112000477A (en) * | 2020-08-21 | 2020-11-27 | 北京浪潮数据技术有限公司 | Method, device, equipment and medium for load balancing in pod |
| CN112187894A (en) * | 2020-09-17 | 2021-01-05 | 杭州谐云科技有限公司 | Container dynamic scheduling method based on load correlation prediction |
| KR20210056655A (en) * | 2019-11-11 | 2021-05-20 | 한국전자기술연구원 | Method for selecting predict-based migration candidate and target on cloud edge |
| KR20220071719A (en) * | 2020-11-24 | 2022-05-31 | 주식회사 솔박스 | Method and apparatus for node labeling |
| US11941453B2 (en) | 2020-12-08 | 2024-03-26 | International Business Machines Corporation | Containerized computing environments |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140066905A (en) * | 2012-11-23 | 2014-06-03 | 현대중공업 주식회사 | An method for analyzing work load |
| CN108009016A (en) * | 2016-10-31 | 2018-05-08 | 华为技术有限公司 | A kind of balancing resource load control method and colony dispatching device |
| CN108182105A (en) * | 2017-12-12 | 2018-06-19 | 苏州大学 | Local dynamic station moving method and control system based on Docker container techniques |
-
2018
- 2018-12-12 CN CN201811519798.5A patent/CN109710376A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140066905A (en) * | 2012-11-23 | 2014-06-03 | 현대중공업 주식회사 | An method for analyzing work load |
| CN108009016A (en) * | 2016-10-31 | 2018-05-08 | 华为技术有限公司 | A kind of balancing resource load control method and colony dispatching device |
| CN108182105A (en) * | 2017-12-12 | 2018-06-19 | 苏州大学 | Local dynamic station moving method and control system based on Docker container techniques |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110262899A (en) * | 2019-06-20 | 2019-09-20 | 无锡华云数据技术服务有限公司 | Monitor component elastic telescopic method, apparatus and controlled terminal based on Kubernetes cluster |
| CN110262899B (en) * | 2019-06-20 | 2021-05-11 | 华云数据控股集团有限公司 | Monitoring component elastic expansion method and device based on Kubernetes cluster and controlled terminal |
| CN110377395A (en) * | 2019-07-03 | 2019-10-25 | 无锡华云数据技术服务有限公司 | A kind of Pod moving method in Kubernetes cluster |
| CN110377395B (en) * | 2019-07-03 | 2021-11-02 | 华云数据控股集团有限公司 | Pod migration method in Kubernetes cluster |
| CN110413380A (en) * | 2019-08-02 | 2019-11-05 | 北京智芯微电子科技有限公司 | The dispatching method of container cluster |
| KR20210056655A (en) * | 2019-11-11 | 2021-05-20 | 한국전자기술연구원 | Method for selecting predict-based migration candidate and target on cloud edge |
| KR102320317B1 (en) * | 2019-11-11 | 2021-11-02 | 한국전자기술연구원 | Method for selecting predict-based migration candidate and target on cloud edge |
| CN111090523A (en) * | 2019-12-18 | 2020-05-01 | 浪潮云信息技术有限公司 | Resource scheduling method based on dynamic priority under kubernets environment |
| CN111787349A (en) * | 2020-05-19 | 2020-10-16 | 视联动力信息技术股份有限公司 | Data caching method, device, equipment and medium |
| CN111787349B (en) * | 2020-05-19 | 2023-12-12 | 视联动力信息技术股份有限公司 | Data caching method, device, equipment and medium |
| CN112000477A (en) * | 2020-08-21 | 2020-11-27 | 北京浪潮数据技术有限公司 | Method, device, equipment and medium for load balancing in pod |
| CN112187894A (en) * | 2020-09-17 | 2021-01-05 | 杭州谐云科技有限公司 | Container dynamic scheduling method based on load correlation prediction |
| CN112187894B (en) * | 2020-09-17 | 2022-06-10 | 杭州谐云科技有限公司 | Container dynamic scheduling method based on load correlation prediction |
| KR20220071719A (en) * | 2020-11-24 | 2022-05-31 | 주식회사 솔박스 | Method and apparatus for node labeling |
| KR102439056B1 (en) * | 2020-11-24 | 2022-09-01 | 주식회사 솔박스 | Method and apparatus for node labeling |
| US11941453B2 (en) | 2020-12-08 | 2024-03-26 | International Business Machines Corporation | Containerized computing environments |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109710376A (en) | The dynamic dispatching method and device of container cluster management system | |
| CN108009016B (en) | Resource load balancing control method and cluster scheduler | |
| US20180203720A1 (en) | Techniques to manage virtual classes for statistical tests | |
| US10523519B2 (en) | Comparative multi-forecasting analytics service stack for cloud computing resource allocation | |
| CN106959894B (en) | Resource allocation method and device | |
| CN108038040A (en) | Computer cluster performance indicator detection method, electronic equipment and storage medium | |
| CN103955398B (en) | Virtual machine coexisting scheduling method based on processor performance monitoring | |
| CN110221915B (en) | Node scheduling method and device | |
| CN113515382B (en) | Cloud resource allocation method and device, electronic equipment and storage medium | |
| CN110506256B (en) | Method for generating source code | |
| CN113157379A (en) | Cluster node resource scheduling method and device | |
| CN114356587B (en) | Calculation power task cross-region scheduling method, system and equipment | |
| CN112256430A (en) | Container deployment method, device, equipment and storage medium | |
| CN111950706A (en) | Data processing method and device based on artificial intelligence, computer equipment and medium | |
| CN107704317A (en) | Smart machine and its application management method and the device with store function | |
| CN114625500A (en) | Method and application for scheduling micro-service application based on topology perception in cloud environment | |
| CN109992408B (en) | Resource allocation method, device, electronic equipment and storage medium | |
| CN115543577B (en) | Covariate-based Kubernetes resource scheduling optimization method, storage medium and device | |
| CN109697105A (en) | A kind of container cloud environment physical machine selection method and its system, virtual resource configuration method and moving method | |
| CN108429704B (en) | Node resource allocation method and device | |
| WO2020086053A1 (en) | Dynamic allocation of computing resources for electronic design automation operations | |
| CN114138453A (en) | Resource optimization allocation method and system suitable for edge computing environment | |
| CN112242959B (en) | Micro-service current-limiting control method, device, equipment and computer storage medium | |
| CN112988383A (en) | Resource allocation method, device, equipment and storage medium | |
| CN108733484A (en) | The method and apparatus of management application |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190503 |