Cloud computing virtual machine scheduling method based on dynamic resource demand and gamma robust optimization

Abstract

The invention discloses a cloud computing virtual machine scheduling method based on dynamic resource requirements and Gamma robust optimization, which comprises a virtual machine scheduling optimization model and a heuristic algorithm, and is characterized in that the Gamma robust optimization model is constructed by introducing Gamma parameters, redistribution of virtual machines is performed through a Gamma robust Firstfit algorithm, and the Gamma robust Firstfit algorithm is used for scheduling the virtual machines. According to the cloud computing virtual machine scheduling method and system, cloud computing virtual machine scheduling based on dynamic resource requirements and gamma robust optimization is achieved, the problems that the cloud computing resource utilization rate is low, the migration cost is high and uncertainty is difficult to deal with are solved, efficient virtual machine integration in a large-scale cloud environment is achieved, the resource utilization rate is effectively increased, host use is reduced, and cost is reduced.

Description

Technical Field

[0001] This invention belongs to the field of cloud computing technology and provides a method for integrating virtual machines in a cloud computing system, specifically involving a cloud computing virtual machine scheduling method based on dynamic resource requirements and robust optimization. Background Technology

[0002] Currently, the utilization rate of cloud computing resource systems is generally low. With the continuous expansion of data center scale, how to efficiently utilize the physical host resources of cloud resource systems to reduce system energy consumption and operation and maintenance costs has become a core technology. Therefore, virtual machine consolidation and scheduling is a key issue in cloud computing resource management. However, accurately predicting the resource requirements of virtual machines in cloud computing resource management is an extremely challenging task. Existing cloud computing system virtual machine prediction methods typically rely on historical data and relatively stable time-series information, but in practical applications, the resource requirements of virtual machines often exhibit significant volatility and uncertainty. This volatility makes traditional prediction methods difficult to handle, easily leading to resource waste or overload problems, thereby reducing the overall performance of the data center.

[0003] Existing technologies employ uncertain optimization methods for virtual machine consolidation in cloud computing systems. While traditional uncertain optimization methods can handle a certain range of uncertainty, they often have limitations regarding data distribution and lack robustness, making it difficult to balance resource utilization and cloud system stability. In the face of dynamic and complex cloud computing environments, existing traditional methods cannot effectively cope with frequent changes in cloud system resource demands, easily exhibiting overly conservative or insufficient optimization practices. This results in low virtual machine resource utilization and difficulty in reducing the host usage costs of cloud computing systems. Summary of the Invention

[0004] To overcome the shortcomings of the existing technologies, this invention proposes a virtual machine scheduling method for cloud computing systems based on dynamic resource requirements and Gamma robust optimization. It proposes a virtual machine scheduling optimization model and a heuristic algorithm to solve the problems of low resource utilization, high migration costs, and difficulty in dealing with uncertainties in the existing technologies. This enables efficient virtual machine integration in large-scale cloud environments, effectively improves resource utilization, reduces host usage, and lowers costs.

[0005] This invention utilizes Γ robustness theory to handle and predict resource demands on virtual machines in cloud computing systems under uncertain environments, and combines this with a dynamic resource scheduling strategy for virtual machine consolidation. According to Γ robustness theory, a risk pool effect exists at a certain scale, meaning that it is not necessary to consider all maximum values, but rather to select different maximum values ​​(Γ values) under different probability conditions. Therefore, this invention establishes a cloud computing system virtual machine scheduling model based on dynamic resource demands and Γ robustness optimization, using Γ robustness theory and the characteristics of virtual machines, to achieve efficient virtual machine consolidation in large-scale cloud environments.

[0006] For simplicity, the present invention defines the following terms and parameter symbols:

[0007] N: The number of virtual machines, with each virtual machine indexed by I.

[0008] M: The number of hosts, with each host indexed by J.

[0009] CPU capacity limit of host j

[0010] Memory capacity limit of host j

[0011] CPU usage of virtual machine i This refers to the average value used by virtual machine i; It is the fluctuation radius of virtual machine i;

[0012] Memory usage of virtual machine i

[0013] Initial placement of virtual machine i This indicates that virtual machine i was initially placed on host j.

[0014] Γ k For different numbers of virtual machines k, the corresponding Γ is calculated in advance; Γ is an uncertain parameter representing the number of virtual machines.

[0015] x ij : Binary decision variable, x ij =1 indicates that virtual machine i is placed on host j.

[0016] Binary decision variables, This indicates that virtual machine i is placed on host j and belongs to the maxset.

[0017] Binary decision variables, This indicates that virtual machine i is placed on host j and belongs to the minset.

[0018] (Here's an explanation: the large set and the small set are also defined based on the Γ robustness; the large set considers Γ maximum values, while the small set only considers the mean.)

[0019] y j : A binary decision variable, representing whether host j is occupied, taking a value of 0 or 1; y j =1 indicates that host j is being used. These decision variables are learned by the model, which determines whether this value should be 0 or 1 when minimizing the objective function; the model makes the decision. y j =1 indicates that host j is in use.

[0020] H jk : Binary decision variables, H jk =1 indicates that the number of virtual machines placed on host j is k (the model determines that the number of virtual machines that should be placed on host j is k).

[0021] W i : Binary decision variable, W i =1 indicates that virtual machine i has migrated, W i =0 indicates that virtual machine i did not migrate.

[0022] S j The threshold used to distinguish between large and small sets is a separating variable.

[0023] M j The maximum radius of the virtual machine allocated on host j.

[0024] The technical solution of the present invention is as follows:

[0025] A cloud computing virtual machine scheduling method based on dynamic resource requirements and robust optimization includes the following steps:

[0026] 1. Construct a robust optimization model to minimize the number of active hosts in the cloud computing system;

[0027] By introducing the Γ parameter to construct a Γ robust optimization model, the robustness of the model to uncertainty can be flexibly adjusted, thereby effectively dealing with the fluctuation of virtual machine resource demand while ensuring resource utilization. Γ is an uncertain parameter, meaning that it is not necessary to consider the maximum utilization of all virtual machines, but rather to select different numbers of Γ maximum virtual machine utilization scenarios under different probability constraints.

[0028] The objective of the constructed Γ robust optimization model is to minimize the number of active hosts;

[0029] The constraints of the optimization model include:

[0030] First constraint: All virtual machines must be deployed on a single host;

[0031] Second constraint: If virtual machine i is deployed on host j, then i is in the MaxSet or MinSet of host j.

[0032] The third constraint: the total number of virtual machines deployed on host j;

[0033] Fourth constraint: The total number of virtual machines in the MaxSet of host j;

[0034] Fifth constraint: Define the maximum radius of the virtual machine on host j;

[0035] Sixth and seventh constraints: Ensure that the delimiter variable S on host j is... j The largest value in the smallest set and the smallest value in the largest set are represented as: as well as

[0036] Eighth constraint: Ensure that the CPU hotspot threshold constraint is not violated;

[0037] Ninth constraint: Ensure that memory capacity constraints are not violated;

[0038] Tenth constraint: Determine whether virtual machine i is migrated;

[0039] All other constraints are constraints on the range of values ​​that the variables can take.

[0040] 2. Design the heuristic algorithm GammaFF to solve the above Γ robust optimization model, and realize efficient integration of virtual machines in a large-scale cloud environment. This includes the following steps:

[0041] A. Select the initial physical host and create a virtual machine queue;

[0042] B. Expanding the virtual machine queue:

[0043] In addition to clearing all virtual machines on the host, a certain number of virtual machines are randomly selected from other hosts and added to this virtual machine queue to form an expanded virtual machine queue. The virtual machines in the queue are sorted according to their usage requirements, and virtual machines with greater resource requirements are processed first to reduce the number of migrations and the impact during virtual machine reallocation.

[0044] C. Reallocate virtual machines using the Γ robust Firstfit algorithm, the steps of which include:

[0045] C1. Set the probability value; calculate the value of Γ.

[0046] C2. Determine if the physical host can be used to host virtual machines;

[0047] Determine whether the host can be placed by using CPU hotspot threshold constraints (the value of the first Γ).

[0048] C3. Place the virtual machine on the selected physical host and update the host's available resources in real time during the placement process.

[0049] Through the above steps, cloud computing virtual machine scheduling based on dynamic resource requirements and robust optimization is achieved.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0051] This invention provides a cloud computing virtual machine scheduling method based on dynamic resource requirements and Gamma-FF robust optimization. By combining the Gamma-FF robust optimization model with the GammaFF heuristic algorithm, this invention can significantly improve resource utilization, reduce virtual machine migration costs, and ensure system stability and efficiency under uncertain conditions. This method is particularly suitable for large-scale cloud computing environments, providing data centers with a more intelligent and adaptive resource management solution. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the execution steps of the GammaFF robust priority adaptation algorithm proposed in this invention;

[0053] In this process, grayed-out hosts represent those selected for clearing, where all currently placed virtual machines (VMs) will be cleared and added to a VM queue. A certain proportion of VMs from other hosts will also be randomly selected and added to the VM queue, thus constructing a complete VM sequence. Then, the Firstfit method, incorporating Gamma-Ray robustness theory, is used to reallocate the VMs currently in the VM sequence. If they can fit, meaning hosts 1 and 2 can accommodate all VMs, then host 3 can be cleared. In this case, host 1 or host 2 is selected again to clear and construct a VM sequence to see if further optimization is possible. If they cannot fit, meaning host 3 cannot be cleared due to insufficient space, the process returns to step one, and host 1 or host 2 is selected again to clear and construct a VM sequence once more.

[0054] Figure 2 This is a schematic diagram of the selection of Γ and the Minset and Maxset in the Γ robustness theory. Detailed Implementation

[0055] The present invention will be further described below with reference to the accompanying drawings and embodiments, but the scope of the invention is not limited in any way.

[0056] This invention provides a cloud computing virtual machine scheduling method based on dynamic resource requirements and robust optimization. Figure 1This is a schematic diagram of the execution steps of the GammaFF robust priority adaptation algorithm proposed in this invention.

[0057] The Γ robustness theory proposed in the literature (Bertsimas D, Sim M. The price of robustness[J]. Operationsresearch,2004,52(1):35-53) states that in practice, virtual machine consolidation means concentrating multiple virtual machines onto a few hosts. Since the load uncertainty of each virtual machine differs, when the load of some virtual machines increases, the load of other applications may be at a low point. Γ robustness optimization can consolidate these load fluctuations through the "risk pool effect," thereby offsetting the impact of local peak loads. For example, when we appropriately select the Γ value, even if some applications experience load peaks, the model reduces the risk of a single virtual machine crashing due to excessive load by considering overall uncertainty and scenario allocation. That is, it is not necessary to consider all maximum values, but rather to select different numbers of maximum values ​​under different probability conditions. Figure 2 As shown, regarding the selection of Γ and the significance of Minset and Maxset, for example, if there are 50 hosts, the probability that 25 virtual machines will simultaneously reach their maximum value is 0.02%. Therefore, if we set the violation probability to be less than 0.02, we only need to consider the maximum value of the usage of the 25 largest virtual machines. So these 25 virtual machines are divided into the large set Maxset, and the other 25 virtual machines only need to have their mean calculated, so they are divided into the small set minset.

[0058]

[0059] where

[0060]

[0061] and

[0062]

[0063] The Γ robustness theory indicates that a risk pooling effect exists at a certain scale, meaning that it is not necessary to consider all maximum values, but rather to select different maximum values ​​under different probability conditions. Therefore, this invention establishes a cloud computing system virtual machine scheduling model based on dynamic resource requirements and Γ robustness optimization, using Γ robustness theory and combining it with the characteristics of virtual machines, to achieve efficient virtual machine consolidation in large-scale cloud environments.

[0064] This invention provides a cloud computing virtual machine scheduling method based on dynamic resource requirements and robust optimization, comprising the following steps:

[0065] 1. Construct a robust optimization model to minimize the number of active hosts in the cloud computing system;

[0066] This invention first constructs a Γ-based robust optimization model. This model flexibly adjusts its robustness to uncertainty by introducing a Γ parameter (Γ is an uncertain parameter, meaning it doesn't need to consider the maximum utilization of all virtual machines, but rather selects the utilization of different numbers (Γ) of the largest virtual machines under different probability constraints). This allows it to effectively address the fluctuations in virtual machine resource demand while ensuring resource utilization. The objective of the optimization model is to minimize the number of active hosts, and the specific form of the objective function is as follows:

[0067]

[0068] Here, parameters α and β are priority constraints for the two parts of the objective function (the first term is the host cost used, and the second term is the migration cost); J is the total number of hosts; I is the total number of virtual machines i; since minimizing the number of active hosts is our primary objective, we can set α = 10. 3 And β = 1.

[0069] The objective of the optimization model has the following constraints:

[0070] Constraint 1:

[0071]

[0072] Each virtual machine is deployed on a unique host.

[0073] Constraint 2:

[0074]

[0075] If virtual machine i is deployed on host j, then i is in host j's MaxSet or MinSet.

[0076] Constraint 3:

[0077]

[0078] The total number of virtual machines deployed on host j is equal to k.

[0079] Another constraint should be added:

[0080]

[0081] Constraint 4:

[0082]

[0083] The constraint on the total number of virtual machines in the MaxSet of host j.

[0084] Constraint 5:

[0085]

[0086] Define the maximum radius of the virtual machine on host j.

[0087] Constraints 6-7:

[0088]

[0089] Ensure on host j as well as Constraint 8:

[0090]

[0091] Ensure that the CPU hotspot threshold is not violated.

[0092] Constraint 9:

[0093]

[0094] Ensure that the memory capacity is not violated.

[0095] Constraint 10:

[0096]

[0097] Determine whether virtual machine i is migrated. w is the constraint used to determine whether migration is necessary. Constraint 11-18:

[0098]

[0099]

[0100] Define the range of values ​​for the decision variables.

[0101] 2. Design a heuristic algorithm to solve the above Γ robust optimization model;

[0102] For large-scale cloud computing environments, traditional models take a very long time to solve. This invention also proposes a Gamma-Frequency Robust Adaptation (GammaFF) algorithm to quickly and efficiently solve the virtual machine consolidation problem. The algorithm mainly includes the following steps:

[0103] A. Select the initial physical host and create a virtual machine queue:

[0104] Select a target host from physical hosts with resource utilization below 50%, clear all virtual machines on that host, and add these virtual machines to the virtual machine queue. At this point, the virtual machine queue only contains the virtual machines from the host that was cleared.

[0105] B. Expanding the virtual machine queue:

[0106] In addition to clearing all virtual machines from the host machine, a certain number of virtual machines are randomly selected from other hosts and added to this virtual machine queue. This is done to enhance the flexibility of the algorithm. A virtual machine queue is formed based on the resource requirements and importance of the virtual machines, and then sorted according to the size of the demand. Virtual machines with higher resource requirements are given priority. In practice, after forming the virtual machine queue, we sort the virtual machines in the queue according to their usage, prioritizing larger virtual machines to reduce the number of migrations and their impact.

[0107] C. Reallocating virtual machines using the robust Firstfit algorithm:

[0108] The robustness is controlled by the Gamma parameter. The Firstfit algorithm is used to place the virtual machine on the most suitable physical host and the available resources of the host are updated in real time during the placement process. The determination of whether a host can be placed is based on the value of the first Γ (Equation (8)). The classic Firstfit algorithm (Bays C. A comparison of next-fit, first-fit, and best-fit[J]. Communications of the ACM,1977,20(3):191-192.) used to solve the bin packing problem traverses each host and places the virtual machine on the first host that can be placed. That is, the traditional method uses a deterministic value (the maximum value of all virtual machines) and determines whether it can be placed based on the maximum value of the virtual machine. All virtual machines are in a large set. However, this invention uses Γ robustness. Only the maximum value of Γ virtual machines needs to be calculated. The mean value of the virtual machines is calculated for the others without calculating the maximum value of all virtual machines, i.e., Equation (8).

[0109] According to the Γ robustness theory: Where N is the total number of virtual machines, and Pr is the violation ratio; once N and Pr are determined, Γ can be calculated. In specific implementation, this invention obtains the value of Γ by limiting a probability, restricting the consideration to the maximum value of the Γ largest virtual machines, while only the mean is considered for the others. This probability value is pre-set (taken to be less than 1%). For example, with a total of 50 virtual machines, if we limit the probability to 0.0002, we can calculate Γ as 25, meaning that the probability of 25 virtual machines simultaneously having their usage at the maximum value is less than 0.0002.

[0110] D. Resource updates and iterations:

[0111] After placing a virtual machine (VM) once, the resource status of all physical hosts is updated, and it is determined whether further VM migration is needed. If necessary, the VM queue is readjusted, and the above process is iterated until the resource requirements of all VMs are met. In practice, steps A, B, and C are repeated. If the number of hosts can be reduced further, readjustment is required; if it cannot be reduced, the algorithm iteration ends. There are two methods for determining whether to stop the iteration: one is to limit the number of iterations, such as iterating the entire process 50 times, and the final result is taken as the optimal solution; the other is to exit the iteration if there is no change after several consecutive iterations, such as if the number of hosts used has not decreased in five consecutive iterations. Ultimately, this achieves the goal of placing all current VMs with fewer hosts, saving costs and thus meeting the resource requirements of the VMs.

[0112] The following specific examples further illustrate the proposed Γ robust optimization model and GammaFF heuristic algorithm, demonstrating their practical application in virtual machine integration.

[0113] Example 1: Application of the Γ Robust Optimization Model

[0114] In this embodiment, we consider a typical cloud computing environment containing 5-13 physical hosts, each hosting a certain number of virtual machines (VMs), with the resource requirements of each VM dynamically changing. The optimization objective is to optimize the resource utilization of the physical hosts by migrating VMs, while minimizing the number of VM migrations. The optimization steps are as follows:

[0115] 1. Initial configuration: All virtual machines are initially assigned to various physical hosts. In the initial state, the resource utilization of physical hosts is uneven, with some hosts being overloaded and others not being fully utilized.

[0116] 2. Gamma Robust Optimization: Using the Γ robust optimization model, the optimal resource allocation scheme for each physical host is calculated based on the dynamic resource requirements of virtual machines. By adjusting the Γ parameter, the optimal state of the system under different uncertainties is determined.

[0117] 3. Migration Decision: Based on the optimization results, determine which virtual machines need to be migrated and to which physical hosts, in order to achieve resource rebalancing.

[0118] The optimization results are shown in Table 1:

[0119] Table 1 Comparison of model integration performance under different initial host numbers

[0120]

[0121] After using the Γ robust optimization model, the resource utilization of physical hosts was improved, and the number of hosts used was reduced.

[0122] Example 2: Application of the GammaFF heuristic algorithm

[0123] To further verify the effectiveness of this invention, the GammaFF heuristic algorithm was used to optimize virtual machine consolidation in a large-scale cloud computing environment. Eight datasets of different sizes and numbers of hosts were selected as test datasets to examine the effectiveness of virtual machine consolidation. The implementation steps are as follows:

[0124] A. Initial physical host selection: Select the target host from physical hosts with resource utilization below 50%, clear all virtual machines on this host, and add these virtual machines to the virtual machine queue.

[0125] B. Virtual Machine Queue Formation: In addition to all virtual machines on the host being emptied, a certain number of virtual machines are randomly selected from other hosts and added to this queue. This is done to enhance the flexibility of the algorithm. All virtual machines are sorted according to their resource requirements from high to low to form a migration queue, prioritizing the processing of virtual machines with higher resource requirements.

[0126] C. Gamma Robustness Firstfit Algorithm: Uses the Gamma parameter to control robustness, adopts the Firstfit approach, places the virtual machine on the most suitable physical host, and updates the available resources of the host in real time during the placement process.

[0127] D. Resource Update and Iteration: After each placement operation, update the resource status of the physical host. If there are still hosts with underutilized resources, readjust the virtual machine queue and repeat the above process until the resource requirements of all virtual machines are met.

[0128] The experimental results are shown in Table 2:

[0129] Table 2 Comparison of integration effects of different algorithms under different initial host numbers

[0130]

[0131] After processing with the GammaFF algorithm, the resource utilization of physical hosts is improved, the number of hosts used is reduced, and the efficiency of resource utilization is increased.

[0132] The specific embodiments demonstrate the application effect of the Gamma robust optimization model and GammaFF heuristic algorithm of the present invention in virtual machine integration, which can effectively improve resource utilization, reduce migration costs, and improve system stability and reliability in large-scale cloud computing environments.

[0133] The above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. For those skilled in the art, other variations or modifications can be made based on the above description. It is impossible to exhaustively list all the implementation methods here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A cloud computing virtual machine scheduling method based on dynamic resource demand and Γ-robust optimization, characterized in that, The method comprises the following steps: 1) constructing a Gamma robust optimization model to minimize the number of active hosts of the cloud computing system; Gamma is an uncertain parameter, and different numbers of Gamma are selected under different probability condition constraints to maximize the use of virtual machines without considering the maximum use of all virtual machines; 2) designing a heuristic algorithm GammaFF to solve the Gamma robust optimization model to realize efficient integration of virtual machines in a large-scale cloud environment; the method comprises the following steps: A. selecting an initial physical host and creating a virtual machine queue; the virtual machine queue comprises all virtual machines on the host; B. expanding the virtual machine queue: a certain number of virtual machines are randomly selected from other hosts and added to the virtual machine queue to form an expanded virtual machine queue, and the virtual machines in the virtual machine queue are sorted according to the size of the resource demand of the virtual machines, and the virtual machines with large resource demand are processed preferentially; C. performing virtual machine reassignment through the Gamma robust Firstfit algorithm, which comprises the following steps: C1. setting a probability value; the value of Gamma is obtained by calculation; C2. judging whether the physical host can be used to place virtual machines: the host can be placed by determining the CPU hotspot threshold constraint, i.e., the value of Gamma; C3. placing the virtual machines on the selected physical host and updating the available resources of the host in real time during the placement process; Thus, the cloud computing virtual machine scheduling based on dynamic resource demand and Gamma robust optimization is realized.

2. The cloud computing virtual machine scheduling method based on dynamic resource demand and robust optimization of claim 1, wherein, The objective function of the Gamma robust optimization model constructed in step 1) comprises two items, i.e., the host cost used and the migration cost. The probability condition constraints of the Gamma robust optimization model constructed comprise: First constraint: the virtual machines are deployed on a unique host; Second constraint: if the virtual machine i is deployed on the host j, then the virtual machine i is in the large set MaxSet or the small set MinSet of the host j; Third constraint: total number constraint of virtual machines deployed on the host j; Fourth constraint: total number constraint of virtual machines in the large set MaxSet of the host j; Fifth constraint: maximum radius constraint of virtual machines on the host j; Eighth constraint: ensuring that the CPU hotspot threshold constraint is not violated; Sixth and seventh constraints: ensure that the split variable S on host j j greater than the maximum in the minimum set and less than the minimum in the maximum set; Ninth constraint: ensuring that the memory capacity constraint is not violated; Tenth constraint: determining whether the virtual machine i is migrated; and variable range constraint. The objective function of the Gamma robust optimization model constructed has the following form:

3. The cloud computing virtual machine scheduling method based on dynamic resource demand and robust optimization of claim 2, wherein, wherein the first item of the objective function is the host cost used, and the second item is the migration cost; parameters α and β are priority limiting parameters of the two items of the objective function; J is the total number of hosts j; and I is the total number of virtual machines i. The value range of the defined decision variable comprises:

4. The cloud computing virtual machine scheduling method based on dynamic resource demand and robust optimization of claim 3, wherein, Set a = 10 3 , β = 1.

5. The cloud computing virtual machine scheduling method based on dynamic resource demand and robust optimization of claim 2, wherein, 6. The cloud computing virtual machine scheduling method based on dynamic resource demand and Gamma robust optimization according to claim 2, wherein: where x ij is a binary decision variable, x ij = 1 means that the virtual machine i is placed on the host j; I is the index of the virtual machine; J is the index of the host; is a binary decision variable, x is a binary decision variable, x j is a binary decision variable, y j = 1 means that the host j is used; H jk is a binary decision variable, H jk = 1 means that the number of virtual machines placed on the host j is k; W i is a binary decision variable, W i = 1 means that the virtual machine i has migrated, W i = 0 means that the virtual machine i has not migrated; S j is a separation variable used to distinguish between the large set and the small set; M j is the maximum radius of the virtual machines allocated on the host j. the first constraint is represented as: the second constraint is represented as: the third constraint is represented as: the fourth constraint is represented as: the fifth constraint is represented as: the sixth and seventh constraints are represented as: the eighth constraint is represented as: the ninth constraint is represented as: the tenth constraint is represented as: ​ 7. The cloud computing virtual machine scheduling method based on dynamic resource demand and robust optimization of claim 6, wherein, The specific process of step C1 is: using Gamma parameter to control robustness, using firstfit algorithm to determine whether the physical host can be placed according to the calculated value of the first Gamma, and placing the virtual machine on the physical host, and updating the available resources of the host in real time during the placement process.

8. The cloud computing virtual machine scheduling method based on dynamic resource demand and robust optimization of claim 7, wherein, Specifically, according to the Gamma robustness theory, the value of Gamma is obtained by limiting a probability.

9. The cloud computing virtual machine scheduling method based on dynamic resource demand and robust optimization of claim 8, wherein, The value of the probability is set to be less than 1%.

Images