CN109298914A - A kind of Docker based on three-tier architecture and virtual machine initial placement method - Google Patents
A kind of Docker based on three-tier architecture and virtual machine initial placement method Download PDFInfo
- Publication number
- CN109298914A CN109298914A CN201811190209.3A CN201811190209A CN109298914A CN 109298914 A CN109298914 A CN 109298914A CN 201811190209 A CN201811190209 A CN 201811190209A CN 109298914 A CN109298914 A CN 109298914A
- Authority
- CN
- China
- Prior art keywords
- virtual machine
- server
- docker
- resource
- container
- 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
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000012216 screening Methods 0.000 claims abstract description 25
- 238000005457 optimization Methods 0.000 claims abstract description 23
- 230000008569 process Effects 0.000 claims abstract description 21
- 230000015654 memory Effects 0.000 claims description 73
- 230000006870 function Effects 0.000 claims description 58
- 238000004364 calculation method Methods 0.000 claims description 22
- 238000013507 mapping Methods 0.000 claims description 18
- 239000011159 matrix material Substances 0.000 claims description 18
- 238000003860 storage Methods 0.000 claims description 10
- 239000012634 fragment Substances 0.000 claims description 5
- 238000013468 resource allocation Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 230000002349 favourable effect Effects 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000007726 management method Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 239000002023 wood Substances 0.000 claims description 2
- 238000005265 energy consumption Methods 0.000 abstract description 7
- 238000012856 packing Methods 0.000 description 8
- 230000006872 improvement Effects 0.000 description 6
- 238000011160 research Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000003213 activating effect Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
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
- 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
- G06F2009/45562—Creating, deleting, cloning virtual machine instances
-
- 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
- G06F2009/4557—Distribution of virtual machine instances; Migration and load balancing
-
- 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
- G06F2009/45583—Memory management, e.g. access or allocation
-
- 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
- G06F2009/45595—Network integration; Enabling network access in virtual machine instances
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Stored Programmes (AREA)
Abstract
The invention discloses a kind of Docker based on three-tier architecture and virtual machine initial placement method, including the following contents: based on theoretical placement (Docker Placement and Virtual Machine Placement, the abbreviation DVMP) restricted model for proposing Docker, VM to PM under Docker-VM-PM three-tier architecture of vector vanning.Server resource utilization rate optimization aim is introduced on this basis, constructs DVMP initial placement model.The virtual machine sequence for meeting institute's Prescribed Properties simultaneously according to DVMP initial placement model discrimination designs the activated state server series that minimum virtual machine computation rule screening meets constraint.The last decision for carrying out optimal VM, PM respectively in the constraint of two classes according to fitness function completes the initial placement process of Docker and virtual machine based on three-tier architecture until all Docker case.Initial placement method proposed by the present invention efficiently solves the optimization problem of data center resource utilization rate under Docker-VM-PM three-tier architecture, reduces energy consumption.
Description
Technical Field
The invention relates to a method for initially placing a Docker and a virtual machine based on a three-layer architecture (Docker-VM-PM, DVP for short), which is suitable for solving the problem of optimizing the utilization rate of data center resources.
Background
In recent years, cloud computing technology has been rapidly developed, and data centers have been rapidly expanded, which in turn causes severe energy consumption and high operation costs. Data shows that the energy consumption of the data center is closely related to the resource utilization rate of the physical servers, so that the resource utilization rate of each server of the data center is improved, and the reduction of the number of the activated physical servers becomes a key strategy for reducing the energy consumption and the operation cost of the data center. The industry and academia mainly utilize virtual machine technology to realize the fine granularity distribution of physical resources. Therefore, how to reasonably place the virtual machine, so as to improve the utilization rate of physical resources and reduce energy consumption becomes a current research hotspot.
With the mature development of the Docker technology, the PaaS cloud service implementation mode is gradually changed from a VM-PM two-layer architecture mainly based on a VM to a Docker-VM-PM three-layer architecture mainly based on a Docker container, and physical resources are allocated in a finer granularity manner. In the present stage, optimization research aiming at the utilization rate of server resources mainly focuses on the research on the placement of a Docker under a Docker-VM/PM framework and the research on the placement of a virtual machine under a VM-PM framework. The placement of Docker can be regarded as a boxing process, and is a study on how to select an appropriate box (virtual machine or physical machine) for a boxing object (Docker) according to resource requirements. The virtual machine placement comprises virtual machine placement position decision and placement position transformation based on the virtual machine online migration technology. According to different application scenes, the virtual machine placement position decision is divided into virtual machine initial placement position decision and virtual machine aggregation placement position decision. The initial placement of the virtual machine is a process of researching how to select a proper target physical machine in an unloaded cloud data center according to a request of a virtualization type for resources, and can also be described as an online boxing problem which takes the virtual machine as a boxing object and a server as a box, or a variant multidimensional boxing problem, a multidimensional vector boxing problem, a semi-online boxing problem and the like.
If the optimal initial placement decision from Docker to VM only depends on the resource usage of VM and ignores the resource influence on indirectly-loaded PM caused by the establishment of the placement relationship, the container is easily distributed in the data center, and an unnecessary number of servers may be used for load sharing, which results in inefficient resource utilization. At present, no research is available on designing a method for initially placing a container and a virtual machine (Docker-VM) by taking a container and virtual machine and server (VM-PM) initially placing processes into a cooperative consideration. However, research on a virtual machine placement method under a VM-PM two-layer architecture places a certain reference value on initialization of a Docker and a virtual machine under a DVP three-layer architecture. The following 3 patent documents provide different policy approaches for virtual machine placement:
1. a virtual machine placement method in a cloud data center based on an ant colony optimization algorithm. (patent No. CN201711266803.1)
2. A virtual machine initial placement strategy method based on a Rendervous hash algorithm is disclosed. (patent No. CN201710491606.3)
3. A virtual machine placement method and device based on distributed storage. (patent No. CN 201610054601.X)
Document 1 provides an ant colony optimization algorithm-based virtual machine placement method in a cloud data center, which solves the VMP problem through the ant colony optimization algorithm, and finds a virtual machine placement method when a virtual machine request arrives, so that the total energy consumption of the cloud data center is minimized, and the total network bandwidth required by communication between virtual machines is reduced.
Document 2 provides a virtual machine placement strategy method based on a Rendezvous hash algorithm. The method comprises the steps of firstly defining a physical host set existing in a data center and a virtual machine set needing to be initialized and placed, constructing a virtual hierarchical structure by analyzing the number of hosts, wherein the virtual hierarchical structure comprises two parts, namely a virtual hierarchical node and a real host node cluster, and finally reasonably scheduling physical resources by comprehensively considering factors such as load balance, virtual machine host performance, data center energy consumption and the like.
Document 3 provides a virtual machine placement method and apparatus based on distributed storage. Whether the virtual machine is placed or not is determined by judging the available resources of the physical server where the first storage block of the storage volume used by the virtual machine is located. The virtual machine is preferentially placed on the first physical server where the storage volume used by the virtual machine is located, so that resources consumed by network transmission during the work of the virtual machine are reduced, and the utilization rate of the storage resources is improved.
However, the methods described in the above documents mainly have the following problems: document 1, by setting a physical machine list capable of placing a virtual machine without violating constraints, selects a server to be placed according to an ant colony optimization algorithm, and although it is possible to improve resource utilization, it does not effectively save the number of active servers. Document 2 considers CPU and memory resources of the servers in the virtual machine placement process, but also does not consider saving the number of active state servers in the placement process. Document 3 focuses on solving the problems of poor virtual machine performance and low utilization rate and poor expansibility of a storage array of a virtualization system caused by an improper virtual machine placement method, which is different from the goal of optimizing the resource utilization rate of a data center.
Disclosure of Invention
The invention aims to provide a method for initially placing a Docker and a Virtual Machine based on a three-layer framework. And then, screening virtual machine sequences simultaneously meeting all boxing constraints according to the DVMP initial placement model, and designing a minimum virtual machine calculation rule to screen an active state server sequence meeting the constraints. And finally, the optimal small box VM and the optimal large box PM are respectively decided in the two types of sequences according to the fitness function until all Docker packing is completed, so that the initial placement process of the Docker and the virtual machine based on the three-layer architecture is completed, and the utilization rate of a server CPU and a memory of the data center is effectively improved.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
a method for initially placing a Docker and a virtual machine based on a three-layer architecture comprises the following steps:
step1, establishing a Docker and virtual machine placement constraint model based on a Docker-VM-PM three-layer framework, wherein the Docker-VM-PM is referred to as DVP for short, the virtual machine placement is referred to as DVMP for short, the problem of placement of a Docker to a virtual machine and a virtual machine to a server under the DVP three-layer framework is described as a two-layer vector collaborative boxing problem, and the Docker is used as the actual load of the virtual machine and the server to reasonably limit a virtual machine placement scheme;
step2, based on the DVMP constraint model, optimally establishing a Docker and virtual machine initial placement model, referred to as DVMP initial placement model for short, for the resource utilization rate of the server under the DVP three-layer framework;
step3, initializing a newly added scene of the data center and screening VM (virtual machine), initializing the configuration and the quantity of VMs (virtual machine models) and PMs (physical management systems) of the data center by adopting uniformly distributed random functions, and screening out a virtual machine sequence which simultaneously meets all constraint conditions of a DVMP (dynamic virtual machine model) initial placement model from all virtual machines of the data center according to the CPU (central processing unit) and memory resource requests of a newly added container after the algorithm starts;
step 4, designing an optimal VM fitness function and optimal VM selection, calculating the fitness of each virtual machine in the virtual machine sequence according to the designed optimal VM fitness function, and selecting the virtual machine with the maximum fitness value as the optimal virtual machine of the container to be put in based on the BF principle;
step 5, designing a minimum VM calculation rule and PM screening, selecting a minimum virtual machine capable of containing a newly-added container from a virtual machine template selectable in a data center by taking a CPU and a memory resource required by the container as a reference, and screening an active state server sequence meeting the constraint according to the resource requirement of the minimum virtual machine;
step 6, designing an optimal PM fitness function and an optimal PM selection, if the sequence of the screened active state servers is empty, starting a new server, otherwise, sequentially calculating the fitness of each server in the server sequence according to the server fitness function, and selecting the server with the maximum fitness as the optimal server of the container and the minimum virtual machine according to the BF principle to put in;
step 7, starting a server, if the activated state server sequence is empty after the minimum virtual machine resource constraint screening, starting a new server and randomly selecting a virtual machine template supported by a data center, and creating a virtual machine instance as the optimal VM and the optimal PM of the container; if no new container is added, the Docker and VM initial placement method is terminated until a new container comes to activate the scheduler to operate the placement algorithm again.
The method is further improved in that in the step1, a DVMP constraint model with a three-layer framework is constructed, and a DVMP constraint condition based on a vector packing theory is provided according to a two-layer vector packing theory and VM and PM resource usage calculation reference;
the two-layer vector collaborative boxing theory comprises two-layer vector boxing models from Docker to VM and from VM to PM, and is specifically described as two-layer vector boxing models from Docker boxing objects to small boxes VM and from VM small boxes to PM large boxes, and the boxing objects are placed along opposite angles of the boxes;
the method comprises the steps that on the basis of calculation of the resource usage of the VM and the PM, the allocation amount of all Docker resources put into a certain VM does not exceed the resource application amount when the VM is created, when the resource upper limit of the VM is reached, a new container is put into other VMs or a new VM is created for carrying, and the total application amount of all the VM resources put into a certain server cannot exceed the maximum supply amount of the server resources;
DVMP constraint conditions based on the vector encasement theory, specifically comprising integer constraint conditions, integrity constraint conditions, resource utilization condition constraint conditions and resource usage constraint conditions;
the integer constraint is embodied by the following equation:
wherein x isij、yri、zrjRespectively representing elements representing a placement state in a virtual machine to server position mapping matrix X, a container to virtual machine position mapping matrix Y and a container to server position mapping matrix Z, and further judging whether an ith virtual machine is placed on a jth server, whether an ith container is placed on the ith virtual machine and whether the ith container runs on the jth server; v. ofi、pjThe state variables of the virtual machine and the server are represented, the values of the state variables are 1 and 0, and whether the virtual machine i and the server j are in the running state or not is respectively represented;
the integrity constraint is embodied by the following formula:
the integrity constraint formula adopts an accumulation calculation mode, the placement states of any container, virtual machine, all virtual machines and servers are accumulated and summed, and the summation result is limited to 1, so that abnormal conditions except 1 are avoided, and all containers and virtual machines are placed at reasonable positions in a cluster;
the resource utilization condition constraint conditions comprise virtual machine resource utilization condition constraint conditions and server resource utilization condition constraint conditions, and the virtual machine resource utilization condition constraint conditions ensure that the quantity of resources requested by a Docker each time does not exceed the total quantity of resources applied when the virtual machine is created by limiting the resource allocation condition of the virtual machine, and are specifically represented by the following formula:
wherein,respectively representing the amount of CPU and memory resources applied when the r-th Docker is created,the total amount of resources of a CPU and a memory which can be provided by the corresponding ith virtual machine; the server resource utilization condition constraint condition ensures that the quantity of resources requested by the virtual machine each time does not exceed the total quantity of resources provided by the server by limiting the resource allocation condition of the server, and is specifically represented by the following formula:
wherein,respectively representing the CPU and memory resource amount applied when the ith virtual machine is established, the total amount of resources of a CPU and a memory which can be provided by the corresponding jth server;
the resource usage constraint condition limits the calculation mode of the VM and PM resource usage in the DVP framework, and is specifically described as that the resource usage of the VM and the PM at any moment takes the total amount of all Docker container resources borne by the VM and the PM as a statistical basis; the specific resource usage constraint is expressed as the following equation:
wherein, yriAnd zrjState variables for the Docker to virtual machine position mapping matrix Y and the Docker to server position mapping matrix Z respectively,respectively representing the CPU and memory resource amount applied when the r-th Docker is created, taking the total resource amount of all Docker containers borne by the VM and the PM at any moment as a statistical basis,respectively representing the amount of CPU and memory resources used by the container in the jth server,respectively representing the amount of CPU and memory resources used by the container in the ith virtual machine.
The further improvement of the invention is that in step2, a three-layer framework DVMP initial placement model is constructed, the specific content is represented by introducing a resource utilization degree optimization function of the data center resource utilization rate on the basis of the DVMP model, and the optimization target of the data center resource utilization rate is represented by the following formula:
wherein,indicating the number of servers in active state, all server state variables P passing through the data centerjThe sum is calculated from the following equation:
wherein,andrespectively representing the idle CPU and memory resource amount of the jth server; the optimization goal of improving the resource utilization rate of the data center consists of the product of minimizing the number of active state servers and maximizing the resource margin of all the servers in each dimension,andis obtained by the following formula:
wherein,representing the amount of resources server j is used by the container,the total amount of resources of the CPU and the memory which can be provided by the corresponding server j;
in the optimization objective, the process of the method,the part is to prevent the container sets with the same scale and configuration from appearing, and the smaller the number of the used servers is, the more concentrated the container placement is, the higher the resource utilization rate is;the influence of the resource allowance is partially expressed, and the product of the CPU and the memory resource allowance is used as the effective evaluation of the multidimensional resource in the placement process based on the multidimensional resource normalization processing mode provided by Wood T; clothes used when placement is finishedWhen the number of the servers is the same, the larger the sum of the product of the surplus resources of the servers in the data center is, the larger the specification of the surplus resources of some servers is, the more favorable the new large-specification container can be accommodated, therefore, the larger the value of the part is, the better the part is, and α in the optimization target isE、βEIs two constants used for representing the weight of the first part and the second part in the resource utilization function, and the weight satisfies αE＞0＞βEAnd α, andEshould be much greater than | βEThe smaller the sum of the two parts of the resource utilization function, the more appropriate the placement scheme.
The further improvement of the invention is that in step3, the data center new scene initialization and VM screening are specifically represented as the following steps:
step 1: initializing the configuration and the quantity of VMs and PMs in the data center by adopting uniformly distributed random functions, wherein the configuration and the quantity of the VMs and the PMs are preset as reference quantity of CPU resourcesReference quantity of memory resourcesCoacting according to the following formula:
wherein,andthe resource demand of CPU and memory of the container i is expressed, and the numerical value is expressed by the resource demand of CPU and memory and the total CPU and memory of the serverThe percentage of the amount of the storage resource represents;
step 2: and (3) screening virtual machine sequences which simultaneously meet all constraint conditions of the DVMP constraint model in the step (1) and the step (2) from all virtual machines in the data center according to the CPU and memory resource requests of the newly added container.
The further improvement of the invention is that in step 4, an optimal VM fitness function and an optimal VM selection are designed, the specific contents are expressed in that the resource relation among the Docker, the VM and the PM is considered globally, the resource influence of VM placement on the PM is considered in the virtual machine fitness function, the situation that a relatively concentrated placement scheme on a VM level is actually distributed and deployed on a PM physical level to cause a large number of resource fragments and start more servers to share load is prevented, and the optimal VM fitness function is expressed as the following formula:
wherein,representing the CPU and memory resources of the newly added container,and the total amount of resources occupied by the container of the virtual machine i and the total amount of resources provided by the virtual machine i are respectively represented, α and β are two constants and are used for representing the weight of the first part and the second part in the optimal virtual machine fitness function, pbfit (j) represents the optimal server fitness function, and the optimal server fitness function is obtained by the following formula:
wherein,andrepresenting the total amount of resources server j actually occupied by the container,corresponding to the total amount of CPU and memory resources that server j is capable of providing,andis obtained by the following formula, represents the CPU, memory resource requirements, z, of container rrjState variables for the Docker to server location mapping matrix Z:
and respectively calculating the fitness of each virtual machine in the screened virtual machine sequence by adopting a virtual machine fitness function, and selecting the virtual machine with the maximum fitness as the optimal virtual machine of the container to be put in.
The further improvement of the present invention is that, in step 5, a minimum VM calculation rule and PM screening are designed to avoid a situation that a randomly created virtual machine cannot be loaded by any active state server and more servers are started, so that the minimum VM rule is determined as a critical point for starting a new server, which is specifically represented by the following steps:
step 1: inputting the CPU request quantity and the memory request quantity of the newly added container;
step 2: comparing the CPU and memory amount of the newly added container and each virtual machine template one by one, and selecting the minimum virtual machine capable of accommodating the new container;
step 3: determining whether the minimum virtual machine can be loaded in the current activation state of the data center, and determining whether the minimum virtual machine is a critical point for starting a new server;
and screening out the activated state server sequence meeting the constraint according to the minimum virtual machine rule, and carrying out the next step of operation.
The further improvement of the invention is that in step 6, an optimal PM fitness function and an optimal PM selection are designed, which are specifically shown in the steps that when all the active state servers cannot contain loads, a new server is started and a virtual machine bearing container is randomly created, and the virtual machine and the server are used as the optimal VM and PM of the container; the optimal PM decision process of the minimum virtual machine is that the fitness calculation is carried out on the screened activated state servers in sequence according to the fitness function of the optimal server, and the virtual machine with the maximum fitness is selected as the optimal server of the container and the minimum virtual machine; the fitness calculation of each server adopts an optimal PM fitness function PBfit (j) to make a decision, which is specifically represented by the following formula:
compared with the prior art, the invention has the following beneficial technical effects:
1. the invention provides a virtual machine initial placement scheme and a virtual machine initial placement method based on a DVP (socket-VM-PM) three-layer framework constructed by taking a container-to-virtual machine (socket-VM) initial placement process and a virtual machine-to-server (VM-PM) initial placement process into synergistic consideration.
2. The invention provides a virtual machine fitness function model aiming at the packing process from a Docker packing object to an optimal VM small box, simultaneously considers the resource relation among Docker, VM and PM, and after a new container is put into the optimal VM, the VM and a host machine PM can reach the maximum target of the utilization rate of each dimension resource, so that the resource fragments of the VM and the PM are reduced, and the starting number of servers is saved.
3. Aiming at the scene that Docker box objects are subjected to the optimal PM large box packing, the invention provides the minimum virtual machine rule as a judgment basis for judging whether a new server is started or not. And on the basis of CPU and memory resources required by the newly added container, selecting the minimum virtual machine capable of containing the container from the virtual machine templates selectable by the data center, so that the number of active state servers is saved.
4. The invention provides a server fitness function model aiming at a scene from a VM small box packing object to an optimal PM large box packing object, and the actual resource quantity of a server at a certain time is taken as optimal VM and PM decision data, so that the starting quantity of the server is saved, and the decision precision is improved.
Drawings
FIG. 1 is a schematic overall flow diagram of the process of the present invention.
FIG. 2 is a schematic diagram of DVMP two-tier binning under a two-dimensional resource.
FIG. 3 is a two-dimensional resource down-container to virtual machine binning diagram.
FIG. 4 is a schematic diagram of virtual machine to server binning under a two-dimensional resource.
Detailed Description
The following describes the initial placement method of the Docker and the virtual machine based on the three-tier architecture in detail with reference to the accompanying drawings and embodiments.
As shown in fig. 1, in the embodiment of the present application, a method for initially placing a Docker and a virtual machine based on a three-tier architecture includes:
step 101: establishing a Docker and virtual Machine Placement (DVMP for short) constraint model based on a Docker-VM-PM three-layer architecture, wherein the Docker-VM-PM is DVP for short, describing the Placement problem of the Docker to the virtual Machine and the virtual Machine to the server under the DVP three-layer architecture as a two-layer vector collaborative boxing problem, placing boxing objects along the diagonal line of a box, and decomposing the boxing objects into a boxing process from the Docker to the VM shown in FIG. 3 and a boxing process from the VM to the PM shown in FIG. 4 as a schematic diagram of two-layer boxing of DVMP under two-dimensional resources shown in FIG. 2. The CPU and memory resource allocation amounts of the box objects VM1 and VM2 in fig. 3 are (cc1, mc1), (cc2, mc2), the allocation amount of all Docker resources put into the VM cannot exceed the resource application amount at the time of VM creation, and when the VM resource upper limit is reached, a new container needs to be put into another VM or a new VM needs to be created for carrying. The CPU and memory resource application amounts when the small boxes VM1, 2 in fig. 4 are created are (cvm1, mvm1), (cvm2, mvm2), and the total application amount of all VM resources put into the server cannot exceed the maximum resource supply amount of the server.
The two-layer vector collaborative boxing theory comprises two-layer vector boxing models from Docker to VM and from VM to PM, and is specifically described as two-layer vector boxing models from Docker boxing objects to small boxes VM and from VM small boxes to PM large boxes, wherein the boxing objects are placed along opposite corners of the boxes instead of being placed in close proximity to each other between the traditional boxing objects. The reason for placing along the diagonal of the box is that the computing resources of the virtual machine and the server in different dimensions only can provide the total amount of use of a single upper limit, when the virtual machine occupies the resources of a certain dimension of the server, although the resources of other dimensions of the server are abundant, a new virtual machine cannot be loaded, and small cubes similar to the interior of the box are placed along the diagonal of the box and are not adjacent to each other.
The specific content of the calculation reference of the resource usage of the VM and the PM is that the allocation amount of all Docker resources put into a certain VM cannot exceed the resource application amount when the VM is created, and when the resource upper limit of the VM is reached, a new container needs to be put into other VMs or a new VM is created for carrying. The total application amount of all VM resources put into a certain server cannot exceed the maximum supply amount of the server resources.
On the basis of two-layer vector collaborative boxing, four types of placement constraint conditions are set, including an integer constraint condition, an integrity constraint condition, a resource utilization condition constraint condition and a resource usage constraint condition. Docker is used as the actual load of the virtual machine and the server, and the rationality of the virtual machine placement scheme is limited.
The integer constraint condition avoids the condition that the positions of the container and the virtual machine do not exist by placing the container and the virtual machine on the virtual machine and the server with integer numbers, effectively limits the scale of a solution space of a placing problem, and is expressed by the following formula:
wherein x isij、yri、zrjThe mapping matrix X from the virtual machine to the server, the mapping matrix Y from the container to the virtual machine and the mapping matrix Z from the container to the server represent elements representing the placing state, and further whether the ith virtual machine is placed on the jth server, whether the ith container is placed on the ith virtual machine and whether the ith container runs on the jth server. v. ofi、pjThe state variables of the virtual machine and the server are represented, and the values 1 and 0 respectively represent whether the virtual machine i and the server j are in the running state or not. The situation that the placing positions of the container and the virtual machine cannot generate decimal is guaranteed through the state constraint of 0 and 1 integers, and the placing scheme which cannot exist is eliminated, so that the placing problem is limited in the solution space of the integers.
The integrity constraint condition ensures that all containers and virtual machines are placed by requiring any container and virtual machine to be placed on only one virtual machine and one server, and is expressed by the following formula:
the formula adopts an accumulation calculation mode, the placement states of any container, virtual machine, all virtual machines and servers are accumulated and summed, and the summation result is limited to 1, so that abnormal conditions except 1 are avoided, and all containers and virtual machines are placed at reasonable positions in the cluster. As the placement constraint model is applied to different placement scenarios, the X, Y and Z matrix sizes representing the virtual machine to server, container to virtual machine, container to physical machine location mappings will vary.
The resource utilization condition constraint conditions comprise a virtual machine resource utilization condition constraint condition and a server resource utilization condition constraint condition, and are specifically represented by the following formula:
the former two formulas represent the constraint conditions of the virtual machine resource utilization, and the latter two formulas represent the server resourcesSource utilization constraints. Wherein,respectively representing the amount of CPU and memory resources applied when the r-th Docker is created,the total amount of resources of the CPU and the memory which can be provided by the ith virtual machine. The above formula limits the CPU and memory resources of the Docker on each virtual machine, and ensures that the quantity of the CPU and the memory requested by the Docker do not exceed the total quantity of the CPU and the memory which can be provided by the virtual machine.Respectively representing the CPU and memory resource amount applied when the ith virtual machine is established,corresponding to the total amount of resources of the CPU and the memory which can be provided by the jth server. The above formula restricts the CPU and memory resources of the virtual machine on each server, and ensures that the number of virtual CPUs and the memory requested by the virtual machine do not exceed the total amount of CPUs and memories that can be provided by the server.
The resource usage constraint is expressed as the following equation:
wherein,andrepresenting the total amount of resources server j actually occupied by the container,andrepresenting the total amount of resources that virtual machine i actually occupies by the container,respectively representing the amount of CPU and memory resources applied when the r-th Docker is created, taking the total amount of all Docker container resources borne by the VM and the PM as a statistical basis for the resource usage amount of the VM and the PM at any moment, and yriAnd zrjState variables of a Docker to virtual machine position mapping matrix Y and a Docker to server position mapping matrix Z are respectively.
Step 102: and establishing a DVMP initial placement model based on a three-layer architecture. Introducing a resource utilization optimization function of the data center resource utilization on the basis of the DVMP constraint model, wherein the resource utilization optimization function is specifically expressed as the following optimization targets:
wherein,indicating the number of servers in active state, all server state variables P passing through the data centerjThe sum is calculated from the following equation:
wherein,andrespectively representing the idle CPU and memory resource amount of the jth server; the formula is an optimization target for improving the resource utilization rate of the data center, and consists of two parts of the product of the minimum number of the active state servers and the maximum resource margin of all the servers in each dimension,andis obtained by the following formula:
wherein,representing the amount of resources server j is used by the container,corresponding to the total amount of resources of the CPU and the memory which can be provided by the jth server. In the optimization objective, the process of the method,partly to prevent the appearance of container sets of the same size and configuration, the smaller number of servers used means that the more centralized the container placement, the higher the resource utilization.When the number of servers used for placement is the same, the larger the sum of the resource margin products of the servers in the data center is, the larger the specification of the residual resources of some servers is, the larger the specification is, the more favorable the capacity of a new large-specification container is, therefore, the larger the value of the part is, the better the value of the part is, α in the optimization target isE、βEIs two constants used to represent the weight of the first part and the second part in the resource utilization function, the weight needs to satisfy αE＞0＞βECondition (iii) and αEIs much larger than | βEThe smaller the sum of the two parts of the resource utilization function, the more appropriate the placement scheme.
Step 103: initializing a newly added scene of the data center and screening VM. Initializing the configuration and the quantity of VMs and PMs in the data center by adopting uniformly distributed random functions, wherein the configuration and the quantity of the VMs and the PMs are preset as reference quantity of CPU resourcesReference quantity of memory resourcesAnd (4) acting together. According to the following formula
The configuration and number of data centers VMs and PMs are initialized. Wherein,andand (3) expressing the resource demand of the CPU and the memory of the container i, wherein the numerical value is expressed by the percentage of the resource demand of the CPU and the memory to the total CPU and memory resource of the server. And screening virtual machine sequences which simultaneously meet all constraint conditions of the DVMP initial placement model in the step 102 from all virtual machines in the data center according to the CPU and memory resource requests of the newly added container.
Step 104: and designing an optimal VM fitness function and an optimal VM selection. The optimal VM fitness function simultaneously considers the resource relation among Docker, VM and PM, after a new container is put into the optimal VM, the VM and the PM of a host machine can achieve the aim that the utilization rate of all the dimensional resources is the maximum, and VM and PM resource fragments are reduced, so that the starting number of servers is saved; and calculating the fitness of each virtual machine in the virtual machine sequence according to the designed optimal VM fitness function, and selecting the virtual machine with the maximum fitness value as the optimal virtual machine of the container to be put into the container based on the BF (Best Fit) principle. Firstly, judging whether a virtual machine sequence is empty, and if the sequence is not empty, designing an optimal VM and PM fitness function, wherein the optimal VM and PM fitness function is specifically represented by the following formula:
wherein,representing the CPU and memory resources of the newly added container,and respectively representing the total amount of resources occupied by the container and the resources provided by the virtual machine iThe source total amount, α and β, are two constants, and are used to represent the weight of the first part and the second part in the optimal virtual machine fitness function, pbfit (j) represents the optimal server fitness function, and is obtained by the following formula:
wherein,andrepresenting the total amount of resources server j actually occupied by the container,corresponding to the total amount of resources of the CPU and the memory that the jth server can provide,andis obtained by the following formula,CPU, memory resource requirements representing container r:
and respectively calculating the fitness of each virtual machine in the screened virtual machine sequence by adopting a virtual machine fitness function, and selecting the virtual machine with the maximum fitness as the optimal virtual machine of the container to be put in.
Step 105: and designing a minimum VM calculation rule and PM screening. The minimum VM calculation rule is designed, and the significance lies in that with the continuous increase of user service containers, the situation that all virtual machines cannot continuously contain loads necessarily occurs, and new virtual machines need to be created to share the loads. In the continuous container loading process, the situation that all the small boxes VM cannot continuously carry the containers necessarily occurs, and new VM sharing loads need to be added. In order to avoid the situation that the increased virtual machines cause more servers to be started to share the load, a minimum virtual machine rule is designed as a critical point for whether a new server is started or not. If the virtual machine sequence is empty, minimum virtual machine calculation is required. And selecting the minimum virtual machine capable of containing the container from the virtual machine templates selectable by the data center on the basis of the CPU and the memory resources required by the newly added container, and taking the minimum virtual machine as a judgment basis for judging whether to start the new server. The method comprises the following steps:
step 1: inputting the CPU request quantity and the memory request quantity of the newly added container;
step 2: comparing the CPU and memory amount of the newly added container and each virtual machine template one by one, and selecting the minimum virtual machine capable of accommodating the new container;
step 3: and determining whether the minimum virtual machine can be loaded in the current activation state of the data center, namely determining whether the minimum virtual machine is a critical point for starting a new server.
And screening the activated state server sequence meeting the constraint according to the minimum virtual machine rule.
Step 106: and designing an optimal PM fitness function and optimal PM selection. The optimal PM fitness function also considers the resource relation among Docker, VM and PM, and after a new container is put into the optimal VM, the VM and the PM of a host machine can achieve the aim that the utilization rate of resources of each dimension is the maximum, so that the resource fragments of the VM and the PM are reduced, and the starting number of servers is saved. And if the screened active state server sequence is empty, starting a new server, otherwise, sequentially calculating the fitness of each server in the server sequence according to the optimal PM fitness function, and selecting the server with the highest fitness as the optimal server of the container and the minimum virtual machine according to the BF principle. The server fitness function is embodied as the following formula:
step 107: and starting the server. And after the minimum virtual machine resource constraint screening is carried out, activating a server sequence to be empty, starting a new server, randomly selecting a virtual machine template supported by the data center, and creating a virtual machine instance as the optimal VM and the optimal PM of the container. If no new container is added, the initial placing method of the Docker and the VM is terminated until a new container resource applies to activate a scheduler to operate the placing algorithm again.
It will be understood by those skilled in the art that the foregoing is only exemplary of the method of the present invention and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (7)
1. A method for initially placing a Docker and a virtual machine based on a three-layer architecture is characterized by comprising the following steps:
step1, establishing a Docker and virtual machine placement constraint model based on a Docker-VM-PM three-layer framework, wherein the Docker-VM-PM is referred to as DVP for short, the virtual machine placement is referred to as DVMP for short, the problem of placement of a Docker to a virtual machine and a virtual machine to a server under the DVP three-layer framework is described as a two-layer vector collaborative boxing problem, and the Docker is used as the actual load of the virtual machine and the server to reasonably limit a virtual machine placement scheme;
step2, based on the DVMP constraint model, optimally establishing a Docker and virtual machine initial placement model, referred to as DVMP initial placement model for short, for the resource utilization rate of the server under the DVP three-layer framework;
step3, initializing a newly added scene of the data center and screening VM (virtual machine), initializing the configuration and the quantity of VMs (virtual machine models) and PMs (physical management systems) of the data center by adopting uniformly distributed random functions, and screening out a virtual machine sequence which simultaneously meets all constraint conditions of a DVMP (dynamic virtual machine model) initial placement model from all virtual machines of the data center according to the CPU (central processing unit) and memory resource requests of a newly added container after the algorithm starts;
step 4, designing an optimal VM fitness function and optimal VM selection, calculating the fitness of each virtual machine in the virtual machine sequence according to the designed optimal VM fitness function, and selecting the virtual machine with the maximum fitness value as the optimal virtual machine of the container to be put in based on the BF principle;
step 5, designing a minimum VM calculation rule and PM screening, selecting a minimum virtual machine capable of containing a newly-added container from a virtual machine template selectable in a data center by taking a CPU and a memory resource required by the container as a reference, and screening an active state server sequence meeting the constraint according to the resource requirement of the minimum virtual machine;
step 6, designing an optimal PM fitness function and an optimal PM selection, if the sequence of the screened active state servers is empty, starting a new server, otherwise, sequentially calculating the fitness of each server in the server sequence according to the server fitness function, and selecting the server with the maximum fitness as the optimal server of the container and the minimum virtual machine according to the BF principle to put in;
step 7, starting a server, if the activated state server sequence is empty after the minimum virtual machine resource constraint screening, starting a new server and randomly selecting a virtual machine template supported by a data center, and creating a virtual machine instance as the optimal VM and the optimal PM of the container; if no new container is added, the Docker and VM initial placement method is terminated until a new container comes to activate the scheduler to operate the placement algorithm again.
2. The initial placement method of a Docker and a virtual machine based on a three-tier architecture as claimed in claim 1, wherein in step1, a three-tier architecture DVMP constraint model is constructed, and a DVMP constraint condition based on a vector binning theory is proposed according to a two-tier vector binning theory and VM, PM resource usage calculation reference;
the two-layer vector collaborative boxing theory comprises two-layer vector boxing models from Docker to VM and from VM to PM, and is specifically described as two-layer vector boxing models from Docker boxing objects to small boxes VM and from VM small boxes to PM large boxes, and the boxing objects are placed along opposite angles of the boxes;
the method comprises the steps that on the basis of calculation of the resource usage of the VM and the PM, the allocation amount of all Docker resources put into a certain VM does not exceed the resource application amount when the VM is created, when the resource upper limit of the VM is reached, a new container is put into other VMs or a new VM is created for carrying, and the total application amount of all the VM resources put into a certain server cannot exceed the maximum supply amount of the server resources;
DVMP constraint conditions based on the vector encasement theory, specifically comprising integer constraint conditions, integrity constraint conditions, resource utilization condition constraint conditions and resource usage constraint conditions;
the integer constraint is embodied by the following equation:
xij,yri,zrj,pj,vi∈{0,1}i∈N,j∈M,r∈D
wherein x isij、yri、zrjRespectively representing elements representing a placement state in a virtual machine to server position mapping matrix X, a container to virtual machine position mapping matrix Y and a container to server position mapping matrix Z, and further judging whether an ith virtual machine is placed on a jth server, whether an ith container is placed on the ith virtual machine and whether the ith container runs on the jth server; v. ofi、pjThe state variables of the virtual machine and the server are represented, the values of the state variables are 1 and 0, and whether the virtual machine i and the server j are in the running state or not is respectively represented;
the integrity constraint is embodied by the following formula:
the integrity constraint formula adopts an accumulation calculation mode, the placement states of any container, virtual machine, all virtual machines and servers are accumulated and summed, and the summation result is limited to 1, so that abnormal conditions except 1 are avoided, and all containers and virtual machines are placed at reasonable positions in a cluster;
the resource utilization condition constraint conditions comprise virtual machine resource utilization condition constraint conditions and server resource utilization condition constraint conditions, and the virtual machine resource utilization condition constraint conditions ensure that the quantity of resources requested by a Docker each time does not exceed the total quantity of resources applied when the virtual machine is created by limiting the resource allocation condition of the virtual machine, and are specifically represented by the following formula:
wherein,respectively representing the amount of CPU and memory resources applied when the r-th Docker is created,the total amount of resources of a CPU and a memory which can be provided by the corresponding ith virtual machine; server resource utilizationThe constraint condition ensures that the quantity of resources requested by the virtual machine each time does not exceed the total quantity of resources provided by the server by limiting the resource allocation condition of the server, and is specifically represented by the following formula:
wherein,respectively representing the CPU and memory resource amount applied when the ith virtual machine is established, the total amount of resources of a CPU and a memory which can be provided by the corresponding jth server;
the resource usage constraint condition limits the calculation mode of the VM and PM resource usage in the DVP framework, and is specifically described as that the resource usage of the VM and the PM at any moment takes the total amount of all Docker container resources borne by the VM and the PM as a statistical basis; the specific resource usage constraint is expressed as the following equation:
wherein, yriAnd zrjState variables for the Docker to virtual machine position mapping matrix Y and the Docker to server position mapping matrix Z respectively,respectively representing the CPU and memory resource amount applied when the r-th Docker is created, taking the total resource amount of all Docker containers borne by the VM and the PM at any moment as a statistical basis,respectively representing the amount of CPU and memory resources used by the container in the jth server,respectively representing the amount of CPU and memory resources used by the container in the ith virtual machine.
3. The method for initial placement of Docker and virtual machine based on three-tier architecture as claimed in claim 2, wherein in step2, a DVMP initial placement model of three-tier architecture is constructed, the specific content is represented by a resource utilization optimization function introducing resource utilization of data center based on the DVMP model, and the optimization target of the resource utilization of data center is represented by the following formula:
wherein,indicating the number of servers in active state, all server state variables P passing through the data centerjThe sum is calculated from the following equation:
wherein,andrespectively representing the idle CPU and memory resource amount of the jth server; the optimization goal of improving the resource utilization rate of the data center consists of the product of minimizing the number of active state servers and maximizing the resource margin of all the servers in each dimension,andis obtained by the following formula:
wherein,representing the amount of resources server j is used by the container,the total amount of resources of the CPU and the memory which can be provided by the corresponding server j;
in the optimization objective, the process of the method,partly to prevent the appearance of container sets of the same size and configuration, the smaller the number of servers used means that containers are putThe more centralized the placement, the higher the resource utilization;the method is characterized in that the influence of resource allowance is partially represented, the product of CPU and memory resource allowance is used as effective evaluation of multidimensional resources in the placement process based on a multidimensional resource normalization processing mode provided by Wood T, when the number of servers used for placement is the same, the larger the sum of the product of the resource allowance of the data center server is, the larger the specification of certain server residual resources is, the more favorable the storage of new large-specification containers is, the larger the value is, the better the storage is, and α in the optimization target isE、βEIs two constants used for representing the weight of the first part and the second part in the resource utilization function, and the weight satisfies αE＞0＞βEAnd α, andEshould be much greater than | βEThe smaller the sum of the two parts of the resource utilization function, the more appropriate the placement scheme.
4. The method for initially placing a Docker and a virtual machine based on a three-tier architecture as claimed in claim 3, wherein in step3, initialization of a newly added scene and VM screening in a data center are specifically represented as the following steps:
step 1: initializing the configuration and the quantity of VMs and PMs in the data center by adopting uniformly distributed random functions, wherein the configuration and the quantity of the VMs and the PMs are preset as reference quantity of CPU resourcesReference quantity of memory resourcesCoacting according to the following formula:
wherein,andexpressing the resource demand of the CPU and the memory of the container i, wherein the numerical value is expressed by the percentage of the CPU and the memory resource demand to the total CPU and memory resource of the server;
step 2: and (3) screening virtual machine sequences which simultaneously meet all constraint conditions of the DVMP constraint model in the step (1) and the step (2) from all virtual machines in the data center according to the CPU and memory resource requests of the newly added container.
5. The method for initially placing a Docker and a virtual machine based on a three-tier architecture as claimed in claim 4, wherein in step 4, an optimal VM fitness function and an optimal VM selection are designed, specific contents are expressed by globally considering a resource relationship among the Docker, the VM and the PM, and the resource influence of VM placement on the PM is considered in the virtual machine fitness function, so as to prevent a relatively centralized placement scheme on a VM level from being actually deployed dispersedly on a PM physical level, which causes a large number of resource fragments and opens more servers to share a load, and the optimal VM fitness function is expressed by the following formula:
wherein,representing the CPU and memory resources of the newly added container,and the total amount of resources occupied by the container of the virtual machine i and the total amount of resources provided by the virtual machine i are respectively represented, α and β are two constants and are used for representing the weight of the first part and the second part in the optimal virtual machine fitness function, pbfit (j) represents the optimal server fitness function, and the optimal server fitness function is obtained by the following formula:
wherein,andrepresenting the total amount of resources server j actually occupied by the container,corresponding to the total amount of CPU and memory resources that server j is capable of providing,andis obtained by the following formula, represents the CPU, memory resource requirements, z, of container rrjState variables for the Docker to server location mapping matrix Z:
and respectively calculating the fitness of each virtual machine in the screened virtual machine sequence by adopting a virtual machine fitness function, and selecting the virtual machine with the maximum fitness as the optimal virtual machine of the container to be put in.
6. The method for initially placing a Docker and a virtual machine based on a three-tier architecture as claimed in claim 5, wherein in step 5, a minimum VM calculation rule and a PM screening are designed to avoid a situation that a randomly created virtual machine cannot be loaded by any active server and more servers are started, so that the minimum VM rule is determined as a critical point for whether to start a new server, which is specifically represented by the following steps:
step 1: inputting the CPU request quantity and the memory request quantity of the newly added container;
step 2: comparing the CPU and memory amount of the newly added container and each virtual machine template one by one, and selecting the minimum virtual machine capable of accommodating the new container;
step 3: determining whether the minimum virtual machine can be loaded in the current activation state of the data center, and determining whether the minimum virtual machine is a critical point for starting a new server;
and screening out the activated state server sequence meeting the constraint according to the minimum virtual machine rule, and carrying out the next step of operation.
7. The method for initially placing a Docker and a virtual machine based on a three-tier architecture as claimed in claim 6, wherein in step 6, an optimal PM fitness function and an optimal PM selection are designed, specifically, when all active state servers cannot accommodate loads, a new server is started and a virtual machine bearer container is created at random, and the virtual machine and the server are used as optimal VMs and PMs of the container; the optimal PM decision process of the minimum virtual machine is that the fitness calculation is carried out on the screened activated state servers in sequence according to the fitness function of the optimal server, and the virtual machine with the maximum fitness is selected as the optimal server of the container and the minimum virtual machine; the fitness calculation of each server adopts an optimal PM fitness function PBfit (j) to make a decision, which is specifically represented by the following formula:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811190209.3A CN109298914A (en) | 2018-10-12 | 2018-10-12 | A kind of Docker based on three-tier architecture and virtual machine initial placement method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811190209.3A CN109298914A (en) | 2018-10-12 | 2018-10-12 | A kind of Docker based on three-tier architecture and virtual machine initial placement method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109298914A true CN109298914A (en) | 2019-02-01 |
Family
ID=65162473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811190209.3A Pending CN109298914A (en) | 2018-10-12 | 2018-10-12 | A kind of Docker based on three-tier architecture and virtual machine initial placement method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109298914A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109960585A (en) * | 2019-02-02 | 2019-07-02 | 浙江工业大学 | A kind of resource regulating method based on kubernetes |
| CN110941495A (en) * | 2019-12-10 | 2020-03-31 | 广西大学 | Container collaborative arrangement method based on graph coloring |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8706782B2 (en) * | 2011-06-12 | 2014-04-22 | International Business Machines Corporation | Self-contained placement of data objects in a data storage system |
| CN103778020A (en) * | 2014-02-08 | 2014-05-07 | 中国联合网络通信集团有限公司 | Virtual machine placing method and device |
| CN104350460A (en) * | 2012-04-30 | 2015-02-11 | 惠普发展公司，有限责任合伙企业 | Determining virtual machine placement |
| CN104503826A (en) * | 2015-01-04 | 2015-04-08 | 中国联合网络通信集团有限公司 | Virtual machine mapping method and device for cloud computing data center |
| CN108073449A (en) * | 2017-11-21 | 2018-05-25 | 山东科技大学 | A kind of virtual machine dynamic laying method |
-
2018
- 2018-10-12 CN CN201811190209.3A patent/CN109298914A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8706782B2 (en) * | 2011-06-12 | 2014-04-22 | International Business Machines Corporation | Self-contained placement of data objects in a data storage system |
| CN104350460A (en) * | 2012-04-30 | 2015-02-11 | 惠普发展公司，有限责任合伙企业 | Determining virtual machine placement |
| CN103778020A (en) * | 2014-02-08 | 2014-05-07 | 中国联合网络通信集团有限公司 | Virtual machine placing method and device |
| CN104503826A (en) * | 2015-01-04 | 2015-04-08 | 中国联合网络通信集团有限公司 | Virtual machine mapping method and device for cloud computing data center |
| CN108073449A (en) * | 2017-11-21 | 2018-05-25 | 山东科技大学 | A kind of virtual machine dynamic laying method |
Non-Patent Citations (2)
| Title |
|---|
| QINGHUA ZHENG, RUI LI, XIUQI LI, JIE WU: "A Multi-Objective Biogeography-Based Optimization for Virtual Machine Placement", 《2015 15TH IEEE/ACM INTERNATIONAL SYMPOSIUM ON CLUSTER, CLOUD AND GRID COMPUTING》 * |
| RONG ZHANG,A-MIN ZHONG,BO DONG,FENG TIAN,RUI LI: "Container-VM-PM Architecture: A Novel Architecture for Docker Container Placement", 《CLOUD COMPUTING-CLOUD 2018》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109960585A (en) * | 2019-02-02 | 2019-07-02 | 浙江工业大学 | A kind of resource regulating method based on kubernetes |
| CN110941495A (en) * | 2019-12-10 | 2020-03-31 | 广西大学 | Container collaborative arrangement method based on graph coloring |
| CN110941495B (en) * | 2019-12-10 | 2022-04-05 | 广西大学 | Container collaborative arrangement method based on graph coloring |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102222014B (en) | Dynamic memory management method based on memory hot plug for virtual machine | |
| CN104679594B (en) | A kind of middleware distributed computing method | |
| CN106528266A (en) | Resource dynamic adjustment method and device in cloud computing system | |
| CN105847385B (en) | A kind of cloud computing platform dispatching method of virtual machine based on operation duration | |
| CN109298914A (en) | A kind of Docker based on three-tier architecture and virtual machine initial placement method | |
| CN108073457A (en) | A kind of hierarchical resource management method of super fusion architecture, apparatus and system | |
| Li et al. | A wholistic optimization of containerized workflow scheduling and deployment in the cloud–edge environment | |
| ITRM20110433A1 (en) | ENERGY SAVING SYSTEM IN THE COMPANY DATE CENTERS. | |
| CN105975341A (en) | Multicore processor-oriented internal memory management system | |
| Kiani et al. | A network-aware and power-efficient virtual machine placement scheme in cloud datacenters based on chemical reaction optimization | |
| Kim et al. | Partition placement and resource allocation for multiple DNN-based applications in heterogeneous IoT environments | |
| Batra et al. | Best Fit Sharing and Power Aware (BFSPA) Algorithm for VM placement in cloud environment | |
| Wu et al. | Heterogeneous virtual machine consolidation using an improved grouping genetic algorithm | |
| CN110308991B (en) | Data center energy-saving optimization method and system based on random tasks | |
| CN108073449B (en) | Dynamic virtual machine placement method | |
| CN114003238B (en) | Container deployment method, device, equipment and storage medium based on transcoding card | |
| CN106227600B (en) | A kind of multidimensional virtual resource allocation method based on Energy-aware | |
| CN107590000A (en) | Secondary random sources management method/system, computer-readable storage medium and equipment | |
| CN113094149A (en) | Data center virtual machine placement method, system, medium and equipment | |
| CN115269110A (en) | Resource allocation method and device for cloud computing platform, electronic equipment and storage medium | |
| Zhu et al. | Balanced Cloud Edge resource allocation based on conflict conditions | |
| CN112346861A (en) | Resource allocation method, device, equipment and super-fusion system | |
| CN109240805A (en) | Virtual machine distribution method and device | |
| Yan et al. | Storage-aware server consolidation for cloud services utilizing local storage | |
| Gao et al. | Deep reinforcement learning based compute-intensive workload allocation in data centers with high energy efficiency |
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 |
Application publication date: 20190201 |
|
| RJ01 | Rejection of invention patent application after publication |