Tasks Scheduling in Cloud environment using PSO-BATS

doi:10.21203/rs.3.rs-945515/v1

Cloud computing plays an important role in the IT industry to deliver scalable resources as a service. Cloud users can virtually access resources over the internet. One of the most important factor to increase the performance and efficiency of the cloud server by maximizing resource utilization as task scheduling. The performance of the cloud mainly depends on resource management and optimally job scheduling. Task scheduling is a technique that defines the choosing of most appropriate resources for the execution of tasks which means the distribution of tasks on available resources by taking some parameters. The main advantage of this scheduling is to maximize the performance and minimize the time loss. There are various researchers are examined numerous scheduling methods to achieve QoS (Quality of Service) and to reduce execution time. However, it had the disadvantages of low throughput and high response time. The main objective of the proposed system is to schedule the task efficiently and to eliminate the faults in scheduling the tasks to the VMs. This research proposed the novel PSO-BATS with MLRHE (Multi-Layered Regression Host Employment) to sort out the issues of task scheduling and ease the scheduling operation with balancing of the load. The proposed efficient scheduling provides benefits to both cloud users and servers. The performance evaluation was done in terms of Makespan, cost, and PIR and found that the proposed method surpassed the other taken methods. In a comparative analysis, the proposed method proved that the best performance compared to existing methods.

Particle Swarm Optimization (PSO)

Bandwidth Aware divisible Task Scheduling (BATS)

Virtual Machines (VM)

Multi-layered regression host employment (MLRHE)

Cloud computing

Cloud computing provides a flexible way to access data from anywhere. The efficiency of the cloud has been driven in part by virtualization and it allows two or more operating systems on one PC. Virtualization assists in effective resource utilization and builds an effective system. This technology enables the service providers to provide VM (Virtual Machine) for work other than physical server machines. VM provides mobility and flexibility over easy migration that enables VMs dynamic mapping to available resources. In a cloud environment, the scheduling task is a challenging task. This technique is used to allocate certain jobs to particular resources at a particular time. The main aim of this strategy is to reduce execution time and less response time.

For the purposes of scheduling the tasks, many algorithms were found to be used predominantly throughput the literature. Some of those algorithms were: GA – Genetic Algorithm, PSO – Particle Swarm Optimization, CS – Cuckoo Search Algorithm and ACO - Ant Colony Optimization.

To face more user’s task requests, the cloud data centres not only finished the designated large tasks but also it was able to satisfy service requirements of the users [1]. In order to divide the systems’ resources and proceed with scheduling, it was important to solve the problems in cloud computing. Quality of service was mainly related in allocating the goal of scheduling. By implementing price model in scheduling algorithm, it was possible to detect multiple targets and also the service cost optimization was found to be depended on PSO-Particle Swarm Optimization. PSO algorithm was found to use dynamic cloud environment that minimizes the task completion time, in turn reduces user’s task cost.

GA-Genetic Algorithm and PSO were combined by a novel hybrid meta-heuristic method [2] for the tasks presented in directed acyclic graph that were showing the communication each inter-tasks. This proposed meta-heuristic methodology was different from the earlier hybrid meta-heuristics concepts which stand apart by improving the PSO-GA solution. PSO was mainly used to for the sake of diversifying and GA was adapted for the sake of intensifying the process in this meta-heuristic methodology. The PSO-GA consisted two standard tasks in it for scheduling the problem of linear algebra with the Elimination of Gauss-Jordan and decomposition of LU. This PSO-GA was performance wise better than other compared methods. The scheduling operations were being carried out in cloud computing environments to provision the Dynamic on-demand resource [3]. Delegation of tasks for virtual machines (VM) were done by task scheduling with the help of a nondeterministic polynomial, which increases the performance and usability, thereby, decreasing the time of response and maintains the balance of the entire system. When introducing a method of static typed task scheduling depending upon the algorithm of PSO, the designated tasks will be independent. This work performance was majorly based on the technique of load maintenance. This devised methodology of scheduling the task was validated with the round robin type of scheduling the task, enhanced scheduling of task by PSO along with the instance of load balancing. The outcomes reveal that this PSO based scheduling of the task outperforms the compared existing algorithms. This work didn’t investigate the cost deduction and ability of their proposed method to withstand failure tolerance.

Digital array radar’s task scheduling issue defines the optimal execution order for subject of the tasks to restrict the resources by using entropy dependent PSO methodology. This work [4] aims to achieve better performance in many phases. For the tough problem of N-P, an algorithm of hybridized PSO along with the optimization by integer programming models found to establish a structure for the full radar operation, which created a complete function of objectives. An enhanced PSO was introduced for comprehensively exploring the better performance schemes. The chaotic sequences were deployed to enhance the initialized solution quality. The entropy of Shannon was involved to show the population variation and then, tune the concerned variables.

Then, HPSO-Hybrid PSO [5] was a method for resolving the TAP - Task Assignment Problem. This method was associated with the issue of np-hard. PSO was a technologically advanced population dependent technique of heuristic optimization. The procedure had been established for dynamical scheduling of heterogeneous tasks and heterogeneous processors in a setup that were distributed. Balancing of load was an important problem in scheduling of task which was also considered in this work. This work of HPSO gave good performance than the Normal PSO to the issues pertaining to the assignment of task. The HPSO and PSO results were also correlated with some techniques of heuristic optimization namely Genetic Algorithm.

In general, Computing in Cloud environments were having huge heterogeneous self-operational systems collection along with an architecture of flexible computation. Task scheduling was a main step contributing to the cloud computing effectiveness. This scheduling was most important to enhance the service provider’s profit and consumption of low power, thereby minimizing the time taken for processing. [6, 7] made use of TSPSO – Task Scheduling using a multi objected nested PSO and orthogonal PSO for improvising processing time and energy optimization specifically. This proposed method outcomes were given rise by simulation with Cloudsim. In the environment of cloud computing, PSO was playing an crucial role in solving the issues of scheduling the task followed by optimizing the work flow pertaining to the system [8]. But, this adopted procedure falls easily into the local optimality for scheduling the scheme cost and time of execution rather than other methods. For addressing this difficulty and enhancing this method of computation, the methodology of traditional adaptive inertia weight particle group task scheduling was used by [8]. This method was able to equalize the particle search capacity among both the local and global levels to eliminate many local particles. This method gave rise to promising outcomes in terms of weight of inertia and state of the particle. Finally, this work was able to enhance the overall accuracy in convergence.

The cloud computing gave rise to many newer ways for application building, thereby giving many services to the users via the web virtualization. In cloud computing, task scheduling was the important element for using payable resources depending on the time. Thus, the load allocation was easily feasible among the resources of system by enhancing the usability and minimizing the executing time taken for executing the task. Hence, this research [9] made use of DAPSO – Dynamic Adaptive Particle Swarm Optimization for attaining the better performance and reducing the task set makespan along with the effective utilization of resources. This algorithm was a combination of CS and DAPSO methodologies known as MDAPSO methodology. Finally, they were able to observe that their hybridized methodologies surpassed the conventional PSO through the results.

As we already know, Cloud computing carried out client’s resource sharing and some services on-demand. For task scheduling, a new algorithm of hybridized GA-PSO methodology [10] was used. The goals of this methodology was also to provide minimum cost and then, balancing the load of designated task. Its performance was better than the methodologies of PSO, GA, WSGA-Work Scheduling with GA, HSGA-Hybrid heuristic scheduling with GA and MTCT. The proposed algorithm was quicker and much effective than the remainder algorithms.

Scheduling of appropriate resources to the workloads of cloud was a tough task, based on the requirements of Quality of Service in the applications of cloud. But still difficulties arising in dispersion, uncertainty and heterogeneity. So, [11] made a comprehensive preview of resource maintenance with scheduling of the task in the cloud environments. Their preview commented on how the researchers should choose appropriate methodologies for determining the workload schedule.

Since, the Cloud computing was the important method for enabling the connection capability of the VMs by TM-eFCFS – Task Migration based Scheduling methodology [12] using the enhanced-FCFS. This work make use of Non-live operation/task migration method for transmitting the tasks which were executed partially to another virtual machines for quick operations. The validation of this devised methodology was subjected to simulation using the cloudSim. Main factors like resource deployment and Makespan were investigated to validate the effectiveness of the TM-eFCFS.

A conventional yet and popular methodology of CMEFCFS – Cloudlet Migration based scheduling algorithm using Enhanced-First Come First Serve was used by [13]. This suggested work was subjected to simulation with the package of Cloudsim in order to validate its effectiveness. The validation was carried out to enhance parameters like execution time, completion time, and cost.

In distributed and parallel computing, task scheduling was an important task [14]. For the better task scheduling, UMR – Uniform Multi-Round was preferred to be used. Hence a novel methodology was suggested based on a modified form of UMR methodology, regarded as MSUMR – Master Service Uniform Multi-Round [14] was suggested. This algorithm increased the efficiencies of computing and scheduling irrespective of the bandwidth availability. Finally, on comparing with the existing works like UMR, extended multi-instalment and Multi-Instalment, this method outperformed by reducing the count of computer nodes that were unutilized.

[15] Divisible utilizations were used for the load partitioning which gave rise to few tiny fractions. This research contributes by portioning the whole divisible utilizations based on the load applications, available resources capacities and solutions. Its main objective was to reduce the time taken for the task completion by the intended application.

The dynamic nature of cloud computing provides ‘model of pay as per service pricing’ [16]. But cloud computing also had some problems like allocation of resources and scheduling the tasks to VMs. Thus, [16] founded the minimized time for completing the desired tasks with the aid of methodologies like BATS-Bandwidth Aware Divisible Scheduling, IDEA, ARIMA, and queuing vacation energy optimization. This research proved that the BATS would give rise to better completion time than the considered model of Berger source Data.

A novel meta-heuristic methodology comprising the integrated Genetic cum Firefly [17] was used to schedule the task in the cloud computing. It blends the mathematically formulated optimization by Firefly methodology along with the methodology of GA to contribute to the robust metaheuristic exploration. This methodology devised was analysed to cross verify its effectiveness against the existing methods of First In First Out and GA.

Cloud computing could process large amount of data by distributed services and its complex computational ability [18].This work [18] was objected to analyse many factors of PSO methodology in terms of its limitations, effectiveness and robustness. This work also made an investigation by examining the parameter suitability and cloud architecture virtualization in the platform with cloud computing as base.

A heuristic approach by [19] combined MAHP – Modified Analytic Hierarchy Process, LEPT – Longest Expected Processing Time pre-emption along with BATS. The resources were designated with the integration of BAR and BATS optimizing algorithms. The technique of dividing and conquering found to enhance the utilized IDEA – Improved Differential Evolutional Algorithm.

Cloud computing was the important technology in the services that were on demand [20]. The earlier method of scheduling tasks on the basis of requirements of resources for all the tasks processing without any reference to the storage, bandwidth, and memory. By introducing task scheduling method that primarily meets the requirements of users were able to yield better bandwidth, storage and memory within affordable costs.

TOPSIS – Technique of Order Precedence with Similarity to Ideal Solution [21] were majorly used for scheduling the task in the cloud environment. They divided this work into 2 phases where in TOPSIS was done in the initializing phase of task getting scheduled for concerned VMs and PSO was implemented in the next and final phase to schedule the predefined hierarchy. This integrated methodology of scheduling the tasks had better performance than the correlated existing methods such as ACO, multiple ACO and FUGE.

1.1 The major objective of this work is:

To schedule the task efficiently
To eliminate the faults in scheduling the tasks to the VMs.
To reduce the overlapping of the tasks to the VM’s in the hosts and to balance the load, we can use MLRHE for the management of load.
Finally, to validate the outcomes by comparing with the existing method

This section provides a detailed description of various existing scheduling methods with their advantages and disadvantages.

There were not that much work relating both the difficulties of VM load management with the Multi-layered regression host employment (MLRHE) to successfully schedule the tasks to the randomized VMs without any failure being occurring. In the literature, some of the works were found to coincide with our work. Some of them are discussed in the form of a summary table as follows:

Table 1

Summarizing table for regression type of load balancing
Citation Sequence	Base Methodology	Inference out of the work
[22]	Surveyed fundamentals of simulator in the cloud environment with reference cloudsim (Load controlling in VMs)	Working principle, arrangement, attributes, and functional abilities of cloudsim.
[23]	Simulation-annealing integrated GA for Migrating the VMs	Tackled issues concerning the scheduling and delegation of resources
[24]	Surveyed maintenance of energy aware cloud resource with dynamic VMC-VMs consolidation	Concluded that not much work with VMC methodologies
[25]	AI-Artificial Intelligence	Investigated AI regarding its ability to identify and track the areas of failures in the cloud environment to manage its outcomes
[26]	GWMA-Generally Weighted Moving Average	Addressed the problems of migration occurring in the VMs

This section contains the proposed workflow for task scheduling in the most economical that all the tasks could be completed in the shortest time possible with fewer resources without any failure.

For task scheduling in this work, PSO we make use of PSO and BATS for the successful scheduling of the task in the cloud environment. But, there will be an overloading of VMs in the host because there might be overlapping in the tasks that are being allocated to the VMs in the cloud environment. So, to tack this issue, we integrate the PSO-BATS task scheduling operation with the MLRHE to manage the load and reduce the chances of overloading as shown in the below Fig. 2

3.1. Task Scheduling with Novel PSO-BATS

In this research, efficient task scheduling is achieved with PSO-BATS methodology. As shown in Fig. 2 PSO is a method that operates primarily based on population. In this model, the set of particles presents and each particle represents a solution for the problem in search space. In the proposed task scheduling, the PSO model represents a set of VMs that are allocated to tasks in particles. All the particles in the predefined swarm will be knowing earlier optimal experience and find the global optimal experience.

Then, we introduce the BATS with PSO for making aware of the bandwidth. The flow for the proposed method is shown in the below Fig. 3.

In PSO, Particles update their exploration directions (their weights) using the following equations:[27]

$${v}_{i,j}=w \times {v}_{i,j} +{c}_{1} \times {r}_{1} \times \left({p}_{i,j}-{x}_{i,j}\right)+ {c}_{2}\times {r}_{2} \times \left({p}_{g,j}-{x}_{i,j}\right)$$

$${x}_{i,j}= {x}_{i,j}+ {v}_{i,j}$$

Here, ‘w’: inertia attribute depending on the local and global capabilities of the particle in the swarm, ${x}_{i,j}$: 'I' particle velocity in the dimension of j, ${c}_{1}\&{c}_{2}$ are weights influencing the social and cognitive attributes correspondingly. Immediately when the updating of velocity is done, the new positions will be determined. Likewise, this operation will be iterated for each measure and every particle existing in the swarm.

Then, BATS will be introduced to make it serve better by scheduling with an awareness of the bandwidth. This combination part is the novel part of this task scheduling. The algorithm for the PSO-BATS task scheduling is shown below.

3.1.1 Novel PSO with BAT Algorithm

Procedure BATS () //scheduling algorithm of Broker

If (TS()) //if there is a set of tasks submitted //TS = Task Submitted

GetVmInformation() //get computing power and bandwidth of VMs owned by Broker

GetTsInformation () //get information about tasks (Ts), including total number and size.A = Schedule (PSO model) //solve the model and obtain the optimized task allocation scheme End If

While (Exist_Idle()) // There is an idle VM in Virtual machine list vmlist

Vmli = GetNI(vmlist) // NI = Next Idle, get next idle VM vmli from vmlist

Ti = GetTN(C) //Get the Number of tasks, Ti, Allocated vmli in C

Bind(C, vmli) //bind Ti tasks to vmli

Set BW(vmli, b & Prime ; i) // BW = Bandwidth, set BW used in Task Transmission: b & Prime ; i = Xi / (T-Wi)

Schedule(PSO model)

{

1. Initialize EA //EA=$\cup E$

2. For i = 1 to m (m represents size of the Particle Swarm (PS))

3. Initialize ${A}_{j }$&${B}_{j }$ //PS and Velocity

4. For V = 1 to L // L represents number of Iteration

{

5. For I = 1 To M

E[j] = PSO (${A}_{j }$) // E[j] is the archive for particle swarm${A}_{j }$

6. Update the archive (EA) of Non-Dominated Solutions // External Archive

7. Choose head particle from EA

8. Update V //v = velocity

9. Update P //p = position

}

10. Return (Non-dominated Solution)

}

11. PSO (${A}_{j }$)

{

${A}_{j}$ : It defines the group of user tasks of jth particle that allocate to various Data Center (${D}_{c}), c=\text{1,2},\dots .P$

12. For t = 1 to P // P is the no of an available Datacenter in cloud

{

13. For i = 1 to N

{

14. Initialize${A}_{j }$

15. Initialize the velocity V of each particle V[i]

16. Initialize the Personal Best pB of each Particle: pB[i] =${A}_{j }$

Object evaluation of each Particle: Evaluate${A}_{j }$

17. Initialize the Global Best Particle (gB) with the best one among the ‘N’ particles: gB = Best Particle found in S

}

18. Add the Non-Dominated Solutions found in S into EAE [t] // EAE [t] is the External Archive storing the Pareto front for the task assign to Data Center${D}_{c}$

19. Initialize the Iteration Number (K) = 0

20. Repeat until K > Z

{

21. For i = 1 to N (Swarm Size)

{

Randomly select the Global Best Particle for S from the External Archive EAE[t] and store its position in gB. //gB = Global Best Particle

22. Calculate the new Velocity V[i] according to (7)

23. Calculate the new position of S[i] according to (8)

24. If $t<G*ProbMUT) then$ //(ProbMUT is the Probability of Mutation)

Perform Mutation on S[i]

Evaluate S[i] using (2) and (3)

25. Update Best Solution of each particle S[i]

26. Update the External Archive EAE[t]

}

27. Retain the Best Pereto Solution in EAE[t]

}

28. Return$(Min\left\{EAE\left[t\right].{MS}_{t=1..P}\right\},Sum\{EAE\left[t\right].{E}_{t=1..P}$}) //MS =MakeSpan, E=Energy

}

3.2. Multi-Layered Regression Host Employment MLRHE) to Tackle the Overloading of VMs

The crucial aspect of performing the regression operation is the relation to associate the independent attributes together for deriving a dependent measure based utilization of host wholly.

The non-trival component is considered as the host utilization that cannot be measured. To achieve the host utilization, the Geometric Relation (GR) function is used and it does not closely trace the behavior of actual host utilization. In this work, there are two alternative models are considers as the space distance for host utilization such as Euclidean Distance (ED) and Absolute Summation (AS). It enhances the energy consumption and service level agreement. The Euclidean distance between current and previous host utilizations as normalization constant and it is calculated by,

$$normVariED= \sqrt{{d\left(CPU\right)}^{2}+ {d\left(RAM\right)}^{2} + {d\left(BW\right)}^{2}}$$

We should be cautioned that $d\left(CPU\right)$, $d\left(BW\right)$, and $d\left(RAM\right)$indicate the relative difference intermediary to the recent and previous CPU occupying, memory and BW utilizations.

normConstAS = |d (CPU)|+|d (RAM)|+|d (BW)|

The GR function is a multi-factorization relationship which combines multiple parameters in same metric. The considered critical factors are CPU, memory and BW for VMs. But those factors of absolute values are not required to use. The mentioned factors of utilization relative to maximum permissible utilization is used to make the dimensionless factors and for host overload. The highest utilization is defined by CSP (Cloud Service Provider) or absolute host utilization. The available data is to be normalized to represent a functional utilization per factor. The GR function for host utilization evaluation is used in[28] [29] as shown in Eq. 1. However, The GR lacks the consideration of the orthogonality of the multi-dimensional space among the different factors.

$$Utilization of Virtual Machines is given by$$

$$= \frac{{w}_{1}}{1-CPU}\times \frac{{w}_{2}}{1-RAM}\times \frac{{w}_{3}}{1-BW}$$

where w₁The weight of for factor i, where the factors are CPU, memory and BW CPU The relative CPU utilization RAM The relative memory utilization BW The relative network utilization.[30]modified the VMs utilization relation as

$$Utilization of Virtual Machines=\frac{CPU+RAM+BW}{normVariED}$$

$$Utilization of Virtual Machines=\frac{CPU+RAM+BW}{normConstAS}$$

We make use of all these relations and concepts for overcoming the overlapping of the tasks in the task scheduling to achieve the faultless outcomes out of our work.

The proposed method of task scheduling by PSO-BATS and load management by MLRHE is evaluated and validated in terms of Makespan, cost, and PIR-Performance Improvement Rate with the existing methods of [31], [32], and [33].

4.1 Performance Comparison with Makespan

The Makespan is the highest/peak period or the period at which the recently assigned task is been completed by IaaS-Infrastructure as a Service in the cloud environment. We use the existing work of [31]to make the comparison of the Makespan yielded by this proposed method.

The above graph shows the comparison of the makespan of both the existing and proposed methods. For the counts of 100 virtual machines, counts of tasks of 250, 500, 750, and 1000 are taken to validate the effectiveness of the proposed method. For all the number if tasks, the proposed method was found to yielded lesser makespan than all the existing methods.

Similarly, the above Fig. 6 plotted for 200-Vms too, the proposed task scheduling with load management methodology yielded lesser values of make span for the task counts of 250, 500, 750, and 1000.

4.2 Performance comparison with Cost

The task cost could be comprehended as the ratio of the cost incurred to generate the appropriate task scheduling depending upon on the VM counts in motion to the initial VM counts defined in a specific operating machine.

We use the existing work of [31]to make a comparison of the cost incurred while scheduling the tasks via our method.

4.3 Performance Comparison with PIR

The performance metric PIR-Performance Improvement Rate could be found by % of performance betterment or improvisation contributed by the proposed methods against the remaining methods under comparison. PIR can be given by the relation,

$$\text{P}\text{I}\text{R}=\text{M}\text{a}\text{k}\text{e}\text{s}\text{p}\text{a}\text{n}\left(\text{r}\text{e}\text{m}\text{a}\text{i}\text{n}\text{d}\text{e}\text{r} \text{o}\text{u}\text{t}\text{l}\text{i}\text{n}\text{e}\right)-Makespan\left(PSO-BATS and MLRHE\right)\times \frac{100}{Makespan\left(PSO-BATS and MLRHE\right)}$$

When we compare the PIR of the existing and proposed method, we consider the data taken from [32] to make the comparison.

The below table is tabulated to show the PIR values of the proposed method against the existing method concerning overall makespan respectively.

Table 2

PIR of the proposed method against the existing methods
	MTCT	MAXMIN	ACO	NSGA-II	DCLCA	Proposed
Total makespan	14,042.70	13,671.90	11,057.40	10,099.50	8898.8	7994.4
PIR % over MTCT		2.7	27	39	57.8	75.7
PIR % over MAXMIN			23.6	35.4	53.6	71
PIR % over ACO				9.5	24.3	38.3
PIR % over NSGA-II					13.5	26.3
PIR % over DCLCA						11.31

The above Fig. 8. shows that the proposed method yielded PIRs of 75.7 %, 71 %, 38.3 %, and 26.3 and 11.31 % than the existing methods of MTCT, MAXMIN, ACO, NSGA-II, and DCLCA respectively.

4.4 Performance depiction with mean numbers and mean marking

The CloudSim is used to create a simulation and it contains the classes for various cloud entities like Host, Datacenter and VMs. The Cloudlet represents a submitted task in the cloud. Each Cloudlet is assigned to a VM. Datacenter broker acts on behalf of a user, it hides VM management. Here, the mean numbers of brokers and cloudlets generated via the CloudSim concerning time are taken and plotted in the graph to know the behavior and outcomes arising out of the performed task scheduling.

This above graph is plotted to show the cloud model of the proposed method by computing and plotting the mean numbers of brokers operating in the system along with the cloudlets count. We cloud see that there were more faults when there is an increase in the number of brokers working in the cloud model.

This above graph is plotted to show the cloud model of the proposed method by computing and plotting the mean numbers of brokers operating in the system, count of cloudlet concerning VMs on hosts 1& 2 respectively. We cloud see that there were more faults when there is an increase in the number of brokers working in the cloud model.

This above graph is plotted to show the cloud model of the proposed method by computing and plotting several places/locations of marking like p1, p3, p5, p6, and p8 in the cloud mode. All these locations will remain marked unless the values tend to the value of 0. The location p3 is found to produce decreasing mean markings for the tasks which are being executed in the cloud environment (i.e.) only the minimum number of tasks will be needed to complete the tasks.

In cloud computing, task scheduling is a challenging factor to meet thousands of requests by cloud resources for task manager. Due to the disadvantages of existing methods of task scheduling, this research proposed novel PSO-BATS integrated with MLRHE. It will meet all the challenges and provide efficient performances compared to traditional scheduling methods. Also, traditional methods had the difficulty of scheduling the tasks, overloading of VMs in the host in the physical machines. This research aims to schedule the task efficiently and to eliminate the faults in scheduling the tasks to the VMs. The proposed system effectively extends the profits of cloud computing with the job scheduling system. The proposed method of results analyzed by using various parameters such as Makespan, cost, and PIR and found that the proposed method surpassed the other taken methods. We compared the performance of our proposed method with existing OCSA, PSO, GA and IDEA methods, it proved that the proposed method performs better than other existing methods.

This research only involves the optimization of task allocation with the measures of Makespan, cost, and PIR. In the future, the relationship between resources and their impact will investigate to improve cloud scheduling.

ACO	Ant Colony Optimization
AI	Artificial Intelligence
ARIMA	Autoregressive Integrated Moving Average
AS	Absolute Summation
BW	Bandwidth
CMEFCFS	Cloudlet Migration using Enhanced-First Come First Serve
CPU	Central Processing Unit
DCLCA	Dynamic Clustering League Championship Algorithm
DAPSO	Dynamic Adaptive Particle Swarm Optimization
EA	External Archive
ED	Euclidean Distance
FIFO	First In First Out
FUGE	combination of fuzzy idea and GA
GA	Genetic Algorithm
GR	Geometric Relation
GWMA	Generally Weighted Moving Average
HPSO	Hybrid-PSO
IDEA	Improved Differential Evolutional Algorithm
IaaS	Infrastructure as a Service
LU	Logical Unit
MAXMIN	Maximum-Minimum
MLRHE	Multi-Layered Regression Host Employment
MDAPSO	Modified Dynamic Adaptive Particle Swarm Optimization
MSUMR	Master Service Uniform Multi-Round
MAHP	Modified Analytic Hierarchy Process
MTCT	Min–min based Time and Cost Trade-off
NSGA-II	Non-domination Sorting Genetic Algorithm-II
OCSA	Oppositional Cuckoo Search Algorithm
PSO	Particle Swarm Optimization
PIR	Performance Improvement Rate
PSO-BATS	PSO- Bandwidth Aware divisible Task Scheduling
ProbMUT	Probability of Mutation
RAM	Random Access Memory
TAP	Task Assignment Problem
TSPSO	Task Scheduling-PSO
TM-eFCFS	Task Migration-enhanced First Come First Serve
TPOSIS	Technique of Order Precedence with Similarity to Ideal Solution
TS	Task Submitted
UMR	Uniform Multi-Round
VM	Virtual Machines
VMC	VMs Consolidation
WSGA	Work Scheduling with GA

Funding

This research work was not funded by any organization/institute/agency.

Conflicts of interest

I confirm that this work is original and has either not been published elsewhere, or is currently under consideration for publication elsewhere. None of the authors have any competing interests in the manuscript.

Data Availability

https://doi.org/10.14257/ijunesst.2016.9.1.36

https://doi.org/10.1007/s00366-015-0396-z

https://doi.org/10.1002/cpe.4368

https://doi.org/10.1016/j.asoc.2018.09.022

https://doi.org/10.1504/ijbic.2009.024726

https://doi.org/10.1016/j.procs.2015.07.419

https://doi.org/10.23940/ijpe.17.07.p8.10631069

https://doi.org/10.14257/ijgdc.2015.8.5.24

https://doi.org/10.1155/2018/1934784

https://doi.org/10.1007/s10723-015-9359-2

https://doi.org/10.1007/978-981-10-8660-1_9

https://doi.org/10.3233/jifs-169034

https://doi.org/10.3390/su10051611

https://doi.org/10.1109/eco-friendly.2016.7893233

https://doi.org/10.1007/978-3-030-24318-0_77

https://doi.org/10.2139/ssrn.3349577

https://doi.org/10.1186/s13677-018-0105-8

https://doi.org/10.1007/978-981-10-5146-3_10

https://doi.org/10.1007/s10586-019-02915-3

https://doi.org/10.17148/ijarcce.2015.4126

https://doi.org/10.1109/dasc-picom-datacom-cyberscitec.2016.108

https://doi.org/10.1007/978-3-319-62238-5_6

https://doi.org/10.1016/j.comnet.2019.106950

https://doi.org/10.1007/978-3-030-34986-8_26

https://doi.org/10.1109/icnn.1995.488968

https://doi.org/10.1016/j.eij.2016.12.002

https://doi.org/10.1038/sj.jors.2600611

https://doi.org/10.22266/ijies2018.0630.29

https://doi.org/10.1007/s00521-016-2448-8

https://doi.org/10.1016/j.simpat.2018.09.018

Code availability

N\A

Author’s Contributions

I Am Anwar R Shaheen Hereby State That The Manuscript Title Entitled “Tasks Scheduling in Cloud environment using PSO-BATS” Submitted To The Wireless Personal Communications, I and my Co-author Sundar Santhosh kumar Confirm That This Work Is Original And Has Not Been Published Elsewhere, Nor Is It Currently Under Consideration For Publication Elsewhere. And I Am Research Scholar in the Department of Computer Science at Alagappa University, Karaikudi, and Tamil Nadu, India.

Ethics approval

No animals or human participants are involved in this research work.

Informed consent

I confirm that any participants (or their guardians if unable to give informed consent, or next of kin, if deceased) who may be identifiable through the manuscript (such as a case report), have been given an opportunity to review the final manuscript and have provided written consent to publish.

Xue, S., Shi, W., & Xu, X. (2016). "A heuristic scheduling algorithm based on PSO in the cloud computing environment,". International Journal of u-and e-Service, Science and Technology, 9, 349–362
Kumar, N., & Vidyarthi, D. P. (2016). "A novel hybrid PSO–GA meta-heuristic for scheduling of DAG with communication on multiprocessor systems,". Engineering with Computers, 32, 35–47
Ebadifard, F., & Babamir, S. M. (2018). "A PSO-based task scheduling algorithm improved using a load‐balancing technique for the cloud computing environment,". Concurrency and Computation: Practice and Experience, 30, 4368
Zhang, H., Xie, J., Ge, J., Lu, W., & Zong, B. (2018). "An Entropy-based PSO for DAR task scheduling problem,". Applied Soft Computing, 73, 862–873
Visalakshi, P., & Sivanandam, S. (2009). "Dynamic task scheduling with load balancing using hybrid particle swarm optimization,". Int. J. Open Problems Compt. Math, 2, 475–488
Sivanandam, S., & Visalakshi, P. (2009). "Dynamic task scheduling with load balancing using parallel orthogonal particle swarm optimisation,". International Journal of Bio-Inspired Computation, 1, 276–286
Jena, R. (2015). "Multi objective task scheduling in cloud environment using nested PSO framework,". Procedia Computer Science, 57, 1219–1227
Zhao, S., Fu, X., Li, H., Dong, G., & Li, J. (2017). "Research on cloud computing task scheduling based on improved particle swarm optimization,". International Journal of Performability Engineering, 13, 1063–1069
Al-Maamari, A., & Omara, F. A. (2015). "Task scheduling using PSO algorithm in cloud computing environments,". International Journal of Grid and Distributed Computing, 8, 245–256
Manasrah, A. M., & Ba Ali, H. "Workflow scheduling using hybrid ga-pso algorithm in cloud computing," Wireless Communications and Mobile Computing, vol(2018). 2018.
Singh, S., & Chana, I. (2016). "A survey on resource scheduling in cloud computing: Issues and challenges,". Journal of grid computing, 14, 217–264
Panwar, N., Negi, S., & Rauthan, M. M. S. "Non-live task migration approach for scheduling in cloud based applications," in International Conference on Next Generation Computing Technologies(2017). pp. 124–137.
Panwar, N., Negi, S., Rauthan, M. S., Aggarwal, M., Jain, P., & "AN ENHANCED SCHEDULING APPROACH WITH CLOUDLET MIGRATIONS FOR RESOURCE INTENSIVE APPLICATIONS,".,. (2018). Journal of Engineering Science and Technology, 13, 2299–2317
Zhao, T., & Jing, M. (2016). "Bandwidth-aware multi round task scheduling algorithm for cloud computing,". Journal of Intelligent & Fuzzy Systems, 31, 1053–1063
Li, B., He, M., Wu, W., Sangaiah, A., & Jeon, G. "Computation Offloading Algorithm for Arbitrarily Divisible Applications in Mobile Edge Computing Environments: An OCR Case," Sustainability, vol. 10, p. 1611(2018).
Gawali, M. B., & Shinde, S. K. "Implementation of IDEA, BATS, ARIMA and queuing model for task scheduling in cloud computing," in(2016). Fifth International Conference on Eco-friendly Computing and Communication Systems (ICECCS), 2016, pp. 7–12.
Rajagopalan, A., Modale, D. R., & Senthilkumar, R. (2020). "Optimal Scheduling of Tasks in Cloud Computing Using Hybrid Firefly-Genetic Algorithm,". in Advances in Decision Sciences, Image Processing, Security and Computer Vision (pp. 678–687). ed: Springer
Kumar, N. (2019). "Parameters Analysis for PSO based Task Scheduling in Cloud Computing,"Available at SSRN3349577,
Gawali, M. B., & Shinde, S. K. (2018). "Task scheduling and resource allocation in cloud computing using a heuristic approach,". Journal of Cloud Computing, 7, 4
Alworafi, M. A., Al-Hashmi, A., Dhari, A., & Darem, A. B. "Task-scheduling in cloud computing environment: cost priority approach," in Proceedings of International Conference on Cognition and Recognition(2018). pp. 99–108.
Panwar, N., Negi, S., Rauthan, M. M. S., & Vaisla, K. S. (2019). "TOPSIS–PSO inspired non-preemptive tasks scheduling algorithm in cloud environment,". Cluster Computing, 22, 1379–1396
Vahora, S., & Patel, R. (2015). "CloudSim-A Survey on VM Management Techniques,". International journal of advanced research in Computer and communication engineering, 4, 128–133
Li, Y., Meng, X., & Dong, H. "A simulated annealing combined genetic algorithm for virtual machine migration in cloud datacenters," in 2016 IEEE 14th Intl Conf on Dependable, Autonomic and Secure Computing, 14th Intl Conf on Pervasive Intelligence and Computing, 2nd Intl Conf on Big Data Intelligence and Computing and Cyber Science and Technology Congress (DASC/PiCom/DataCom/CyberSciTech)(2016). pp. 572–577.
Khan, M. A., Paplinski, A., Khan, A. M., Murshed, M., & Buyya, R. (2018). "Dynamic virtual machine consolidation algorithms for energy-efficient cloud resource management: a review,". in Sustainable Cloud and Energy Services (pp. 135–165). ed: Springer
Gupta, L., Salman, T., Zolanvari, M., Erbad, A., & Jain, R. (2019). "Fault and performance management in multi-cloud virtual network services using AI: A tutorial and a case study,". Computer Networks, 165, 106950
Chen, J. H., & Lu, S. L. "GWMA Algorithm for Host Overloading Detection in Cloud Computing Environment," in International Conference on e-Business Engineering(2019). pp. 358–370.
Eberhart, R., & Kennedy, J. "Particle swarm optimization," in Proceedings of the IEEE international conference on neural networks(1995). pp. 1942–1948.
Abdelsamea, A., El-Moursy, A. A., Hemayed, E. E., & Eldeeb, H. (2017). "Virtual machine consolidation enhancement using hybrid regression algorithms,". Egyptian Informatics Journal, 18, 161–170
Anandharajan, T., BHARGAVAN, D., & Bhagyaveni, M. (2013). "VM CONSOLIDATION TECHNIQUES IN CLOUD DATACENTER,"Journal of Theoretical & Applied Information Technology, vol. 53,
Romero, C., Tamiz, M., & Jones, D. (1998). "Goal programming, compromise programming and reference point method formulations: linkages and utility interpretations,". Journal of the Operational Research Society, 49, 986–991
Krishnadoss, P., & Jacob, P. (2018). "OCSA: task scheduling algorithm in cloud computing environment,". Int J Intell Eng Syst, 11, 271–279
Latiff, M. S. A., Madni, S. H. H., & Abdullahi, M. (2018). "Fault tolerance aware scheduling technique for cloud computing environment using dynamic clustering algorithm,". Neural Computing and Applications, 29, 279–293
Barbierato, E., Gribaudo, M., Iacono, M., & Jakóbik, A. (2019). "Exploiting CloudSim in a multiformalism modeling approach for cloud based systems,". Simulation Modelling Practice and Theory, 93, 133–147

Tasks Scheduling in Cloud environment using PSO-BATS

Status:

Version 1

Abstract

Figures

1. Introduction

2. Background

3. Preview Of The Proposed Method

3.1. Task Scheduling with Novel PSO-BATS

3.1.1 Novel PSO with BAT Algorithm

3.2. Multi-Layered Regression Host Employment MLRHE) to Tackle the Overloading of VMs

4. Performance Investigation

4.1 Performance Comparison with Makespan

4.2 Performance comparison with Cost

4.3 Performance Comparison with PIR

4.4 Performance depiction with mean numbers and mean marking

5. Conclusion

Abbreviations

Declarations

References

Status:

Version 1