CN111679907B - Cloud computing task scheduling method - Google Patents

Abstract

The invention discloses a cloud computing task scheduling method, which comprises the steps of acquiring a merged cell population by adopting a reverse learning method, increasing the diversity of the cell population to ensure that the population can be effectively and uniformly distributed in an interval range, then carrying out iterative computation on the acquired merged cell population by utilizing a cuttlefish algorithm, searching cells in a new merged cell population space, increasing the probability of searching an optimal solution by the cells until the maximum iteration times are met to obtain an optimal scheduling scheme, and then distributing M sub-tasks which are independent and non-separable by utilizing the optimal scheduling scheme, thereby effectively reducing the defect of lacking the diversity of the cell population caused by the conventional method, increasing the diversity of the cell population, balancing the load and reducing the waiting time.

Description

Cloud computing task scheduling method

Technical Field

The invention relates to cloud computing low-energy-consumption task scheduling, in particular to a cloud computing task scheduling method.

Background

With the explosion development of the internet and the arrival of a big data era, the traditional network technology mode cannot meet the urgent requirements of users on rapid processing, storage, access and the like of mass data, and the cloud computing technology is produced. Cloud computing is in fact the development of grid computing, parallel computing, distributed computing, utility computing, or in other words the commercial implementation of these computing science concepts. A large number of computing resources, storage resources, service resources and the like are connected through a network to form a super resource pool, and then the resources are uniformly scheduled and managed according to the requirements of users. Cloud computing is used as a novel commercial service computing mode, IT service resources are provided by cloud service providers, and cloud users only need to pay according to needs without considering the building, management and maintenance of hardware infrastructure. The advent of the cloud computing era has brought about more opportunities and greater challenges to people.

One of the key technologies in cloud computing is task scheduling, which schedules M tasks to N virtual resource nodes. The rationality of scheduling not only directly influences the experience quality of a user on cloud services, but also relates to the commercial benefits of a cloud platform provider, determines the running performance of the whole cloud system, and is an important factor for healthy development of cloud computing. At present, more and more cloud computing task scheduling starts to use a heuristic method. The cuttlefish optimization algorithm is a novel heuristic optimization algorithm which was originally proposed by Adel Sabry Eesa in 2013. It is inspired by the color change mechanism of cuttlefish cells to find the optimal solution in the optimization problem. The conventional method for initializing the seed group uses random initialization, so that the uniform distribution of the initialization of the seed group in an interval range cannot be ensured, the optimization speed is low, the final optimization result is not ideal, and the solving precision is not high.

Disclosure of Invention

The invention aims to provide a cloud computing task scheduling method to solve the problems that the prior art cannot ensure that population initialization is uniformly distributed in an interval range, so that the optimization speed is slow and the final optimization result is not ideal.

In order to achieve the purpose, the invention adopts the following technical scheme:

a cloud computing task scheduling method comprises the following steps:

step 1), dividing a task to be processed into M sub-tasks which are independent and can not be re-divided;

step 2), acquiring a merged cell population by adopting a reverse learning method, and then performing iterative computation on the acquired merged cell population by utilizing a cuttlefish algorithm until the maximum iteration times are met or a cost function meets the condition, so that an optimal scheduling scheme can be obtained;

and 3) scheduling the M subtasks according to the optimal scheduling scheme obtained in the step 2).

Further, the step 2) of obtaining the initialization cell population by using a reverse learning method specifically comprises the following steps:

step 2.1), randomly generating an initial cell population x with the size of NP;

and 2.2) generating a reverse cell population by adopting a reverse learning method according to the initialized cell population x generated randomly, and combining the initialized cell population x and the reverse cell population to form a combined cell population with the size of 2 NP.

Further, the diversity of cell population is increased by using a reverse learning strategy, and x = (x) ₁ ,x ₂ ,x ₃ ,....x _D ) One point (feasible solution) in the D-dimensional space; x is the number of _j Belongs to [ a _j ,b _j ]Wherein j belongs to (1,2,3.. D); [ a ] A _j ,b _j ]Is x _j Is bounded by an upper and lower bound, i.e. x _j The search range of (2); generation of a reverse cell population using a reverse learning approach

Comprises the following steps: />

The solution space of the initialization cell population x and the solution space of the reverse cell population are combined into one to form a combined cell population.

Further, the cost function is used for calculating the cost of all cells in the combined cell population, and the cell with the minimum cost is taken as the global optimal solution S ^best Sorting the costs of all cells in the combined cell population from small to large, and taking the first NP cells to form a search cell population;

randomly dividing the search cell population into G ⁽¹⁾ 、G ⁽²⁾ 、G ⁽³⁾ 、G ⁽⁴⁾ And 4 groups of collaborative search are carried out, each group is iteratively solved corresponding to different reflection and visibility processes until the maximum iteration times gmax is reached or the cost function meets the condition, and the optimal solution can be obtained.

Further, for G ⁽¹⁾ The two processes of cell in the group, which are visible after reflection, are as follows: reflection _j ＝R*G _i,j ⁽¹⁾ ，

R＝random()*(r ₁ -r ₂ )+r ₂ ，V＝random()*(v ₁ -v ₂ )+v ₂ ；

In the formula: j = {1,2,3.. D } represents a D-dimensional problem; r is the degree of reflection; v is the degree of visibility;

is G ⁽¹⁾ (ii) a solution in dimension j for the ith cell in the group; random () generates a random number on (0,1); r is a radical of hydrogen ₁ 、r ₂ Constants that limit the cell expansion interval, respectively; v. of ₁ 、v ₂ Constants that limit the degree of visibility, respectively; at G ⁽¹⁾ And medium V is set to 1.

Further, for G ⁽²⁾ The cells in the group, both in reflection and visibility, are represented as follows:

V＝random()*(v ₁ -v ₂ )+v ₂ ；

in the formula

Is G ⁽²⁾ Solution in dimension j of the ith cell in the group, at G ⁽²⁾ Where R is 1,v ₁ 、v ₂ Respectively, constants that limit the degree of visibility.

Further, for G ⁽³⁾ The cells in the group, both reflex and visibility processes are represented as follows: reflection _j ＝R*G _i,j ⁽³⁾ ，

V＝random()*(v ₁ -v ₂ )+v ₂ ；

In the formula

Is->

Average value of (1) in G ⁽³⁾ Where R is 1,v ₁ 、v ₂ Respectively, constants that limit the degree of visibility.

Further, for G ⁽⁴⁾ Cells in the group randomly seek new solutions within the search range; namely:

in the formula (II)>

Is G ⁽⁴⁾ Randomly searching the new solution on the j dimension by the cells in the group; />

Respectively, the upper and lower bounds of the search range in the jth dimension.

Further, the optimal scheduling scheme output in the step 2) is used as a cloud computing task scheduling strategy, and task scheduling is executed to schedule the M tasks to the N virtual resource nodes.

Compared with the prior art, the invention has the following beneficial technical effects:

the invention relates to a cloud computing task scheduling method, which adopts a reverse learning method to obtain a merged cell population, increases the diversity of the cell population so as to ensure that the population can be effectively and uniformly distributed in an interval range, then utilizes a cuttlefish algorithm to carry out iterative computation on the obtained merged cell population so as to enable cells to be searched in a new merged cell population space, increases the probability of searching an optimal solution by the cells until the maximum iteration times are met to obtain an optimal scheduling scheme, and then utilizes the optimal scheduling scheme to distribute M sub-tasks which are independent and non-separable, thereby effectively reducing the defect of lacking the diversity of the cell population caused by the existing method, increasing the diversity of the cell population, balancing the load and reducing the waiting time.

Furthermore, the population number is divided into four groups by adopting a cuttlefish algorithm, each group works independently to share the respective optimal solution, the searching speed is high, and the convergence time is short.

Drawings

FIG. 1 is a flow chart of an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are merely illustrative of the basic idea of the invention and do not limit the invention. Other advantages and effects of the present invention will be apparent to those skilled in the art from the description of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention.

As shown in fig. 1, a cloud computing task scheduling method includes the following steps:

step 1), dividing a task to be processed into M sub-tasks which are independent and can not be divided;

step 2), acquiring a merged cell population by adopting a reverse learning method, and then performing iterative computation on the acquired merged cell population by utilizing a cuttlefish algorithm until the maximum iteration times are met to obtain an optimal scheduling scheme;

the step 2) of obtaining the initialized cell population by adopting a reverse learning method specifically comprises the following steps:

step 2.1), randomly generating an initial cell population x of size NP, with a constant limiting the cell expansion interval r ₁ ,r ₂ (ii) a Let the constant of the degree of visibility be v ₁ ,v ₂ (ii) a The maximum number of iterations gmax;

step 2.2), generating a reverse cell population by adopting a reverse learning method according to the initialized cell population x generated randomly, and combining the initialized cell population x and the reverse cell population to form a combined cell population with the size of 2 NP; increasing the diversity of the cell population;

specifically, the diversity of cell population is increased by using a reverse learning strategy, and x = (x) ₁ ,x ₂ ,x ₃ ,....x _D ) One point (feasible solution) in the D-dimensional space; x is a radical of a fluorine atom _j Belongs to [ a _j ,b _j ]Wherein j belongs to (1,2,3.. D); [ a ] A _j ,b _j ]Is x _j Is at an upper and lower bound, i.e. x _j The search range of (2); reverse cell population generation using reverse learning method

Comprises the following steps: />

Combining the solution space of the initialized cell population x and the solution space of the reverse cell population into one to form a merged cell population, and searching the solution space of the merged cell population by the cells to increase the probability of searching the optimal solution by the cells;

calculating the cost of all cells in the merged cell population by using a cost function, and taking the cell with the minimum cost as a global optimal solution S ^best Sorting the costs of all cells in the combined cell population from small to large, and taking the first NP cells to form a search cell population;

randomly dividing the search cell population into G ⁽¹⁾ 、G ⁽²⁾ 、G ⁽³⁾ 、G ⁽⁴⁾ 4 groups of collaborative search are carried out, each group is iteratively solved corresponding to different reflection and visibility processes until the maximum iteration times gmax is reached or a cost function meets a condition, and then an optimal solution can be obtained;

the cuttlefish algorithm summarizes the discoloring behavior of cuttlefish into two processes of reflection (reflection) and visibility (visibility) of cells, wherein the reflection represents a mechanism that the cells reflect incident light, and the visibility represents the visibility of the cuttlefish matched with the surrounding environment; the cuttlefish algorithm uses two processes of reflection and visibility as a search strategy to find a new solution S ^new I.e. S ^new = reflection + visibility. Meanwhile, the cell population is divided into G according to 6 reflected light conditions of cuttlefish discoloration ⁽¹⁾ 、G ⁽²⁾ 、G ⁽³⁾ 、G ⁽⁴⁾ 4 sets of collaborative searches, each set corresponding to a different reflection and visibility process.

For G ⁽¹⁾ Cells in the groupThe two processes for which the reflection remains visible are as follows: reflection _j ＝R*G _i,j ⁽¹⁾ ，

R＝random()*(r ₁ -r ₂ )+r ₂ ，V＝random()*(v ₁ -v ₂ )+v ₂ ；

In the formula: j = {1,2,3.. D } represents a D-dimensional problem; r is the degree of reflection; v is the visibility;

is G ⁽¹⁾ (ii) a solution in dimension j for the ith cell in the group; random () generates a random number on (0,1); r is ₁ 、r ₂ Constants that limit the cell expansion interval, respectively; v. of ₁ 、v ₂ Constants that limit the degree of visibility, respectively; at G ⁽¹⁾ And medium V is set to 1.

For G ⁽²⁾ The cells in the group, both in reflection and visibility, are represented as follows:

V＝random()*(v ₁ -v ₂ )+v ₂ ；

in the formula

Is G ⁽²⁾ Solution in dimension j of the ith cell in the group, at G ⁽²⁾ Where R is 1,v ₁ 、v ₂ Respectively, constants that limit the degree of visibility.

For G ⁽³⁾ The cells in the group, both in reflection and visibility, are represented as follows: reflection _j ＝R*G _i,j ⁽³⁾ ，

V＝random()*(v ₁ -v ₂ )+v ₂ ；

In the formula

Is->

Average value of (1) in G ⁽³⁾ Where R is 1,v ₁ 、v ₂ Constants that limit the degree of visibility, respectively;

for G ⁽⁴⁾ Cells in the group randomly seek new solutions within the search range; namely:

in combination with>

Is G ⁽⁴⁾ Randomly searching the new solution on the j dimension by the cells in the group; />

Respectively, the upper and lower bounds of the search range in the jth dimension.

Repeat pair G constantly ⁽¹⁾ 、G ⁽²⁾ 、G ⁽³⁾ 、G ⁽⁴⁾ And performing iterative solution to obtain the cells with the minimum cell cost calculated by each group, if the cell cost of the obtained cells is less than the current cell cost, replacing the current scheduling scheme with the scheduling scheme corresponding to the group in which the cells are located until the maximum iteration number gmax is reached, and exiting and outputting the optimal scheduling scheme.

Step 3), scheduling the M subtasks according to the optimal scheduling scheme obtained in the step 2): and (3) using the optimal scheduling scheme output in the step (2) as a cloud computing task scheduling strategy, and executing task scheduling to schedule the M tasks to the N virtual resource nodes.

Claims (2)

1. A cloud computing task scheduling method is characterized by comprising the following steps:

step 1), dividing a task to be processed into M sub-tasks which are independent and can not be divided;

step 2), acquiring a merged cell population by adopting a reverse learning method, and then performing iterative computation on the acquired merged cell population by utilizing a cuttlefish algorithm until the maximum iteration times are met or a cost function meets the condition, so that an optimal scheduling scheme can be obtained;

the method for acquiring the initialized cell population by adopting the reverse learning method specifically comprises the following steps:

step 2.1), randomly generating an initial cell population x with the size of NP;

step 2.2), generating a reverse cell population by adopting a reverse learning method according to the initialized cell population x generated randomly, and combining the initialized cell population x and the reverse cell population to form a combined cell population with the size of 2 NP;

increasing the diversity of cell populations using a reverse learning strategy, let x = (x) ₁ ,x ₂ ,x ₃ ,....x _D ) Is a point of the D-dimensional space; x is the number of _j Belongs to [ a _j ,b _j ]Wherein j belongs to (1,2,3.. D); [ a ] A _j ,b _j ]Is x _j Is at an upper and lower bound, i.e. x _j The search range of (2); reverse cell population generation using reverse learning method

Comprises the following steps: />

Combining the solution space of the initialized cell population x and the solution space of the reverse cell population into one to form a combined cell population;

calculating the cost of all cells in the merged cell population by using a cost function, and taking the cell with the minimum cost as a global optimal solution S ^best Sorting the costs of all cells in the combined cell population from small to large, and taking the first NP cells to form a search cell population;

randomly dividing the search cell population into G ⁽¹⁾ 、G ⁽²⁾ 、G ⁽³⁾ 、G ⁽⁴⁾ 4 groups of collaborative search are carried out, each group is iteratively solved corresponding to different reflection and visibility processes until the maximum iteration times gmax is reached or a cost function meets a condition, and then an optimal solution can be obtained;

for G ⁽¹⁾ The two processes of cell in the group, which are visible after reflection, are as follows: reflection _j ＝R*G _i,j ⁽¹⁾ ，

R＝random()*(r ₁ -r ₂ )+r ₂ ，V＝random()*(v ₁ -v ₂ )+v ₂ ；

In the formula: j = {1,2,3.. D } represents a D-dimensional problem; r is the degree of reflection; v is the degree of visibility;

is G ⁽¹⁾ (ii) a solution in dimension j for the ith cell in the group; random () generates a random number on (0,1); r is ₁ 、r ₂ Constants that limit the cell expansion interval, respectively; v. of ₁ 、v ₂ Constants that limit the degree of visibility, respectively; at G ⁽¹⁾ Middle V is set to 1;

for G ⁽²⁾ The cells in the group, both in reflection and visibility, are represented as follows:

V＝random()*(v ₁ -v ₂ )+v ₂ ；

in the formula

Is G ⁽²⁾ Solution in dimension j of the ith cell in the group, at G ⁽²⁾ Wherein R is 1,v ₁ 、v ₂ Constants that limit the degree of visibility, respectively;

for G ⁽³⁾ The cells in the group, both in reflection and visibility, are represented as follows: reflection _j ＝R*G _i,j ⁽³⁾ ，

V＝random()*(v ₁ -v ₂ )+v ₂ ；

In the formula

Is->

Average value of (1) in G ⁽³⁾ Where R is 1,v ₁ 、v ₂ Constants that limit the degree of visibility, respectively;

for G ⁽⁴⁾ Cells in the group randomly seek new solutions within the search range; namely:

in the formula (II)>

Is G ⁽⁴⁾ Randomly searching the new solution on the j dimension by the cells in the group; />

Respectively the upper and lower bounds of the search range on the jth dimension;

and 3) scheduling the M subtasks according to the optimal scheduling scheme obtained in the step 2).

2. The cloud computing task scheduling method according to claim 1, wherein the optimal scheduling scheme output in step 2) is used as a cloud computing task scheduling policy, and task scheduling is performed to schedule the M tasks to the N virtual resource nodes.

Images