A cloud data center high-efficiency container orchestration method and system

Abstract

The application discloses a kind of cloud data center high efficiency container arrangement method and system, it includes obtaining cloud data center overall information, including server quantity, each server resource upper limit, each server minimum power, each server maximum power, average migration network bandwidth budget provided by cloud data center;The information of all containers in a continuous time slot is obtained, including container quantity, each container resource request amount, each container image size, each container running time, each container migration attribute;According to the information of all containers in the continuous time slot and cloud data center overall information, the optimization problem with the minimum cloud data center average energy consumption as the target is constructed and solved, and the container arrangement decision of the continuous time slot is obtained.The application can obtain current time slot container arrangement decision under the condition of meeting various resource constraint conditions in the case where future container creation request quantity and container specification parameters cannot be accurately obtained, and is efficient, energy-saving and environment-friendly.

Description

Technical Field

[0001] This invention relates to the field of cloud computing, and more specifically to a container orchestration method and system for cloud data centers. Background Technology

[0002] Cloud computing is a novel business computing model. Through virtualization technology, it pools the resources of servers within a data center, providing computing resources on demand via the internet. Enterprises or individual users can deploy their businesses to the cloud according to their needs, eliminating the need to purchase, configure, or manage resources themselves; they only pay based on the actual amount and duration of resource usage. Cloud computing makes application deployment easier for businesses and individuals and reduces various maintenance costs. Containers are a type of virtualization technology. Compared to traditional virtual machines, containers offer advantages such as faster start / stop speeds, lightweight design, strong scalability, and better isolation. Therefore, containers have become a replacement technology for virtual machines and are widely used in cloud data centers. Containers create container instances from images, and each container instance has its own resource requests, runtime, and migration attributes.

[0003] As cloud computing demand continues to grow, the scale of cloud data centers is also expanding rapidly, exposing the problem of high energy consumption in cloud data centers. The energy consumption of cloud data centers consists of multiple components, including server energy consumption. The energy consumption of a single server is related to the server's computing resource utilization rate and its own energy consumption attributes; servers without load will enter a dormant state to reduce energy consumption. Within cloud data centers, due to the ever-increasing number of containers and the lack of suitable high-performance container orchestration strategies, containers are scattered across different servers, resulting in a large number of servers operating at low resource utilization, thus leading to increased data center energy consumption. Cloud data centers urgently need a container orchestration method to reduce energy consumption.

[0004] However, the number of container creation requests changes continuously over time, making it impossible to accurately predict the actual number of user requests in the future. Furthermore, the differences in runtime, resource request parameters, and migration attributes among containers make it difficult to determine the optimal high-performance container orchestration strategy over the long term. Existing cloud data center container orchestration strategies simply copy virtual machine orchestration strategies without considering container migration attributes, resulting in suboptimal energy-saving effects and insufficient reduction in cloud data center energy consumption. These issues urgently need to be addressed. Summary of the Invention

[0005] Purpose of the invention: In order to overcome the shortcomings of the prior art, the present invention provides a high-performance container orchestration method for cloud data centers, and also provides a high-performance container orchestration system for cloud data centers.

[0006] Technical Solution: To solve the above technical problems, this invention provides a high-performance container orchestration method for cloud data centers, which includes the following steps:

[0007] Obtain overall information about the cloud data center, including the number of servers, the resource limits of each server, the minimum power of each server, the maximum power of each server, and the average migration network bandwidth budget provided by the cloud data center;

[0008] Obtain information about all containers within a continuous time slot, including the number of containers, the resource request amount of each container, the image size of each container, the runtime of each container, and the migration attributes of each container; the continuous time slot is from time slot 1 to time slot T, where T>1;

[0009] Based on the information of all containers in the continuous time slot and the overall information of the cloud data center, an optimization problem with the goal of minimizing the average energy consumption of the cloud data center is constructed and solved. The solution is used as the container orchestration decision for the continuous time slot. The container orchestration decision is the orchestration and deployment position of each container in each time slot from the beginning time slot 1 to time slot T.

[0010] This invention also provides a high-performance container orchestration system for cloud data centers, comprising:

[0011] The system initial unit is configured to initialize the original values ​​of dynamic parameters;

[0012] The input acquisition unit is configured to acquire information as input for other modules;

[0013] The decision storage unit is configured to acquire and store container orchestration decisions for each time slot;

[0014] The orchestration decision feedback unit is configured to return the deployment effect corresponding to a certain container orchestration decision;

[0015] The orchestration decision generation unit is configured to obtain container orchestration decisions within the current time slot.

[0016] Preferably, the orchestration decision generation unit includes an optimization problem construction module and an optimization problem solving module:

[0017] The optimization problem construction module is configured to construct the container orchestration optimization problem for the current time slot based on the container requests arriving in the current time slot and the cloud data center container orchestration decision of the previous time slot, using any of the methods described above.

[0018] The optimization problem-solving module is configured to solve the container orchestration optimization problem of this time slot using any of the methods described above, and obtain the final container orchestration decision for this time slot.

[0019] This invention also provides a computer device, comprising:

[0020] One or more processors; and

[0021] One or more memories; and

[0022] One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the cloud data center high-performance container orchestration method as described above.

[0023] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the cloud data center high-performance container orchestration method as described above.

[0024] Beneficial effects: This invention provides a chained service entity placement method based on an edge computing environment with the goal of minimizing user response time. By constructing a network model, latency model, and cost model of the edge network environment, and combining the objective function and constraints of the chained service entity placement problem in edge computing, a heuristic algorithm based on K-Means clustering and greedy algorithms is used to obtain a chained service entity placement scheme, which can achieve near-optimal results with low time complexity.

[0025] Compared to existing technologies, the high-performance container orchestration method and system for cloud data centers provided by this invention have the following beneficial effects: This invention creates an optimization problem with the goal of minimizing the average energy consumption of the cloud data center, while satisfying various resource constraints and limitations in various scenarios. It then transforms and solves this optimization problem to obtain the container orchestration decision for each time slot. The high-performance container orchestration method and system for cloud data centers provided by this invention can, at the beginning of each time slot, obtain the container orchestration decision based on container creation requests and the container orchestration situation of previous time slots, even when the number of future container creation requests and container specifications cannot be accurately known. It then orchestrates newly arriving containers and already running containers, achieving the effect of minimizing average energy consumption in cloud data center scenarios while satisfying various resource constraints, effectively reducing the average energy consumption of cloud data centers, and achieving high efficiency, energy saving, and environmental protection. Attached Figure Description

[0026] Figure 1 A flowchart illustrating the high-performance container orchestration method for cloud data centers provided in this embodiment;

[0027] Figure 2 A schematic diagram of the structure of a high-performance container orchestration system for cloud data centers provided in this embodiment;

[0028] Figure 3 for Figure 2A schematic diagram of the decision feedback unit in the middle of the program;

[0029] Figure 4 for Figure 2 A schematic diagram of the middle-level decision generation unit;

[0030] Figure 5 The simulation comparison chart shows the average energy consumption of the high-performance container orchestration method for cloud data centers provided in the example and other orchestration algorithms. Detailed Implementation

[0031] The present invention will be further described in detail below with reference to the embodiments. These embodiments do not constitute a limitation on the present invention.

[0032] This embodiment provides a high-performance container orchestration method for cloud data centers, such as... Figure 1 As shown, it includes the following steps:

[0033] (S1) Obtain / collect overall information about the cloud data center, including the number of servers, the resource limits of each server (including the computing resource limits and memory resource limits of each server), the minimum power of each server, the maximum power of each server, and the average migration network bandwidth budget provided by the cloud data center.

[0034] (S2) Obtain information about all containers within a continuous time slot, including the number of containers, the resource request amount of each container, the image size of each container, the runtime of each container, and the migration attributes of each container; the continuous time slot is from time slot 1 to time slot T, where T>1;

[0035] (S3) Based on the information of all containers in the continuous time slot and the overall information of the cloud data center, construct an optimization problem (1) with the goal of minimizing the average energy consumption of the cloud data center and solve it. The solution result is used as the container orchestration decision for the continuous time slot. The container orchestration decision is the orchestration and deployment position of each container in each time slot from the start time slot 1 to time slot T.

[0036] In the method provided in this embodiment, after constructing the optimization problem (1) with the objective of minimizing the average energy consumption of the cloud data center in the above step (S3), it further includes:

[0037] (S31) The constructed optimization problem (1) aimed at minimizing the average energy consumption of the cloud data center is transformed and decomposed into optimization problems (2) for each time slot;

[0038] (S32) At the beginning of each time slot, the optimization problem (2) is actually solved, that is, the container orchestration optimization problem (2) in that time slot is solved to obtain the final container orchestration decision for that time slot.

[0039] Specifically, in step (S32) above, the container orchestration optimization problem (2) within this time slot is solved to obtain the final container orchestration decision for this time slot, which includes the following steps:

[0040] L1) Obtain an initial container orchestration decision S0 for this time slot: At the beginning of this time slot, obtain the container orchestration decision of the previous time slot (referring to the final container orchestration decision of the previous time slot), and generate an initial container orchestration decision S0 for this time slot based on this decision.

[0041] L2) Through iterative iteration, the container orchestration decision S0 is updated based on the initialization, to obtain the final container orchestration decision S0 for this time slot. f (That is, the initial container orchestration decision S0 is iterated f times, and the container orchestration decision is continuously updated until the final container orchestration decision S0 for this time slot is obtained.) f );

[0042] L3) Update dynamic parameters based on the final container orchestration decision for this time slot, that is: based on the obtained final container orchestration decision S for this time slot. f Update the virtual queue variable q t In this embodiment, it can also be described as, based on the final container orchestration decision S obtained for this time slot. f Based on the final container orchestration decision from the previous time slot, update the virtual queue variables.

[0043] Specifically, the goal of this method is to minimize the average server energy consumption of the cloud data center under the constraints of various resources in the cloud data center: The optimization problem (1) established with the goal of minimizing the average energy consumption of the cloud data center is as follows:

[0044] Optimization goal:

[0045] Constraints:

[0046] 1.1) For any container i, the orchestration decision constraints are:

[0047] 1.2) For any container i, the orchestration decision constraints during runtime are:

[0048] 1.3) For any container i: runtime out-of-time orchestration decision constraints:

[0049] 1.4) For any server j, computational resource constraints:

[0050] 1.5) For any server j, memory resource constraints:

[0051] 1.6) For any container i, the migration attribute constraint is:

[0052] 1.7) Long-term migration network bandwidth constraints:

[0053] In the formula, [x,y] represents the set from integer x to integer y; T represents the total number of time slots on the time axis; M is the total number of servers in the cloud data center; N is the total number of server instances running from time slot 1 to time slot T; E i,j,t The decision of whether to schedule device i in time slot t to server j; P t The sum of power consumption of all servers in the cloud data center within time slot t; t i,a For the first time slot of container i; t i,e For the last time slot in which container i runs; R i,c R represents the amount of computing resources requested by container i; i,m L represents the memory resource request amount for container i; j,c L represents the total computing resources of server j; j,m M represents the total memory resources of server j; i Let C be the migration attribute characterization variable for container i; t C represents the migration network bandwidth overhead caused by container orchestration within time slot t; b Average migration network bandwidth budget provided for cloud data centers;

[0054] The sum of power consumption P of all servers in the cloud data center within time slot t. t The calculation formula is:

[0055]

[0056]

[0057]

[0058] In the formula, μ t,j This indicates the computing resource usage of server j in time slot t; P t,j P represents the energy consumption of server j in time slot t; j,min P represents the minimum power consumption of server j; j,max This represents the maximum power consumption of server j;

[0059] The migration network bandwidth overhead C caused by container orchestration within time slot t. t The calculation formula is:

[0060]

[0061]

[0062] In the formula, H t,i Let I be the characterization variable for whether container i migrates in time slot t. i This indicates the image size of container i.

[0063] In reality, based on a current time slot, such as a time slot before time slot 1, it is difficult to obtain and predict the information of all containers within a future continuous time slot (from time slot 1 to time slot T). Even based on the current time slot t after time slot 1, only the information of all containers from time slot 1 to time slot t can be obtained, and the information of containers in future time slots (from time slot t+1 to time slot T) is still difficult to obtain and predict. Therefore, in order to obtain an effective solution based on reality, this method transforms the optimization problem of a continuous time slot into an optimization problem for each time slot. Further, the transformation of optimization problem (1) includes:

[0064] The optimization problem (1) is decomposed into each time slot. At the beginning of each time slot, the following optimization problem (2) is actually solved:

[0065] Optimization objective: min(q) t *C t +V*P t )

[0066] Constraints: Including constraints 1.1) to 1.6) above.

[0067] In the formula, q t The virtual queue variable is continuously updated, with an initial value of q1 = 0. It represents the amount of migration network bandwidth expenditure caused by the arrangement of the contents of each time slot from the beginning time slot 1 to the previous time slot t-1 of the current time slot t that exceeds the average migration network bandwidth budget (representing the degree of importance attached to the dynamic adjustment of migration network bandwidth expenditure); V is a preset weight adjustment parameter (representing the degree of importance attached to energy consumption).

[0068] Meanwhile, this embodiment provides a high-performance container orchestration method for cloud data centers, which also includes the following steps:

[0069] The steps for storing / retrieving the final container orchestration decision for a specific time slot; the container orchestration decision refers to the orchestration and deployment positions of all containers within that time slot.

[0070] The steps for calculating / obtaining the deployment effect of a certain container orchestration decision within a specific time slot; the deployment effect includes the sum P of the power consumption of all servers in the cloud data center within that time slot under the certain container orchestration decision. t The migration network bandwidth cost C caused by the orchestration decision for a particular type of container in the cloud data center during this time slot. t The deployment effect also includes the corresponding calculation result (also known as the score) U = q t *C t +V*P t .

[0071] In step L1) above, at the beginning of the time slot, the container orchestration decision of the previous time slot is obtained, and based on this, an initial container orchestration decision S0 for the current time slot is generated (that is, at the beginning of each time slot, a feasible initial strategy for optimization problem (2) is obtained), specifically including:

[0072] L11) Obtain the container orchestration decision from the previous time slot;

[0073] L12) Based on the container orchestration decision of the previous time slot, keep the orchestration position of the running containers unchanged, and for each new container arriving in this time slot, randomly and with equal probability, orchestrate it to a server whose available computing resources and available memory resources meet the resource requirements of the container, thereby obtaining an initial container orchestration decision S0 for this time slot.

[0074] Step L2) above is updated through iterative iteration based on the initial container orchestration decision S0 to obtain the final container orchestration decision S0 for this time slot. f This involves performing f iterations on the initial container orchestration decision S0 (where f is the number of iterations), continuously updating the container orchestration decision until the final container orchestration decision S is obtained. f The loop iteration specifically includes:

[0075] After the k-th iteration, the resulting container orchestration decision is S. k ;

[0076] Based on container orchestration decision S k In the newly arrived container or the running container with a migration attribute variable of 1, a container i′ is randomly selected with equal probability.

[0077] Among all servers with available computing and memory resources that meet the container requirements, one server j′ is randomly selected with equal probability.

[0078] Set the orchestration position of container i′ to server j′, thereby enabling S k Based on this, the container orchestration decision S′ is obtained. k ;

[0079] For container orchestration decisions S k and S′ k Substituting these values ​​into the calculation expressions related to the optimization objective of optimization problem (2), we obtain the results U and U′ respectively (which can also be described as the results of the container orchestration decision S respectively). k and S′ k Quantitative scoring is performed to obtain the corresponding calculation results, i.e., scores U and U′); this corresponds to the step in this method of calculating / obtaining the deployment effect of a certain container orchestration decision in a certain time slot (here, this time slot); the calculation expression related to the optimization objective of the above optimization problem (2) is also the calculation expression of the corresponding calculation result, i.e., the score, in this step of this method; where U = q t *C t +V*P t U′=q t *C′ t +V*P′ t Here: C t C′ t These represent the container orchestration decisions S made by the cloud data center within this time slot. k S′ k The resulting network bandwidth costs, P t 、P′ t These represent the container orchestration decisions S within the cloud data center at this time slot. k S′ k The sum of the power consumption of all servers under the specified conditions.

[0080] make Where τ is the policy transition smoothing parameter, and η represents the selection of S′. k As S k+1 The probability of S is such that with probability η, S k+1 S′ k With probability 1-η, S k+1 For S k ;

[0081] After each iteration, k is incremented by 1 to enter the next iteration, which means k = k + 1, where 1 ≤ k ≤ f; after completing f iterations, the final container orchestration decision S for this time slot is obtained. f .

[0082] Step L3) above is based on the final container orchestration decision S for this time slot. f Update the virtual queue variable q t (For, by obtaining the final container orchestration decision S for this time slot) f As feedback, the virtual queue variable q t The update includes:

[0083] Based on the final container orchestration decision S for this time slot f Based on the final container orchestration decision of the previous time slot, calculate the corresponding S in this time slot (based on the final container orchestration decision S in this time slot). f C) Migration network bandwidth overhead caused by container orchestration t Update the virtual queue variable q t+1 =[q t +C t -C b ] + ,in[*] + This refers to max{*,0} (meaning to compare with 0 and take the larger value).

[0084] Meanwhile, this embodiment provides a high-performance container orchestration system for cloud data centers, which, as follows: Figure 2 As shown, it includes:

[0085] The system initial unit is configured to initialize the original values ​​of dynamic parameters;

[0086] The input acquisition unit is configured to acquire information as input for other modules;

[0087] The decision storage unit is configured to acquire and store container orchestration decisions for each time slot;

[0088] The orchestration decision feedback unit is configured to return the deployment effect corresponding to a certain container orchestration decision; (or it can be said to be configured to quantify and score the container orchestration decision, which corresponds to the step in the above method of calculating / obtaining the deployment effect of a certain container orchestration decision in a certain time slot).

[0089] The orchestration decision generation unit is configured to obtain container orchestration decisions within the current time slot.

[0090] In this embodiment, the input acquisition unit includes a data center information acquisition module and a container information acquisition module;

[0091] The data center information acquisition module is configured to acquire overall information about the cloud data center, including the number of servers, the resource limit of each server, the minimum power of each server, the maximum power of each server, and the average migration network bandwidth budget provided by the cloud data center.

[0092] The container information acquisition module is configured to obtain information about all containers in a continuous time slot from the start time slot to the current time slot, including the number of containers, the resource request amount of each container, the image size of each container, the runtime of each container, and the migration attributes of each container.

[0093] In this embodiment, the operation flow of the above-mentioned orchestration decision feedback unit is as follows: Figure 3 As shown, it includes a orchestration decision energy consumption feedback module and an orchestration decision migration expense feedback module;

[0094] The orchestration decision energy consumption feedback module is configured to, for container orchestration decisions within a time slot, return the sum of power consumption P of all servers in the cloud data center under that container orchestration decision within that time slot, based on the overall information of the cloud data center. t (This can also be referred to as the total server energy consumption within the cloud data center under this container orchestration decision during this time slot), the formula is:

[0095]

[0096]

[0097]

[0098] In the formula, μ t,j P represents the computing resource utilization of server j in time slot t; t,j P represents the energy consumption of server j in time slot t; j,min P represents the minimum power consumption of server j; j,max Indicates the maximum power consumption of server j; M is the total number of servers in the cloud data center; N is the total number of instances running in the server from time slot 1 to time slot T; E i,j,t The decision of whether to schedule device i in time slot t to server j; (P t R is the sum of the power consumption of all servers in the cloud data center within time slot t; i,c L represents the amount of computing resources requested by container i; j,c Let J be the total computing resources of server j;

[0099] The orchestration decision migration cost feedback module is configured to, for a specific container orchestration decision within a time slot, return the total migration network bandwidth cost C caused by that container orchestration decision within this time slot, based on the final container orchestration decision information from the previous time slot. t The formula is:

[0100]

[0101]

[0102] In the formula, [x,y] represents the set from integer x to integer y; i Indicates the image size of container i; t i,a For the first time slot of container i; t i,e For the last time slot in which container i runs; R i,m H represents the amount of memory resources requested by container i.t,i C is a characterization variable for whether container i migrates in time slot t. t This represents the migration network bandwidth overhead caused by container orchestration within time slot t.

[0103] The aforementioned orchestration decision generation unit includes an optimization problem construction module and an optimization problem solving module, and its operation flow is as follows: Figure 4 As shown:

[0104] The optimization problem construction module is configured to construct a model for the container orchestration optimization problem in the current time slot based on the container requests arriving in the current time slot and the cloud data center container orchestration decision in the previous time slot, using any of the cloud data center high-performance container orchestration methods described above. That is, based on the container requests arriving in the current time slot and the current cloud data center container orchestration status, the optimization problem (2) is constructed, as shown below:

[0105] Optimization objective: min(q) t *C t +V*P t )

[0106] Constraints:

[0107] 1.1) For any container i, the orchestration decision constraints are:

[0108] 1.2) For any container i, the orchestration decision constraints during runtime are:

[0109] 1.3) For any container i: runtime out-of-time orchestration decision constraints:

[0110] 1.4) For any server j, computational resource constraints:

[0111] 1.5) For any server j, memory resource constraints:

[0112] 1.6) For any container i, the migration attribute constraint is:

[0113] In the formula, M i Let q be the migration attribute characterization variable for container i; tThe virtual queue variable is continuously updated, with an initial value of q1 = 0. It represents the amount of migration network bandwidth expenditure caused by the arrangement of the contents of each time slot from the beginning time slot 1 to the previous time slot t-1 of the current time slot t that exceeds the average migration network bandwidth budget (representing the degree of importance attached to the dynamic adjustment of migration network bandwidth expenditure); V is a preset weight adjustment parameter (representing the degree of importance attached to energy consumption).

[0114] The optimization problem-solving module is configured to use any of the cloud data center high-performance container orchestration methods described above to solve the container orchestration optimization problem within this time slot, thereby obtaining the final container orchestration decision for this time slot. In this embodiment, the specific steps for solving the container orchestration optimization problem within this time slot to obtain the final container orchestration decision for this time slot are as follows:

[0115] L1) At the beginning of this time slot, obtain the container orchestration decision of the previous time slot, and generate an initial container orchestration decision S0 for this time slot based on this decision, including:

[0116] L11) Obtain the container orchestration decision from the previous time slot;

[0117] L12) Based on the container orchestration decision of the previous time slot, keep the orchestration position of the running containers unchanged, and for each new container arriving in this time slot, randomly orchestrate it with equal probability to a server whose available computing resources and available memory resources meet the resource requirements of the container, thereby obtaining an initial container orchestration decision S0 for this time slot.

[0118] L2) Through iterative iteration, the container orchestration decision S0 is updated based on the initialization, to obtain the final container orchestration decision S0 for this time slot. f That is, the initial container orchestration decision S0 is iterated f times, and the container orchestration decision is continuously updated until the final container orchestration decision S0 for this time slot is obtained. f The loop iteration specifically includes:

[0119] L21) After the k-th iteration, the resulting container orchestration decision is S. k ;

[0120] L22) Based on container orchestration decisions S k In the newly arrived container or the running container with a migration attribute variable of 1, a container i′ is randomly selected with equal probability.

[0121] L23) Among all servers with available computing and memory resources that meet the container requirements, one server j′ is randomly selected with equal probability;

[0122] L24) sets the orchestration position of container i′ to server j′, thereby enabling S k Based on this, the container orchestration decision S′ is obtained. k ;

[0123] L25) For container orchestration decisions S k and S′ k Substitute them into the calculation expressions related to the optimization objective of optimization problem (2) respectively to obtain the results U and U′ respectively; this also corresponds to the step of calculating / obtaining the deployment effect of a certain container orchestration decision in a certain time slot (here, this time slot) in the above method, and corresponds to the deployment effect of a certain container orchestration decision calculated / returned by the orchestration decision feedback unit in the above system (which can also be said to be a quantitative score of a certain container orchestration decision);

[0124] L26) Order Where τ is the policy transition smoothing parameter, and η represents the selection of S′. k As S k+1 The probability of S is such that with probability η, S k+1 S′ k With probability 1-η, S k+1 For S k ;

[0125] In iterative steps L21) to L26), after each iteration, k is incremented by 1 to enter the next iteration, which means: k = k + 1, where 1 ≤ k ≤ f; after completing f iterations, the final container orchestration decision S for this time slot is obtained. f .

[0126] L3) Update dynamic parameters based on the final container orchestration decision for this time slot, that is: based on the obtained final container orchestration decision S for this time slot. f Update the virtual queue variable q t In this embodiment, it can also be described as, based on the final container orchestration decision S obtained for this time slot. f Based on the final container orchestration decision from the previous time slot, update the virtual queue variables.

[0127] Figure 5The simulation data provided in this embodiment, comparing the average energy consumption of the high-performance container orchestration method for cloud data centers with other orchestration algorithms (the vertical axis represents average energy consumption), demonstrates the average energy consumption of cloud data centers caused by different container orchestration methods. The high-performance container orchestration method for cloud data centers provided in this embodiment is abbreviated as "LM". The container orchestration methods compared include: Greedy & Migration (Greedy & Migration); NextFit (NF & Migration) (NextFit & Migration); FirstFitDecreasing (FFD & Migration) (FirstFitDecreasing & Migration); and Segmentation (FirstFitDecreasing & Segmentation). Figure 5 As can be seen, the high-efficiency container orchestration method for cloud data centers provided in this embodiment reduces the average energy consumption of servers by more than 16.29% or even up to 31.86% compared with other existing container orchestration methods. It can effectively reduce the average energy consumption of cloud data centers, and is highly efficient, energy-saving and environmentally friendly.

[0128] The present invention also provides a computer device comprising: one or more processors; one or more memories; and one or more programs, wherein the one or more programs are stored in the memories and configured to be executed by the one or more processors, wherein when the programs are executed by the processors, they implement the steps of the cloud data center high-performance container orchestration method as described above.

[0129] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the cloud data center high-performance container orchestration method as described above.

[0130] The container orchestration decisions described in this paper can also be referred to as container orchestration strategies. A particular container orchestration decision described in this paper can also be referred to as a specific container orchestration decision. The high-performance container orchestration method for cloud data centers described in this paper can also be referred to as a high-performance container orchestration method for cloud computing centers. The high-performance container orchestration system for cloud data centers described in this paper can also be referred to as a high-performance container orchestration system for cloud computing centers.

[0131] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware, and the program can be stored in a computer-readable storage medium. In the context of this invention, the computer-readable storage medium can be considered tangible and non-transitory. Non-limiting examples of non-transitory tangible computer-readable storage media include non-volatile memory circuits (e.g., flash memory circuits, erasable programmable read-only memory circuits, or mask read-only memory circuits), volatile memory circuits (e.g., static random access memory circuits or dynamic random access memory circuits), magnetic storage media (e.g., analog or digital magnetic tape or hard disk drives), and optical storage media (e.g., CDs, DVDs, or Blu-ray discs), etc.

[0132] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0133] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the invention. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0134] The above are merely preferred embodiments of the present invention. It should be noted that the above embodiments do not constitute a limitation on the present invention. Equivalent transformations, various changes and modifications made by those skilled in the art without departing from the technical concept of the present invention all fall within the protection scope of the present invention.

Claims

1. A high-performance container orchestration method for cloud data centers, characterized in that... Includes the following steps: Obtain overall information about the cloud data center, including the number of servers, the resource limits of each server, the minimum power of each server, the maximum power of each server, and the average migration network bandwidth budget provided by the cloud data center; Obtain information about all containers within a continuous time slot, including the number of containers, the resource request amount of each container, the image size of each container, the runtime of each container, and the migration attributes of each container; the continuous time slot is from time slot 1 to time slot T, where T>1; Based on the information of all containers in the continuous time slot and the overall information of the cloud data center, an optimization problem with the goal of minimizing the average energy consumption of the cloud data center is constructed and solved. The solution is used as the container orchestration decision for the continuous time slot. The container orchestration decision is the orchestration and deployment position of each container in each time slot from the start time slot 1 to time slot T. The optimization problem (1) aimed at minimizing the average energy consumption of cloud data centers is as follows: Optimization goal: Constraints: (1.1) For any container Decision-making constraints: (1.2) For any container During runtime, orchestration decision constraints are: (1.3) For any container Out-of-runtime orchestration decision constraints: (1.4) For any server Computational resource constraints: (1.5) For any server Memory resource constraints: (1.6) For any container Migration attribute constraints: (1.7) Long-term migration network bandwidth constraints: In the formula, Indicates from integers to integers A set; This indicates the total number of time slots on the time axis; This represents the total number of servers within the cloud data center. From time slot 1 to time slot Total number of running instances of the container; For the contents of time slot t Orchestrate to server Decision-making; This is the sum of the power consumption of all servers in the cloud data center within time slot t; For containers The first time slot to run; For containers The last time slot in operation; For containers The amount of computing resources requested; For containers The amount of memory resources requested; For server Total computing resources; For server Total memory resources; For containers The migration attribute characterizes the variable; This refers to the migration network bandwidth overhead caused by container orchestration within time slot t. Average migration network bandwidth budget provided for cloud data centers; The sum of the power consumption of all servers in the cloud data center within time slot t. for: in: in: In the formula, Indicates server The utilization rate of computing resources in time slot t; Indicates server Energy consumption in time slot t; Indicates server The lowest power consumption; Indicates server Maximum power consumption; The migration network bandwidth overhead caused by container orchestration within time slot t. for: in For containers Characteristic variables for whether migration occurs in time slot t: In the formula Represents a container The size of the mirror image.

2. The high-performance container orchestration method for cloud data centers according to claim 1, characterized in that, The optimization problem (1) aimed at minimizing the average energy consumption of a cloud data center is transformed, including: The optimization problem (1) is decomposed into each time slot. At the beginning of each time slot, the actual optimization problem (2) is solved as follows: Optimization goal: ) Constraints: including constraints (1.1) to constraints (1.6) In the formula, This is a virtual queue variable, with its initial value... ; These are preset weight adjustment parameters.

3. The high-performance container orchestration method for cloud data centers according to claim 2, characterized in that, The container orchestration optimization problem (2) within this time slot is solved to obtain the final container orchestration decision for this time slot, which specifically includes: L1) obtains an initial container orchestration decision for this time slot. ; L2) Through iterative iteration, the container orchestration decision is initialized. Based on this, the final container orchestration decision for this time slot is obtained by updating the existing decision. ; L3) Based on the final container orchestration decision obtained for this time slot Update virtual queue variable .

4. The high-performance container orchestration method for cloud data centers according to claim 3, characterized in that, In step L1), an initial container orchestration decision for this time slot is obtained. ,include: L11) Obtain the container orchestration decision from the previous time slot; L12) Based on the container orchestration decision of the previous time slot, keeping the orchestration positions of already running containers unchanged, for each newly arriving container in this time slot, it is randomly orchestrated with equal probability to a server whose available computing resources and available memory resources meet the container's resource requirements, thus obtaining an initial container orchestration decision for this time slot. .

5. The high-performance container orchestration method for cloud data centers according to claim 3, characterized in that, Step L2) involves iterative iteration to initialize container orchestration decisions. Based on this, the final container orchestration decision for this time slot is obtained by updating the existing decision. This includes decisions regarding the initial container orchestration. conduct The process iterates through the loop, continuously updating the container orchestration decision until the final container orchestration decision is obtained. The loop iteration specifically includes: In the After the round of iterations, the resulting container orchestration decision is: ; Based on container orchestration decisions From the newly arrived container or the running container whose migration attribute characterization variable is 1, randomly select one container with equal probability. ; One server is randomly selected with equal probability from all servers whose available computing and memory resources meet the container's requirements. ; container The orchestration location is set to the server. Thus in Based on this, container orchestration decisions are obtained. ; For container orchestration decisions and Substituting these values ​​into the calculation expressions related to the optimization objective of optimization problem (2), we obtain the results respectively. and ; make ,in For policy transition smoothing parameters, Indicates selection As The probability of is such that is The probability, for ,by The probability, for ; Finish After several iterations, the final container orchestration decision for this time slot is obtained. .

6. The high-performance container orchestration method for cloud data centers according to claim 3, characterized in that, Step L3) is based on the final container orchestration decision obtained for this time slot. Update virtual queue variable Specifically, it includes: Based on the final container orchestration decision for this time slot Calculate the migration network bandwidth overhead caused by container orchestration within this time slot. Update the virtual queue variable ,in[*] + It refers to max{*,0}.

7. A high-performance container orchestration system for cloud data centers, characterized in that, include: The system initial unit is configured to initialize the original values ​​of dynamic parameters; The input acquisition unit is configured to acquire information as input for other modules; The decision storage unit is configured to acquire and store container orchestration decisions for each time slot; The orchestration decision feedback unit is configured to return the deployment effect corresponding to a certain container orchestration decision; The orchestration decision generation unit is configured to obtain container orchestration decisions within the current time slot; Optimization goal: Constraints: (1.1) For any container Decision-making constraints: (1.2) For any container During runtime, orchestration decision constraints are: (1.3) For any container Out-of-runtime orchestration decision constraints: (1.4) For any server Computational resource constraints: (1.5) For any server Memory resource constraints: (1.6) For any container Migration attribute constraints: (1.7) Long-term migration network bandwidth constraints: In the formula, Indicates from integers to integers A set; This indicates the total number of time slots on the time axis; This represents the total number of servers within the cloud data center. From time slot 1 to time slot Total number of running instances of the container; For the contents of time slot t Orchestrate to server Decision-making; This is the sum of the power consumption of all servers in the cloud data center within time slot t; For containers The first time slot to run; For containers The last time slot in operation; For containers The amount of computing resources requested; For containers The amount of memory resources requested; For server Total computing resources; For server Total memory resources; For containers The migration attribute characterizes the variable; This refers to the migration network bandwidth overhead caused by container orchestration within time slot t. Average migration network bandwidth budget provided for cloud data centers; The input acquisition unit includes: a data center information acquisition module and a container information acquisition module; The data center information acquisition module is configured to acquire overall information about the cloud data center, including the number of servers, the resource limit of each server, the minimum power of each server, the maximum power of each server, and the average migration network bandwidth budget provided by the cloud data center. The container information acquisition module is configured to obtain information about all containers from the start time slot to the current time slot, including the number of containers, the resource request amount of each container, the image size of each container, the runtime of each container, and the migration attributes of each container. The orchestration decision feedback unit includes an orchestration decision energy consumption feedback module and an orchestration decision migration expense feedback module; The orchestration decision energy consumption feedback module is configured to, for a container orchestration decision within a time slot, return the sum of the power consumption of all servers in the cloud data center under that container orchestration decision within that time slot, based on the overall information of the cloud data center. for: in: in: In the formula, Indicates server The utilization rate of computing resources in time slot t; Indicates server Energy consumption in time slot t; Indicates server The lowest power consumption; Indicates server Maximum power consumption; This represents the total number of servers within the cloud data center. From time slot 1 to time slot Total number of running instances of the container; For the contents of time slot t Orchestrate to server Decision-making; For containers The amount of computing resources requested; The orchestration decision migration cost feedback module is configured to, for a specific container orchestration decision within a time slot, return the total migration network bandwidth cost caused by that container orchestration decision within this time slot, based on the final container orchestration decision information from the previous time slot. for: in For containers Characteristic variables for whether migration occurs in time slot t: In the formula, Indicates from integers to integers A set; Represents a container The size of the mirror image; For containers The first time slot to run; For containers The last time slot in operation; For containers The amount of memory resources requested.

8. The high-performance container orchestration system for cloud data centers according to claim 7, characterized in that, The orchestration decision generation unit includes an optimization problem construction module and an optimization problem solving module: The optimization problem construction module is configured to construct the container orchestration optimization problem for the current time slot based on the container requests arriving in the current time slot and the cloud data center container orchestration decision of the previous time slot, using the method described in claim 1 or 2. The optimization problem-solving module is configured to solve the container orchestration optimization problem of this time slot using the method described in any one of claims 3 to 6, and obtain the final container orchestration decision for this time slot.

Images