Professional Documents
Culture Documents
Multi-Objective Task
Multi-Objective Task
H O S T E D BY
Alexandria University
Civil Aviation University of China, Basic Experiment Center, TianJin 300300, China
KEYWORDS Abstract A cloud computing multi-objective task scheduling optimization based on fuzzy self-
Fuzzy self-defense algorithm; defense algorithm is proposed. Select the shortest time, the degree of resource load balance and
Cloud computing; the cost of multi-objective task completion as the goal of cloud computing multi-objective task
Multi-objective task; scheduling, establish a mathematical model to measure the effect of multi-objective task scheduling,
Scheduling optimization and construct the objective function of cloud computing multi-objective task scheduling. The exper-
imental results show that the cloud computing multi-objective task scheduling optimization method
based on fuzzy self-defense algorithm can improve the performance of multi-objective maximum
completion time, deadline violation rate and virtual machine resource utilization.
Ó 2021 THE AUTHORS. Published by Elsevier BV on behalf of Faculty of Engineering, Alexandria
University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).
and the load balance function, and also gives the initial pher- tations to cloud assets for preparing in a circulated climate is a
omone. Then, it improves the heuristic function and phero- well‐known NP‐hard issue. To determine this issue, an energy‐
mone updating method of ant colony optimization efficient task‐scheduling calculation is proposed in [26] depen-
algorithm. It obtains the objective constraint function through dent on best‐worst (BWM) and the Technique for Order Pref-
the ant colony optimization algorithm, further obtaining the erence by Similarity to Ideal Solution (TOPSIS) procedure.
global optimal solution. Then, it simulates the cloudsim, and Initial, a decision‐making bunch distinguishes the assessment
compares Min-Min algorithm and ACO algorithm. As demon- rules. From that point forward, a BWM cycle is applied to
strated in experiment, such algorithm is better than the above allot the significance loads for every basis, in light of the fact
two algorithms in terms of either the task execution time, or that the chose standards have shifted significantly. At that
cost, or load balance; Literature[5] proposes a new Q- point, TOPSIS utilizes these weighted models as contributions
learning-based multi-objective optimization task scheduling to assess and quantify the performance of every other option.
algorithm. At the task scheduling stage, the better task In Ref. [27], an task planning solution is proposed considering
sequence may be obtained through the Q-learning algorithm. the game hypothesis intended for energy management in cloud
During the virtual machine distribution process, at the begin- computing. The game hypothesis can more readily facilitate
ning, the linear weighting method is used to compute the node the dispersion of errands and the conveyance of energy. Like-
and completion time, optimizing the multi-objective problem. wise, The task scheduling model for processing hubs by setting
The objective function is established for the shorter completion up a numerical model is proposed in [27] to manage huge
time and smaller total cost. As demonstrated in experiment, information. Tests show that the calculation can improve
qmts algorithm sorts the tasks based on the Q-learning, which energy planning in distributed computing.
may obtain shorter sorting time than heft. In addition, after Based on the above background, maximum completion
the optimization of multi-objective scheduling strategy, it time, high deadline violation rate and low utilization of virtual
may reduce the total cost and realize the task of makespan. machine resources play an important role in the process of
Ref. [21] introduced a novel antlion algorithm for taking care scheduling optimization for cloud computing multi-objective
of multi-objective task scheduling issues in cloud computing tasks. However, so far, these parameters have not been prop-
conditions. In the proposed technique, the multi-target nature erly satisfied in a comprehensive solution. Therefore, in order
of the issue gets from the need to at the same time limit make- to improve the performance of cloud computing multi-
span while boosting asset use. The proposed algorithm was objective task in terms of maximum completion time, deadline
upgraded by using elite-based differential advancement as a violation rate and virtual machine resource utilization, this
neighborhood search strategy to improve its misuse capacity paper applies the fuzzy self-defense algorithm to the multi-
and to try not to get caught in nearby optima. In another study objective task scheduling optimization in cloud computing,
in [22], a trust-mindful scheduling methodology for distributed designs the cloud computing multi-objective task scheduling
computing is proposed. This method considers Virtual Machi- optimization algorithm, and verifies the effectiveness of such
nes’ trust level calculation and undertakings need assurance. In method via the simulation experiment.
Ref. [23], a Q-learning-based task scheduling system for
energy-proficient cloud computing is proposed to handle the 2. Design the cloud computing multi-objective task scheduling
issue of energy utilization in both scheduling and assignment optimization method
planning. It was shown that the applying of M/M/S lining
model can yield a more limited errand reaction time, which 2.1. Select the goal of cloud computing multi-objective task
prompts improved energy effectiveness. In Ref. [24], In this scheduling
article, the metaheuristics whale optimization algorithm
(WOA) is utilized for cloud task planning with a multiobjective
For the scheduling optimization of cloud computing multi-
streamlining model, targeting improving the presentation of a
objective task, select the shortest time that customers need to
cloud framework with given registering assets. On that pre-
wait, the degree of resource load balance and the cost of
mise, a high-level methodology called Improved WOA is pro-
multi-objective task completion as the goal of cloud computing
posed for Cloud task scheduling to additionally improve the
multi-objective task scheduling[6].
searchability of the WOA-based technique. This strategy can
likewise accomplish better execution on framework asset
(1) Shortest time that customers need to wait
usage, within the sight of both little and enormous scope
undertakings. In another study, the adaptable assignment
The computing equation for the shortest time that cus-
planning issue in a distributed computing framework is con-
tomers need to wait is as follows:
centrated by a hybrid discrete artificial bee colony (ABC) cal-
culation in Ref. [25], where the issue is first demonstrated as a sumðVMj Þ
m X
mixture flowshop scheduling issue. An improved variant of the WFi ¼ max timeij ð1Þ
j¼1
versatile irritation structure is implanted in the proposed calcu- i¼1
lation to adjust the misuse and investigation capacity. A deter- In which, timeij refers to the execution time of cloud com-
mination and refreshed methodology are applied to upgrade puting multi-objective task Ti on the virtual machine VMj ;
the misuse interaction. To additionally improve the abuse
sum VMj refers to the total number of multi-objective tasks
capacities, a profound misuse administrator is planned. An
assigned on the virtual machine VMj . The processing time of
improved scout honey bee utilized with various neighborhoods
the multi-objective task is the ratio of task instruction length
scan strategies for the best food source or the unwanted
to execution speed of virtual machine VMj , which is repre-
arrangement is planned and can expand the intermingling
sented as follows:
capacity of the proposed calculation. Planning of client solici-
Multi-objective task scheduling optimization in cloud computing 5605
TAi For the operating cost ctotal required for processing the
timeij ¼ ð2Þ
VAj cloud computing multi-objective task set, assume that the cost
ci for multi-objective task Ti on the virtual machine within the
(2) Degree of resource load balance unit time is ci , then ctotal may be expressed as follows:
The computing equation for the degree of resource load X
p
balance on the virtual machine is as follows: ctotal ¼ ci ðte ðTi Þ ts ðTi ÞÞ ð7Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi i¼1
P
m
ðVMFj VMFj Þ
2
If the computing performance of multi-objective task on
j¼1
m
the virtual machine is good enough, the value of ci in equation
-¼ ð3Þ (7) is larger[9].
n
However, for the load balance of cloud computing multi-
In which, VMFj refers to the total execution time of the
objective task on the virtual machine, although more multi-
multi-objective task on the virtual machine; VMFj refers to objective tasks are assigned on the virtual machine, the load
the mean execution time of the multi-objective task on the vir- balance effects may not be better[10]. Define the expected exe-
tual machine. cution time of all multi-objective tasks on the virtual machine
as VLj , which is expressed as follows:
(3) Cost of multi-objective task completion !
Xu1
amountðTi Þ
VLj ¼ ð8Þ
The computing equation for the cost of multi-objective task ability Vj
i¼0
completion is as follows:
In which, u refers to the number of multi-objective tasks
X
m
costVMj ¼ sum VMj Vcpuj þ Vramj þ Vbwj ð4Þ assigned on the virtual machine. Define the expected mean exe-
j¼1 cution time of all multi-objective tasks on the virtual machine
as VL, and the computing equation is as shown below:
The equation (4) fully reflects the cost of cloud computing
Pw1
multi-objective task completion on the virtual machine, which j¼0 VLj
is closely connected with central processing unit (CPU), band- VL ¼ ð9Þ
p
width performance and memory of the virtual machine.
Starting from three objectives, namely, shortest time that In which, w refers to the number of multi-objective tasks
customers need to wait, degree of resource load balance and assigned on the virtual machine; p refers to the number of
cost of multi-objective task completion, measure the effect of multi-objective tasks in the cloud computing multi-objective
multi-objective task scheduling and establish the following task set.
mathematical model: Define the load balance variance of multi-objective task
scheduling as BL, which is expressed as follows:
minWF; min-; mincost sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pm ð5Þ P VLj 2
j¼1 aij ¼ 1 aij 2 ½0; 1 i ¼ 1; 2; ; n w1
s:t:
u
VL
j¼0
Compute the time of multi-objective task completion by BL ¼ ð10Þ
utilizing the goal of shortest time that customers need to wait. w
In combination with the computing of degree of resource load The smaller the value of BL is, the better the load balance
balance and cost of multi-objective task completion, establish of multi-objective task in the data center is during the process
the mathematical model that measures the effect of multi- of cloud computing multi-objective task scheduling.
objective task scheduling, and select the goal of cloud comput- For the cloud computing users, it is required to spend the
ing multi-objective task scheduling. minimum cost within the shortest time. Therefore, take the
minimum values[11] for ttotal and ctotal . Considering that cloud
2.2. Construct the objective function of cloud computing multi- computing belongs to a commercial service pattern, during the
objective task design of Pareto domination relationship, the time and cost
related to user interests are higher than load balance[12] in
A certain constraint relationship[7] exists among optimized terms of priority. As a result, construct the objective
objectives of multi-objective task in cloud computing. Hence, function of cloud computing multi-objective task, as shown
based on understanding the impact of various steps upon such in equation (11):
8
relationship during the process of cloud computing multi- > minðttotal Þ ¼ minðte ðTlast Þ ts ðTfirst ÞÞ
>
> p
objective task scheduling, the optimized multi-objective func- >
> P
>
< minðctotal Þ ¼ min ci ðte ðTi Þ ts ðTi ÞÞ
tion[8] of cloud computing multi-objective task scheduling
may be determined.
i¼1
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! ð11Þ
>
> 2
>
> P
w1
Assume that, in the cloud computing multi-objective task >
> VLj
set, Tfirst refers to the first processed multi-objective task and : minðBLÞ ¼ min u
VL =w
j¼0
Tlast refers to the last processed multi-objective task. Then,
the equation (6) may be utilized to compute the processing Define the number of multi-objective tasks in the multi-
time of cloud computing multi-objective task set, namely: objective task set through computing the processing time of
cloud computing multi-objective task set. According to the
ttotal ¼ te ðTlast Þ ts ðTfirst Þ ð6Þ expected execution time of all multi-objective tasks on the vir-
5606 X. Guo
deadline as the evaluation goal. As a result, it is impossible to [3] S.Z. Mirjalili, S. Mirjalili, S. Saremi, et al, Grasshopper
achieve the load balance among virtual machines and the fluc- optimization algorithm for multi-objective optimization
tuation of resource utilization rate is greater. problems, Appl. Intell. 48 (4) (2018) 805–820.
[4] Jun-wei Ge, Qiang Guo, Yi-qiu Fang, A Multi-objective
Optimization Algorithm for Cloud Computing Task
4. Conclusion Scheduling Based on Improved Ant Colony Algorithm,
Microelectron. Comput. 34 (11) (2017) 63–67.
A cloud computing multi-objective task scheduling optimiza- [5] Zhao Tong, Xiao-mei Deng, Hong-jian Chen, et al, Multi-
tion based on fuzzy self-defense algorithm is proposed in this objective Task Scheduling Algorithm Based on Reinforcement
paper. Based on the selection of the goal of cloud computing Learning in Cloud Envi-ronments, J. Chinese Comput. Syst. 41
multi-objective task scheduling, an objective function of cloud (02) (2020) 285–290.
[6] D. Vazquez, M.J. Fernandez-Torres, R. Ruiz-Femenia, et al,
computing multi-objective task scheduling is constructed.
MILP method for objective reduction in multi-objective
After solving the objective function with the fuzzy self- optimization, Comput. Chem. Eng. 108 (4) (2018) 382–394.
defense algorithm by search, the global optimal solution of [7] H. Dkhil, A. Yassine, H. Chabchoub, Multi-objective
the objective function is obtained, realizing the cloud comput- optimization of the integrated problem of location assignment
ing multi-objective task scheduling optimization. As shown and straddle carrier scheduling in maritime container terminal at
from the results, this method has better performance in terms import, J. Oper. Res. Soc. 69 (2) (2018) 247–269.
of maximum completion time, deadline violation rate, and uti- [8] A. Cheaitou, P. Cariou, Greening of maritime transportation: a
lization of virtual machine resources. Therefore, it has better multi-objective optimization approach, Ann. Oper. Res. 273 (1–
applicability. The proposed algorithm plans the three objec- 2) (2019) 501–525.
tives, namely, shortest time that customers need to wait, degree [9] Dhiraj, Kumar, Suhasini, et al. Abrasive waterjet machining of
Ti/CFRP/Ti laminate and multi-objective optimization of the
of resource load balance and cost of multi-objective task com-
process parameters using response surface methodology. J.
pletion as a single objective evaluation function, and takes it as Compos. Mater. 54(13) (2019) 1741-1759.
the scheduling goal. As a result, it obtains a better scheduling [10] Haifeng, Jin, Mohammad, et al. Multiobjective Optimization for
effect, with the shortest maximum scheduling completion time. Scaffolding Space Planning in Industrial Piping Construction
Furthermore, as the number of cloud computing multi- Using Model-Based Simulation Programming. J. Comput. Civ.
objective tasks increases, a low deadline violation rate is main- Eng., 2020, 34(1):107-112.
tained for proposed algorithm so that the deadline violation [11] X. Blasco, G. Reynoso-Meza, E.A. Sánchez-Pérez, et al,
rate is less than other methods, namely, about 5%. In addition, Computing Optimal Distances to Pareto Sets of Multi-
under a certain number of multi-objective tasks, the results Objective Optimization Problems in Asymmetric Normed
obtained with the proposed algorithm are relatively balanced Lattices, Acta Appl. Math. 3 (2018) 1–19.
[12] M. Abd Elaziz, S. Xiong, K.P.N. Jayasena, et al, Task
with regard to the task scheduling, which indicates that the
scheduling in cloud computing based on hybrid moth search
resource utilization rate on the virtual machine is relatively algorithm and differential evolution, Knowl.-Based Syst. 169 (1)
adequate, namely, always above 95%. (2019) 39–52.
[13] A. Asghari, M.K. Sohrabi, F. Yaghmaee, Task scheduling,
Declaration of Competing Interest resource provisioning, and load balancing on scientific
workflows using parallel SARSA reinforcement learning
The authors declare that they have no known competing agents and genetic algorithm, J. Supercomput. (2020) 1–29.
financial interests or personal relationships that could have [14] Rey Antón, F.D. Igual, Prieto-Matı́as Manuel, Variable intra-
appeared to influence the work reported in this paper. task threading for power-constrained performance and energy
optimization in DAG scheduling, J. Supercomput. 75 (2) (2019)
104–116.
Acknowledgment [15] E. Barboza, C. Bastos-Filho, J.F.M. Filho, Adaptive Control of
Optical Amplifier Operating Point using VOA and Multi-
This paper is funded by the central university basic scientific Objective Optimization, J. Lightwave Technol. 37 (16) (2019)
research business fee project of Civil Aviation University of 3994–4000.
[16] S.N. Naserabad, A. Mehrpanahi, G. Ahmadi, Multi-Objective
China. The project name is the application research of desktop
Optimization of HRSG Configurations on the Steam Power
cloud in university computer laboratory, and the project num-
Plant Repowering Specifications, Energy 159 (15) (2018) 277–
ber is 3122018S009. 293.
[17] A. Stupar, T. Mcrae, N. Vukadinovic, et al, Multi-Objective
Availability of data and material Optimization of Multi-Level DC-DC Converters using
Geometric Programming, IEEE Trans. Power Electron. 34
(12) (2019) 11912–11939.
All the data can be requested from the corresponding author
[18] Dkhil, Hamdi, Yassine, et al. Multi-objective optimization of
upon reasonable request.
the integrated problem of location assignment and straddle
carrier scheduling in maritime container terminal at import. J.
References Oper. Res. Soc. 2018, 69(2):247-269.
[19] A M A A, B M T R. Multi-objective optimization of reliability–
[1] M.A. Elaziz, I. Attiya, An improved Henry gas solubility redundancy allocation problem with cold-standby strategy using
optimization algorithm for task scheduling in cloud computing, NSGA-II. Reliability Engineering & System Safety, 2018,
Artif. Intell. Rev. (2020) 1–39. 172:225-238.
[2] D.V. Kunhambu, R.V.S. Balan, An Efficient Multi-Objective [20] M. Sadeghi, A.S. Mehr, M. Zar, et al, Multi-objective
Particle Swarm Optimization Based Ranking System for Cloud optimization of a novel syngas fed SOFC power plant using a
Service Selection, IET Commun. 13 (3) (2018) 297–304. downdraft gasifier, Energy 148 (1) (2018) 16–31.
Multi-objective task scheduling optimization in cloud computing 5609
[21] Laith Abualigah, Ali Diabat, A novel hybrid antlion for task scheduling in cloud computing systems, IEEE Syst. J. 14
optimization algorithm for multi-objective task scheduling (3) (2020) 3117–3128.
problems in cloud computing environments, Cluster [25] Jun-qing Li, Yun-qi Han, A hybrid multi-objective artificial bee
Computing (2020) 1–19. colony algorithm for flexible task scheduling problems in cloud
[22] Rjoub Gaith, Jamal Bentahar, Omar Abdel Wahab, computing system, Cluster Computing 23 (4) (2020) 2483–2499.
BigTrustScheduling: Trust-aware big data task scheduling [26] Reihaneh Khorsand, Mohammadreza Ramezanpour, An
approach in cloud computing environments, Future energy-efficient task-scheduling algorithm based on a multi-
Generation Computer Syst. 110 (2020) 1079–1097. criteria decision-making method in cloud computing, Int. J.
[23] Ding Ding, Xiaocong Fan, Yihuan Zhao, Kaixuan Kang, Qian Commun Syst 33 (9) (2020) e4379.
Yin, Jing Zeng, Q-learning based dynamic task scheduling for [27] Jiachen Yang, Bin Jiang, Zhihan Lv, Kim-Kwang Raymond
energy-efficient cloud computing, Future Generation Computer Choo, A task scheduling algorithm considering game theory
Systems 108 (2020) 361–371. designed for energy management in cloud computing, Future
[24] Xuan Chen, Long Cheng, Cong Liu, Qingzhi Liu, Jinwei Liu, Generation computer systems 105 (2020) 985–992.
Ying Mao, John Murphy, A woa-based optimization approach