CN113296893A - Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm - Google Patents

Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm Download PDF

Info

Publication number
CN113296893A
CN113296893A CN202110595616.8A CN202110595616A CN113296893A CN 113296893 A CN113296893 A CN 113296893A CN 202110595616 A CN202110595616 A CN 202110595616A CN 113296893 A CN113296893 A CN 113296893A
Authority
CN
China
Prior art keywords
population
optimal solution
virtual machine
individual
resource
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.)
Granted
Application number
CN202110595616.8A
Other languages
Chinese (zh)
Other versions
CN113296893B (en
Inventor
李雷孝
李少旭
邓丹
万剑雄
马志强
刘东江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Inner Mongolia University of Technology
Original Assignee
Inner Mongolia University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Inner Mongolia University of Technology filed Critical Inner Mongolia University of Technology
Priority to CN202110595616.8A priority Critical patent/CN113296893B/en
Publication of CN113296893A publication Critical patent/CN113296893A/en
Application granted granted Critical
Publication of CN113296893B publication Critical patent/CN113296893B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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

Abstract

The invention belongs to the technical field of cloud computing, and particularly relates to a cloud platform low-resource-loss virtual machine placement method based on a hybrid sine and cosine particle swarm optimization algorithm. The method comprises the following steps: step 1: acquiring a resource state of a physical machine of a data center and a set of virtual machines to be placed, and randomly generating N particles as an initial population according to a mapping relation between the virtual machines and the physical machine; step 2: generating a reverse population of the initial population in the step 1 by using a reverse learning strategy, comparing individuals in the initial population with reverse individuals thereof, and selecting more optimal individuals as initial individuals of the population; and step 3: and (4) iterative updating, namely updating the global optimal solution of the population and the historical optimal solution of each individual, and updating the speed and the position of each individual of the population by using a hybrid sine and cosine particle swarm optimization algorithm until the maximum iteration time T is reached, so as to obtain the global optimal solution of the population. And 4, step 4: and (4) outputting the population global optimal solution obtained in the step (3), wherein the placement scheme corresponding to the global optimal solution position is the optimal placement scheme.

Description

Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm
Technical Field
The invention belongs to the technical field of cloud computing, and particularly relates to a cloud platform low-resource-loss virtual machine placement method based on a hybrid sine and cosine particle swarm optimization algorithm.
Background
Cloud computing has been rapidly developed in the global scope as a new computing mode following distributed computing, parallel computing and grid computing, and due to the characteristics of a resource pool mechanism which is low in cost, safe and reliable and can be allocated according to needs. With the continuous expansion of the scale of the cloud platform, the problems of serious resource loss and low utilization rate are increasingly highlighted. The virtual machine migration mechanism is an important means for improving the resource utilization rate and reducing the energy consumption of the cloud platform. Virtual machine placement is an important link of virtual machine migration, and the quality of the placement position of the virtual machine affects the resource allocation condition of the cloud platform. If the resource loss of the cloud platform is serious, more physical machines are necessarily required to be started to place the virtual machines, so that resource waste is caused, and the operation cost is increased. Therefore, the method has important application value in constructing the virtual machine placement model with low energy consumption.
N virtual machines are placed on m physical machines, each virtual machine has m choices, and the placement scheme is m in totalnThis is a typical NP-Hard problem, and it is difficult to find the optimal solution in a polynomial time. An important goal of the virtual machine placement problem is to find a near-optimal solution.
The traditional scheme for solving the virtual machine placement problem is a heuristic algorithm based on a greedy thought, such as a First-time adaptation (FF), an optimal adaptation (Best Fit, BF) and the like, and aims to reduce the number of physical machines of a cloud platform in an active state as much as possible to reduce energy consumption and improve resource utilization rate. However, the virtual machine placing strategy based on the greedy thought may cause some physical machines to be overloaded and reduce the service quality of the virtual machines running on the physical machines, and some physical machines are in a low-load state and cause load imbalance.
In recent years, some machine learning based methods have been used to solve the virtual machine placement problem. One is a virtual machine placement strategy under reinforcement Learning, the strategy takes energy consumption as an optimization target, a Q-Learning (lambda) algorithm is optimized from two aspects of state aggregation and time reliability, and the energy consumption of a cloud platform can be effectively reduced. One is a virtual machine placement strategy based on fuzzy membership, and by means of the idea of fuzzy clustering, a membership matrix of a virtual machine and a physical machine is calculated, and an optimal placement scheme is obtained by local search in the matrix.
The meta-heuristic algorithm is also widely applied to the virtual machine placement problem, wherein a weight fitness-based particle swarm optimization algorithm takes communication delay, server load and processing delay of a CPU (central processing unit) as the basis for virtual machine placement, rewrites the weight factor omega of the PSO algorithm, and can better reduce the application execution time. Another virtual machine placement strategy that mixes the sine and cosine algorithm and the casn sea squirt group algorithm reduces energy consumption by minimizing the number of active physical machines. And a hybrid fuzzy ant colony optimization algorithm for scheduling the virtual machines takes response time, processing time and load balance as optimization targets, calculates pheromone values according to historical information, and selects a proper physical machine for the virtual machines.
The above approach lacks a tradeoff between resource consumption and service level.
Disclosure of Invention
Aiming at the existing problems, the invention provides a cloud platform low-resource-loss virtual machine placement method based on a hybrid sine-cosine particle swarm optimization algorithm, which can reduce the resource loss of the cloud platform as much as possible and ensure the service quality of users.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
1. a cloud platform low-resource-loss virtual machine placement method based on a hybrid sine and cosine particle swarm optimization algorithm is characterized by comprising the following steps of: the method comprises the following steps:
step 1: acquiring resource states of a physical machine of a data center and a set of virtual machines to be placed, randomly generating N particles as an initial population according to a mapping relation between the virtual machines and the physical machine, and encoding positions of the N particles to obtain N placing schemes;
step 2: generating a reverse population of the initial population in the step 1 by using a reverse learning strategy, taking Lose as a fitness index, comparing the individuals in the initial population with the corresponding reverse individuals, and selecting better individuals as the initial individuals of the population;
and step 3: and (3) iterative updating, namely updating the global optimal solution of the population and the historical optimal solution of each individual by taking the Lose as a fitness index, and updating the speed and the position of each individual of the population by utilizing a hybrid sine and cosine particle swarm optimization algorithm until the maximum iteration time T is reached to obtain the global optimal solution of the population.
And 4, step 4: and (4) outputting the population global optimal solution obtained in the step (3), wherein the placement scheme corresponding to the global optimal solution position is the optimal placement scheme.
Preferably, the step 1 comprises:
step 11: establishing a physical machine set:
pmj＝{idj,cpuj,memj,bwj,diskj,alivej} (1)，
wherein:
idjthe unique identification of the jth physical machine on the cloud platform is represented,
cpujindicates the CPU core number of the j physical machine,
memjthe memory size of the jth physical machine is shown,
bwjindicates the bandwidth size of the jth physical machine,
diskjindicating the disk size of the jth physical machine,
aliveje {0, 1} represents whether the jth physical machine is in an active state;
step 12: let vmiCreation information representing the ith virtual machine, virtual machine vm of cloud platformiThe resource requirements of (a) can be formally described as a formula:
vmi＝{idi,cpui,memi,bwi,diski} (2)，
wherein:
idia number indicating that the ith virtual machine is on the cloud platform,
cpuiindicates the number of CPU cores of the ith virtual machine,
memithe memory size of the ith virtual machine is shown,
bwiindicating the bandwidth size of the ith virtual machine,
diskirepresenting the disk size of the ith virtual machine;
step 13: placing a virtual machine onto a physical machine can be abstracted as a resource mapping from the virtual machine to the physical machine, and is described as: coding the placement scheme according to the mapping relation between the virtual machines and the physical machines, wherein each virtual machine can only map to one physical machine by taking the virtual machine as a unit, coding of a solution is represented by using an m-dimensional vector, each attribute in the vector is a physical machine number, and x is { x ═ x { (X })1，x2，…，xn}，xiE {1, 2.., m }, i e {1, 2, …, n }, and vector xiIs n, and each dimension takes the value of [1, m]The whole number in between is an integer of,
randomly generating N particles, namely, the size of a population is N, the dimensionality is D, the maximum iteration time is T, and the position of the ith particle is as follows: x is the number ofi＝(xi1，xi2，…，xiD) The velocity of the ith particle is: v. ofi＝(vi1，vi2，…，viD) (i ═ 1, 2, …, N), where xidRepresents the position of the ith particle in the d-dimension, and the range of the particle position is [ x ]min，xmax]Obtaining N placement schemes, v, by using the above-mentioned encoding method according to the positions of N particlesidRepresents the speed of the ith particle in the direction of the d dimension, and the range of the particle speed is [ vmin，vmax]；
Step 14: after the virtual machine is placed on the physical machine, the resource loss degree of the data center is described as follows: suppose a physical machinepmjK virtual machines are operated on the system, and the resource set of the physical machine is as follows: r { cpu, mem, bw, disk }, R ∈ RjIndicating a physical machine pmjR type resource allocation amount, riRepresentation deployment at physical machine pmjVirtual machine vm of (3)i(i ═ 1, 2, …, k) of resource allocation amounts of r type;
physical machine pmjR type resource utilization rate Uj rThe formula is as follows:
Figure BDA0003090946750000051
taking the average value of the utilization rates of various resources in the cloud platform as the average resource utilization rate of the cloud platform, taking m as the number of physical machines in the cloud platform, and using UavgThe total average resource utilization rate of the cloud platform is represented by the following calculation formula:
Figure BDA0003090946750000052
measuring the deviation degree Lose of the cloud platform resources by using the standard deviation of the resource utilization rate of each dimension of each physical machine, wherein m is the number of the physical machines in the cloud platform, and the calculation formula of Lose is as follows:
Figure BDA0003090946750000053
preferably, the specific process of step 2 is as follows:
x for the ith individual in the populationi：
xi＝(xi1，xi2，…，xiD)，
xiIs solved by xi -Comprises the following steps:
xi -＝(m+1-xi1,m+1-xi2,...,m+1-xiD) (6)；
where m is the number of physical machines and the value of the inverse solution is each of the positions of the solutionsTaking the inverse of a bit, using equation (6) to generate xiIs solved by xi -If xi -The Lose value of the corresponding placement scheme is less than xiLose value of corresponding Placement scheme, Explanation of xi -Ratio xiExcellent, x is usedi -Substitution xi(ii) a Otherwise, x is statedi -Is not comparable to xiExcellent, no need of replacement.
Preferably, the specific process of step 3 includes:
step 31: taking the Lose evaluation index in the step 1 as a fitness function, and updating the global optimal solution of the population and the historical optimal solution of each individual;
step 32: acquiring a weight factor of the current iteration number according to a weight updating formula;
step 33: applying the weight obtained in the step 32 to a mixed sine and cosine particle swarm optimization algorithm, and updating the speed and the position of each individual in the population by using the mixed sine and cosine particle swarm optimization algorithm;
step 34: if the iteration of step 33 reaches the maximum iteration number, the next step is performed; otherwise, the iteration is continued by returning to step 31.
Preferably, the specific process of step 31 is:
the global optimal solution of the population is the individual with the minimum Lose value in the iterative process of the whole population, and the updating rule is as follows:
selecting the optimal solution of the current population, namely the individual with the minimum Lose value, from the population of the 1 st generation, setting the optimal solution as the global optimal solution of the population, selecting the optimal solution of the current population, namely the individual with the minimum Lose value, from the population of each generation thereafter, comparing the optimal solution with the global optimal solution of the population, and if the Lose value of the optimal solution of the current population is smaller than the global optimal solution, setting the optimal solution of the current population as the global optimal solution;
the historical optimal solution of the individual is the individual with the minimum Lose value experienced by the current individual in the iterative process, and the updating rule is as follows:
in the 1 st generation, the Lose value of each individual is set as the historical optimal solution of each individual, in each generation of the population, the Lose value of a new individual is compared with the historical optimal solution of the individual, and if the Lose value of the new individual is smaller than the historical optimal solution of the individual, the new individual is set as the historical optimal solution of the new individual.
Preferably, the specific process of step 32 is: let the weight factor keeping the original speed be ω, and the update formula of ω is:
Figure BDA0003090946750000071
wherein, ω ismaxAnd ωminThe maximum weight and the minimum weight of the population are obtained, T is the current iteration frequency, and T is the maximum iteration frequency of the population.
Preferably, the specific process of step 33 is:
the optimal positions searched for by the ith particle so far are: p is a radical ofi＝(pi1，pi2，…，piD) The population has so far been searched for the optimal positions: g ═ g (g)1，g2，…，gD) (ii) a The particle velocity and position are updated using the following equations:
if r > rc：
vi(t+1)＝w·vi(t)+c1r1[pi(t)-xi(t)]+c2r2[g(t)-xi(t)] (8)；
Figure BDA0003090946750000072
Otherwise:
vi(t+1)＝vi(t) (10)；
Figure BDA0003090946750000073
wherein r iscFor the disturbance probability, the value is 0.1, r is [0,1]]M is the number of physical machines, t represents the current iteration number, and omega is the hold valueWeighting factor of the original velocity, c1Learning ability factors for individuals into themselves, c2The individual learning ability factors to the population are generally 1.5 r1To the extent of learning to self, r2 to the extent of learning to the population, r1And r2Is [0,1]]A uniform random number between r3To the extent of learning towards the optimal solution, r3Take the value of [0, 2 pi]Uniformly distributed random numbers r4Is [0, 2 ]]Uniformly distributed random numbers, controlling the distance between the individual and the current global optimal solution, r5To take on a value of [0,1]The random number of (2) controls the sine and cosine transition probability.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a virtual machine placing method based on a hybrid sine and cosine particle swarm optimization algorithm. The strategy considers four computer system resources including a CPU, a memory, a network bandwidth and a disk, analyzes the ubiquitous resource loss in a cloud platform, provides an optimized objective function model for minimizing the resource loss, and improves the codes and the weights on the basis of a standard particle swarm algorithm, so that the strategy is more suitable for solving the virtual machine placement problem. The hybrid sine and cosine algorithm can effectively avoid trapping in a local optimal solution, and improves the solving precision. Meanwhile, the population convergence speed can be increased by combining a population initialization strategy based on reverse learning.
The simulation experiment results show that under the condition that the resource allocation of the cloud platform and the virtual machine request are certain, the method can reduce the resource loss of the cloud platform better than a standard particle swarm algorithm, an optimal adaptation method and a first-time adaptation method, further reduce the energy consumption and reduce the operation cost of the cloud platform.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention.
In the drawings:
FIG. 1 is a schematic flow chart of a method for placing a cloud platform low-resource-consumption virtual machine based on a hybrid sine-cosine particle swarm optimization algorithm;
FIG. 2 is a diagram of a virtual machine placement scenario in accordance with the present invention;
FIG. 3 is an exemplary graph of resource consumption according to the present invention;
FIG. 4 is a transformation of the position and code of the present invention;
FIG. 5 resource consumption graph after virtual machine placement;
FIG. 6 SLA violation rate after virtual machine placement;
FIG. 7 is a graph of convergence rate versus number of virtual machines.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
Referring to the accompanying drawings 1-7, the method for placing the cloud platform low-resource-loss virtual machine based on the hybrid sine-cosine particle swarm optimization algorithm comprises the following steps:
step 1: acquiring resource states of physical machines of the data center and a set of virtual machines to be placed, randomly generating N particles as initial populations according to the mapping relation between the virtual machines and the physical machines, and obtaining N placing schemes for position codes of the N particles. The method comprises the following specific steps:
step 11: first, a cloud platform resource structure is described. The cloud platform virtualizes a large amount of idle computer system resources to form a virtual resource pool which can be allocated according to needs and dynamically expanded. The virtual machine is a basic unit for resource allocation of the cloud platform to the user. The virtual machine placement problem refers to that a proper physical host is selected for the virtual machine according to a certain placement strategy under the condition of meeting resource constraints, so that the purposes of reducing resource loss, saving operation cost, improving operation efficiency and the like are achieved.
Then, a physical machine set is established, physical resources provided by the cloud platform mainly comprise a CPU, a memory, a network bandwidth and a disk, and a physical machine pm is establishedjFormula of set of:
pmj＝{idj,cpuj,memj,bwj,diskj,alivej} (1)，
wherein:
idjthe unique identification of the jth physical machine on the cloud platform is represented,
cpujindicates the CPU core number of the j physical machine,
memjthe memory size of the jth physical machine is shown,
bwjindicates the bandwidth size of the jth physical machine,
diskjindicating the disk size of the jth physical machine,
aliveje {0, 1} indicates whether the jth physical machine is in an active state.
Step 12: let vmiCreation information representing the ith virtual machine, virtual machine vm of cloud platformiThe resource requirements of (a) can be formally described as a formula:
vmi＝{idi,cpui,memi,bwi,diski} (2)，
wherein:
idia number indicating that the ith virtual machine is on the cloud platform,
cpuiindicates the number of CPU cores of the ith virtual machine,
memithe memory size of the ith virtual machine is shown,
bwiindicating the bandwidth size of the ith virtual machine,
diskirepresenting the disk size of the ith virtual machine;
step 13: fig. 1 illustrates a scenario in which a user makes a resource request to a cloud platform, and the cloud platform allocates the resource request to a physical node. User requests to create n virtual machines { vm) from cloud platformiI is 1, 2, …, n, m existing physical machines of the cloud platform { pm }jJ-1, 2, …, m, these virtual machines are reasonably placed on m physical machines. One physical machine can place a plurality of virtual machines, and one virtual machine can be placed on only one physical machine.
Coding the placement scheme according to the mapping relation between the virtual machines and the physical machines, wherein each virtual machine can only map to one physical machine by taking the virtual machine as a unit, coding of a solution is represented by using an m-dimensional vector, each attribute in the vector is a physical machine number, and x is { x ═ x { (X })1，x2，…，xn}，xiE {1, 2.., m }, i e {1, 2, …, n }, and vector xiIs n, and each dimension takes the value of [1, m]An integer in between.
Randomly generating N particles, namely, the size of a population is N, the dimensionality is D, the maximum iteration time is T, and the position of the ith particle is as follows: x is the number ofi＝(xi1，xi2，…，xiD) The velocity of the ith particle is: v. ofi＝(vi1，vi2，…，viD) (i ═ 1, 2, …, N), where xidRepresents the position of the ith particle in the d-dimension, and the range of the particle position is [ x ]min，xmax]Obtaining N placement schemes, v, by using the above-mentioned encoding method according to the positions of N particlesidRepresents the speed of the ith particle in the direction of the d dimension, and the range of the particle speed is [ vmin，vmax]。
Step 14: because a plurality of heterogeneous virtual machines are often operated on the same physical machine, the virtual machines with different specifications have different occupancy rates on computer resources, which may cause the resource utilization rate of a certain dimension of the physical machine to be too high, and the resource utilization rates of other dimensions to be too low. Unreasonable virtual machine placement positions can cause load unbalance of physical machines, resources of other dimensions cannot be fully utilized, and computer resources are greatly lost.
Fig. 2 is a resource consumption case of the cloud platform when only CPU and memory resources are considered. 3 heterogeneous virtual machines are placed on a certain physical machine, and the remaining available resources are 20% of CPU and 30% of memory. If the user requests a virtual machine with resource requirements of 30% of CPU and 30% of memory, the cloud platform starts a new physical machine to process the virtual machine creation request due to insufficient CPU resources of the physical machine. This approach will result in wasted resources and system power consumption.
After the virtual machine is placed on the physical machine, the resource loss degree of the data center is described as follows: suppose a physical machine pmjK virtual machines are operated on the system, and the resource set of the physical machine is as follows: r { cpu, mem, bw, disk }, R ∈ RjIndicating a physical machine pmjR type resource allocation amount, riRepresentation deployment at physical machine pmjVirtual machine vm of (3)i(i ═ 1, 2, …, k) of resource allocation amounts of r type;
physical machine pmjR type resource utilization rate Uj rThe formula is as follows:
Figure BDA0003090946750000121
taking the average value of the utilization rates of various resources in the cloud platform as the average resource utilization rate of the cloud platform, taking m as the number of physical machines in the cloud platform, and using UavgThe total average resource utilization rate of the cloud platform is represented by the following calculation formula:
Figure BDA0003090946750000122
measuring the deviation degree Lose of the cloud platform resources by using the standard deviation of the resource utilization rate of each dimension of each physical machine, wherein m is the number of the physical machines in the cloud platform, and the calculation formula of Lose is as follows:
Figure BDA0003090946750000123
step 2: and (3) generating a reverse population of the initial population in the step (1) by using a reverse learning strategy, taking the Lose as a fitness index, comparing the individuals in the initial population with the corresponding reverse individuals, and selecting better individuals as the initial individuals of the population. The diversity of the initial solution is increased, and the specific steps of enabling the initial population to be closer to the optimal solution are as follows:
x for the ith individual in the populationi：
xi＝(xi1，xi2，…，xiD)，
xiIs solved by xi -Comprises the following steps:
xi -＝(m+1-xi1,m+1-xi2,...,m+1-xiD) (6)；
wherein m is the number of physical machines, the reverse solution value is the reverse value of each bit of the solution position, and x is generated by using a formula (6)iIs solved by xi -If xi -The Lose value of the corresponding placement scheme is less than xiLose value of corresponding Placement scheme, Explanation of xi -Ratio xiExcellent, x is usedi -Substitution xi(ii) a Otherwise, x is statedi -Is not comparable to xiExcellent, no need of replacement.
The specific process of the step 3 comprises the following steps:
step 31: and (3) updating the global optimal solution of the population and the historical optimal solution of each individual by taking the Lose evaluation index in the step (1) as a fitness function. The method specifically comprises the following steps: the global optimal solution of the population is the individual with the minimum Lose value in the iterative process of the whole population, and the updating rule is as follows:
selecting the optimal solution of the current population, namely the individual with the minimum Lose value, from the population of the 1 st generation, setting the optimal solution as the global optimal solution of the population, selecting the optimal solution of the current population, namely the individual with the minimum Lose value, from the population of each generation thereafter, comparing the optimal solution with the global optimal solution of the population, and if the Lose value of the optimal solution of the current population is smaller than the global optimal solution, setting the optimal solution of the current population as the global optimal solution;
the historical optimal solution of the individual is the individual with the minimum Lose value experienced by the current individual in the iterative process, and the updating rule is as follows:
in the 1 st generation, the Lose value of each individual is set as the historical optimal solution of each individual, in each generation of the population, the Lose value of a new individual is compared with the historical optimal solution of the individual, and if the Lose value of the new individual is smaller than the historical optimal solution of the individual, the new individual is set as the historical optimal solution of the new individual.
Step 32: and acquiring the weight factor of the current iteration number according to the weight updating formula. The specific process is as follows: let the weight factor keeping the original speed be ω, and the update formula of ω is:
Figure BDA0003090946750000141
wherein, ω ismaxAnd ωminThe maximum weight and the minimum weight of the population are obtained, T is the current iteration frequency, and T is the maximum iteration frequency of the population.
Step 33: and (4) applying the weight obtained in the step (32) to a mixed sine and cosine particle swarm optimization algorithm, and updating the speed and the position of each individual in the population by using the mixed sine and cosine particle swarm optimization algorithm. The method specifically comprises the following steps:
the optimal positions searched for by the ith particle so far are: p is a radical ofi＝(pi1，pi2，…，piD) The population has so far been searched for the optimal positions: g ═ g (g)1，g2，…，gD) (ii) a The particle velocity and position are updated using the following equations:
if r > rc：
vi(t+1)＝w·vi(t)+c1r1[pi(t)-xi(t)]+c2r2[g(t)-xi(t)] (8)；
Figure BDA0003090946750000142
Otherwise:
vi(t+1)＝vi(t) (10)；
Figure BDA0003090946750000143
wherein r iscFor the disturbance probability, the value is 0.1, r is [0,1]]M is the number of physical machines, t represents the current iteration number, and omega is the original valueWeight factor of velocity, c1Learning ability factors for individuals into themselves, c2The individual learning ability factors to the population are generally 1.5 r1To the extent of learning to self, r2 to the extent of learning to the population, r1And r2Is [0,1]]A uniform random number between r3To the extent of learning towards the optimal solution, r3Take the value of [0, 2 pi]Uniformly distributed random numbers r4Is [0, 2 ]]Uniformly distributed random numbers, controlling the distance between the individual and the current global optimal solution, r5To take on a value of [0,1]The random number of (2) controls the sine and cosine transition probability.
Step 34: if the iteration of step 33 reaches the maximum iteration time T, the next step is carried out; otherwise, the iteration is continued by returning to step 31.
Example (b):
the method adopts Python language to realize the algorithm, and verifies the effectiveness of the sine and cosine particle swarm optimization algorithm in reducing resource loss on the problem of virtual machine placement by comparing and analyzing the algorithm with a standard particle swarm optimization algorithm, an optimal adaptation method and a first-time adaptation method.
The experiment simulates 50 isomorphic physical machines to form a cloud platform, and the physical machines are configured to 36-core CPU, 95GB memory, 10000MB bandwidth and 10T disk.
In the experiment, the optimal adaptation algorithm sequentially selects the physical machine which can meet the requirements of the virtual machine and has the minimum residual resources among all the physical machines in the active state to place. Firstly, the physical machines are sorted in an ascending order by taking the memory utilization rate as a standard, and whether the physical machines are placed is sequentially judged, so that the selected physical machines meet the requirements. And if the last physical machine is checked to be still not placed, the new physical machine is required to be awakened for placing.
And the first-time adaptation algorithm sequentially judges whether the virtual machine can be placed according to the serial number of the physical machine, if so, the virtual machine is placed on the physical machine, and otherwise, the next physical machine is judged. And if the physical machines cannot be placed, the new physical machine is required to be awakened to be placed.
According to the virtual machine parameter specification provided by the browser package of the OpenStack, the bandwidth of the virtual machine is limited, and the specific parameter specification of the virtual machine is shown in the table 1.
TABLE 1 virtual machine Specifications
Figure BDA0003090946750000161
The main parameters of the sine and cosine particle swarm optimization algorithm are shown in table 2. [ x ] ofmin，xmax]Is the range of particle positions, [ v ]min，vmax]The particle speed range is used for restraining the particles so as to prevent the particles from crossing the boundary or missing the optimal solution at an excessive speed. r iscIs the transition probability of equations (8), (9), (10), (11) and is used to switch the location update equations.
TABLE 2 sine and cosine particle swarm optimization algorithm parameters
Figure BDA0003090946750000162
And (3) analyzing an experimental result:
(1) resource loss analysis:
the resource loss of the invention is an evaluation index, under the request of virtual machines with the same quantity and scale, the resource loss of each placement strategy is compared, and the resource loss of the optimal placement scheme obtained by each strategy is obtained according to a formula (5). Because the value interval of the resource utilization rate is [0,1], the data obtained when the standard deviation is calculated is small, the percentage number of the resource utilization rate is removed, and the average value of each strategy is taken as a result after 20 times of execution.
Fig. 4 illustrates resource consumption conditions requested by 4 sets of virtual machines by the sine and cosine particle swarm optimization algorithm, the BF algorithm, and the FF algorithm. As can be seen from the figure, the resource consumption of the SCPSO algorithm is smaller than that of the BF algorithm, the FF algorithm, and the PSO algorithm for the same number of virtual machine requests. The experimental result shows that the resource loss of the cloud platform does not change linearly with the number of the virtual machines, but is directed at the resource distribution condition of the cloud platform. If the current cloud platform resources are not enough to place all the virtual machines, the cloud platform needs to start a new physical machine to bear the resource request of the virtual machine. After the optimization strategy is carried out, the virtual machines are uniformly distributed to all the physical machines, so that the resource distribution of the cloud platform is relatively uniform, and the resource loss is reduced.
(2) SLA violation rate analysis:
the service quality is one of important indexes for evaluating the cloud service, and the SLA is generally related to the resource utilization rate of a physical machine. As the resource utilization of the physical machine increases, the quality of service quickly degrades. The embodiment defines the SLA violation rate of the cloud platform as:
fSLA＝ln(2+Uavg-θ) (12)，
fig. 5 illustrates the change of the service quality of the cloud platform as the number of requests of the virtual machines increases in the SCPSO algorithm, the PSO algorithm, the FF algorithm and the BF algorithm. Compared with the traditional FF algorithm, the SLA violation rate of the SCPSO algorithm is reduced by 7 percent on average, 6 percent compared with the SLA violation rate of the BF algorithm and 1 percent compared with the SLA violation rate of the standard PSO algorithm. Therefore, the SCPSO algorithm can better ensure the service quality of the user. Traditional heuristic algorithms based on greedy strategies, such as BF (back-propagation) algorithm and FF (field-off) algorithm, mainly aim at reducing the number of physical machines in an active state as much as possible to improve the resource utilization rate, conflict with the aim of improving the service quality as much as possible, and inevitably cause the reduction of the service quality. The virtual machine placement strategy provided by the invention can reduce resource loss as much as possible and ensure the service quality of users.
(3) And (3) comparison of convergence rate:
FIG. 6 illustrates the average convergence speed of the SCPSO algorithm and the PSO algorithm under different virtual machine requests. The convergence rate of the SCPSO algorithm is improved by 50% compared with the PSO algorithm. The SCPSO algorithm is fused with a reverse learning strategy, an encoding method is improved, a proper weight updating formula is used, and a high-quality solution can be obtained while the convergence rate is improved by mixing a sine and cosine algorithm.
In summary, a virtual machine placement strategy based on the SCPSO algorithm is proposed herein. The strategy considers four computer system resources including a CPU, a memory, a network bandwidth and a disk, analyzes the ubiquitous resource loss in a cloud platform, provides an optimized objective function model for minimizing the resource loss, and improves the codes and the weights on the basis of a standard particle swarm algorithm, so that the strategy is more suitable for solving the virtual machine placement problem. The hybrid sine and cosine algorithm can effectively avoid trapping in a local optimal solution, and improves the solving precision. Meanwhile, the population convergence speed can be increased by combining a population initialization strategy based on reverse learning.
The simulation experiment results show that under the condition that the resource allocation of the cloud platform and the virtual machine request are certain, the method can reduce the resource loss of the cloud platform better than a standard particle swarm algorithm, an optimal adaptation method and a first-time adaptation method, further reduce the energy consumption and reduce the operation cost of the cloud platform. And next, performing multi-objective optimization on the virtual machine scheduling problem, and further improving the comprehensive performance of the cloud platform.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (7)

1. A cloud platform low-resource-loss virtual machine placement method based on a hybrid sine and cosine particle swarm optimization algorithm is characterized by comprising the following steps of: the method comprises the following steps:
step 1: acquiring resource states of a physical machine of a data center and a set of virtual machines to be placed, randomly generating N particles as an initial population according to a mapping relation between the virtual machines and the physical machine, and acquiring N placing schemes according to positions of the N particles;
step 2: generating a reverse population of the initial population in the step 1 by using a reverse learning strategy, taking Lose as a fitness index, comparing the individuals in the initial population with the corresponding reverse individuals, and selecting better individuals as the initial individuals of the population;
and step 3: performing iterative updating, namely updating the global optimal solution of the population and the historical optimal solution of each individual by using Lose as a fitness index, and updating the speed and the position of each individual of the population by using a hybrid sine-cosine particle swarm optimization algorithm until the maximum iteration time T is reached to obtain the global optimal solution of the population;
and 4, step 4: and (4) outputting the population global optimal solution obtained in the step (3), wherein the placement scheme corresponding to the global optimal solution position is the optimal placement scheme.
2. The method for placing the cloud platform low-resource-consumption virtual machine based on the hybrid sine-cosine particle swarm optimization algorithm according to claim 1, is characterized in that: the step 1 comprises the following steps:
step 11: establishing a physical machine set:
pmj＝{idj,cpuj,memj,bwj,diskj,alivej} (1)，
wherein:
idjthe unique identification of the jth physical machine on the cloud platform is represented,
cpujindicates the CPU core number of the j physical machine,
memjthe memory size of the jth physical machine is shown,
bwjindicates the bandwidth size of the jth physical machine,
diskjindicating the disk size of the jth physical machine,
aliveje {0, 1} represents whether the jth physical machine is in an active state;
step 12: let vmiCreation information representing the ith virtual machine, virtual machine vm of cloud platformiThe resource requirements of (a) can be formally described as a formula:
vmi＝{idi,cpui,memi,bwi,diski} (2)，
wherein:
idiediting of ith virtual machine on cloud platformThe number of the mobile station is,
cpuiindicates the number of CPU cores of the ith virtual machine,
memithe memory size of the ith virtual machine is shown,
bwiindicating the bandwidth size of the ith virtual machine,
diskirepresenting the disk size of the ith virtual machine;
step 13: placing a virtual machine onto a physical machine can be abstracted as a resource mapping from the virtual machine to the physical machine, and is described as: coding the placement scheme according to the mapping relation between the virtual machines and the physical machines, wherein each virtual machine can only map to one physical machine by taking the virtual machine as a unit, coding of a solution is represented by using an m-dimensional vector, each attribute in the vector is a physical machine number, and x is { x ═ x { (X })1，x2，…，xn}，xiE {1, 2.., m }, i e {1, 2, …, n }, and vector xiIs n, and each dimension takes the value of [1, m]An integer in between;
randomly generating N particles, namely, the size of a population is N, the dimensionality is D, the maximum iteration time is T, and the position of the ith particle is as follows: x is the number ofi＝(xi1，xi2，…，xiD) The velocity of the ith particle is: v. ofi＝(vi1，vi2，…，viD) (i ═ 1, 2, …, N), where xidRepresents the position of the ith particle in the d-dimension, and the range of the particle position is [ x ]min，xmax]Obtaining N placement schemes, v, by using the above-mentioned encoding method according to the positions of N particlesidRepresents the speed of the ith particle in the direction of the d dimension, and the range of the particle speed is [ vmin，vmax]；
Step 14: after the virtual machine is placed on the physical machine, the resource loss degree of the data center is described as follows: suppose a physical machine pmjK virtual machines are operated on the system, and the resource set of the physical machine is as follows: r { cpu, mem, bw, disk }, R ∈ RjIndicating a physical machine pmjR type resource allocation amount, riRepresentation deployment at physical machine pmjVirtual machine vm of (3)i(i＝1，2, …, k);
physical machine pmjR type resource utilization rate Uj rThe formula is as follows:
Figure FDA0003090946740000031
taking the average value of the utilization rates of various resources in the cloud platform as the average resource utilization rate of the cloud platform, taking m as the number of physical machines in the cloud platform, and using UavgThe total average resource utilization rate of the cloud platform is represented by the following calculation formula:
Figure FDA0003090946740000032
measuring the deviation degree Lose of the cloud platform resources by using the standard deviation of the resource utilization rate of each dimension of each physical machine, wherein m is the number of the physical machines in the cloud platform, and the calculation formula of Lose is as follows:
Figure FDA0003090946740000033
3. the method for placing the cloud platform low-resource-loss virtual machine based on the hybrid sine-cosine particle swarm optimization algorithm according to claim 2, is characterized in that: the specific process of the step 2 is as follows:
x for the ith individual in the populationi：
xi＝(xi1，xi2，…，xiD)，
xiIs solved by xi -Comprises the following steps:
xi -＝(m+1-xi1,m+1-xi2,...,m+1-xiD) (6)；
wherein m is the number of physical machines, the reverse solution value is the reverse value of each bit of the solution position, and formula (6) is used to generateTo xiIs solved by xi -If xi -The Lose value of the corresponding placement scheme is less than xiLose value of corresponding Placement scheme, Explanation of xi -Ratio xiExcellent, x is usedi -Substitution xi(ii) a Otherwise, x is statedi -Is not comparable to xiExcellent, no need of replacement.
4. The method for placing the cloud platform low-resource-consumption virtual machine based on the hybrid sine-cosine particle swarm optimization algorithm according to claim 3, wherein the method comprises the following steps: the specific process of the step 3 comprises the following steps:
step 31: taking the Lose evaluation index in the step 1 as a fitness function, and updating the global optimal solution of the population and the historical optimal solution of each individual;
step 32: acquiring a weight factor of the current iteration number according to a weight updating formula;
step 33: applying the weight obtained in the step 32 to a mixed sine and cosine particle swarm optimization algorithm, and updating the speed and the position of each individual in the population by using the mixed sine and cosine particle swarm optimization algorithm;
step 34: if the iteration of step 33 reaches the maximum iteration number, the next step is performed; otherwise, the iteration is continued by returning to step 31.
5. The method for placing the cloud platform low-resource-consumption virtual machine based on the hybrid sine-cosine particle swarm optimization algorithm according to claim 3, wherein the method comprises the following steps: the specific process of step 31 is as follows:
the global optimal solution of the population is the individual with the minimum Lose value in the iterative process of the whole population, and the updating rule is as follows:
selecting the optimal solution of the current population, namely the individual with the minimum Lose value, from the population of the 1 st generation, setting the optimal solution as the global optimal solution of the population, selecting the optimal solution of the current population, namely the individual with the minimum Lose value, from the population of each generation thereafter, comparing the optimal solution with the global optimal solution of the population, and if the Lose value of the optimal solution of the current population is smaller than the global optimal solution, setting the optimal solution of the current population as the global optimal solution;
the historical optimal solution of the individual is the individual with the minimum Lose value experienced by the current individual in the iterative process, and the updating rule is as follows:
in the 1 st generation, the Lose value of each individual is set as the historical optimal solution of each individual, in each generation of the population, the Lose value of a new individual is compared with the historical optimal solution of the individual, and if the Lose value of the new individual is smaller than the historical optimal solution of the individual, the new individual is set as the historical optimal solution of the new individual.
6. The method for placing the cloud platform low-resource-consumption virtual machine based on the hybrid sine-cosine particle swarm optimization algorithm according to claim 4, wherein the method comprises the following steps: the specific process of the step 32 is as follows: let the weight factor keeping the original speed be ω, and the update formula of ω is:
Figure FDA0003090946740000051
wherein, ω ismaxAnd ωminThe maximum weight and the minimum weight of the population are obtained, T is the current iteration frequency, and T is the maximum iteration frequency of the population.
7. The method for placing the cloud platform low-resource-consumption virtual machine based on the hybrid sine-cosine particle swarm optimization algorithm according to claim 5, wherein the method comprises the following steps: the specific process of step 33 is as follows:
the optimal positions searched for by the ith particle so far are: p is a radical ofi＝(pi1，pi2，…，piD) The population has so far been searched for the optimal positions: g ═ g (g)1，g2，…，gD) (ii) a The particle velocity and position are updated using the following equations:
if r > rc：
Figure FDA0003090946740000061
Otherwise:
vi(t+1)＝vi(t) (10)；
Figure FDA0003090946740000062
wherein r iscFor the disturbance probability, the value is 0.1, r is [0,1]]M is the number of physical machines, t represents the current iteration number, omega is a weight factor for keeping the original speed, c1Learning ability factors for individuals into themselves, c2The individual learning ability factors to the population are generally 1.5 r1To the extent of learning to self, r2 to the extent of learning to the population, r1And r2Is [0,1]]A uniform random number between r3To the extent of learning towards the optimal solution, r3Take the value of [0, 2 pi]Uniformly distributed random numbers r4Is [0, 2 ]]Uniformly distributed random numbers, controlling the distance between the individual and the current global optimal solution, r5To take on a value of [0,1]Random number of (c), control of sine and cosine transition probability, vi(t) is the velocity of the particles in the t-th generation, vi(t +1) is the velocity of the particle at the t ═ t +1 generation, xi(t) is the position of the t-th generation particle, xi(t +1) is the position of the particle at the t-th (t +1) th generation, pi(t) is the individual optimal solution in the t generation, and g (t) is the global optimal solution in the t generation.
CN202110595616.8A 2021-05-29 2021-05-29 Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm Active CN113296893B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110595616.8A CN113296893B (en) 2021-05-29 2021-05-29 Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110595616.8A CN113296893B (en) 2021-05-29 2021-05-29 Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm

Publications (2)

Publication Number Publication Date
CN113296893A true CN113296893A (en) 2021-08-24
CN113296893B CN113296893B (en) 2022-09-02

Family

ID=77326084

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110595616.8A Active CN113296893B (en) 2021-05-29 2021-05-29 Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm

Country Status (1)

Country Link
CN (1) CN113296893B (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8972405B1 (en) * 2012-07-26 2015-03-03 Emc Corporation Storage resource management information modeling in a cloud processing environment
CN104484715A (en) * 2014-11-28 2015-04-01 江苏大学 Neural network and particle swarm optimization algorithm-based building energy consumption predicting method
CN106657238A (en) * 2016-10-13 2017-05-10 内蒙古工业大学 Interactive-workload-oriented data center coarse-grained dynamic server reservation algorithm
US20170329048A1 (en) * 2016-05-12 2017-11-16 The Climate Corporation Statistical blending of weather data sets
CN107491341A (en) * 2017-08-31 2017-12-19 福州大学 A kind of virtual machine distribution method based on particle group optimizing
US20210073286A1 (en) * 2019-09-06 2021-03-11 Digital Asset Capital, Inc. Multigraph verification

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8972405B1 (en) * 2012-07-26 2015-03-03 Emc Corporation Storage resource management information modeling in a cloud processing environment
CN104484715A (en) * 2014-11-28 2015-04-01 江苏大学 Neural network and particle swarm optimization algorithm-based building energy consumption predicting method
US20170329048A1 (en) * 2016-05-12 2017-11-16 The Climate Corporation Statistical blending of weather data sets
CN106657238A (en) * 2016-10-13 2017-05-10 内蒙古工业大学 Interactive-workload-oriented data center coarse-grained dynamic server reservation algorithm
CN107491341A (en) * 2017-08-31 2017-12-19 福州大学 A kind of virtual machine distribution method based on particle group optimizing
US20210073286A1 (en) * 2019-09-06 2021-03-11 Digital Asset Capital, Inc. Multigraph verification

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
SASAN GHAREHPASHA等: "Power efficient virtual machine placement in cloud data centers", 《CLUSTER COMPUTING》 *
SHAYEM SALEH ALRESHEEDI等: "Improved multiobjective salp swarm", 《HUMAN-CENTRIC COMPUTING AND INFORMATION SCIENCES》 *
刘小娟等: "一种基于差分进化的正弦余弦算法", 《工程科学学报》 *
吕立国等: "结合质心思想和柯西变异策略的粒子群优化算法", 《计算机应用》 *
李生: "云计算资源负载预测关键技术及算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
王茂宇等: "一种基于改进粒子群算法的虚拟机放置方法", 《数据通信》 *
裴养等: "基于粒子群优化算法的虚拟机放置策略", 《计算机工程》 *

Also Published As

Publication number Publication date
CN113296893B (en) 2022-09-02

Similar Documents

Publication Publication Date Title
CN107491341B (en) A kind of virtual machine distribution method based on particle group optimizing
Liu et al. Deadline‐constrained coevolutionary genetic algorithm for scientific workflow scheduling in cloud computing
CN112286677B (en) Resource-constrained edge cloud-oriented Internet of things application optimization deployment method
CN110851272A (en) Cloud task scheduling method based on phagocytic particle swarm genetic hybrid algorithm
Wei-guo et al. Research on kubernetes' resource scheduling scheme
CN108469983A (en) A kind of virtual machine deployment method based on particle cluster algorithm under cloud environment
Gao et al. An energy-aware ant colony algorithm for network-aware virtual machine placement in cloud computing
CN109447264B (en) Virtual machine placement genetic optimization method based on VHAM-R model in cloud computing environment
Supreeth et al. Hybrid genetic algorithm and modified-particle swarm optimization algorithm (GA-MPSO) for predicting scheduling virtual machines in educational cloud platforms
CN110795208A (en) Mobile cloud computing self-adaptive virtual machine scheduling method based on improved particle swarm
CN111966495B (en) Data processing method and device
Luo et al. An improved particle swarm optimization algorithm based on adaptive weight for task scheduling in cloud computing
CN110362379A (en) Based on the dispatching method of virtual machine for improving ant group algorithm
CN111176784A (en) Virtual machine integration method based on extreme learning machine and ant colony system
CN113037546A (en) Security virtual network mapping method based on deep reinforcement learning
CN113360245A (en) Internet of things equipment task downloading method based on mobile cloud computing deep reinforcement learning
CN113296893B (en) Cloud platform low-resource-loss virtual machine placement method based on hybrid sine and cosine particle swarm optimization algorithm
Liu et al. Emergency scheduling optimization simulation of cloud computing platform network public resources
CN115617526A (en) Cloud data center energy-saving method based on cloud data center construction and virtual machine integration
CN108073449B (en) Dynamic virtual machine placement method
CN116321189A (en) Server deployment method based on deep reinforcement learning in edge-oriented computing
Tan et al. An energy-aware virtual machine placement algorithm in cloud data center
CN113157431B (en) Computing task copy distribution method for edge network application environment
CN114980216A (en) Dependent task unloading system and method based on mobile edge calculation
CN115016889A (en) Virtual machine optimization scheduling method for cloud computing

Legal Events

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