An AHP based Task Scheduling and Optimal Resource Allocation in Cloud Computing

—Cloud systems by virtue characterize ultimate resource utilization with ever evolving user requirements facilitating adaptivity. With a scope of enhancing the QoS needs of user applications, numerous factors are considered for tunning among which Task scheduling promises to grab focus. The Task Scheduling mechanism ascertains improvement by distributing the subtasks to specific set of resources pertaining to prevailing Quality models. The work emphasizes the need for effective task scheduling and optimizing resource allocation by modelling a modified AHP (Analytical Hierarchy Process) driven approach. The proposed method guarantees the functionality in two phases pertaining to Task ranking and pipelined with Optimized scheduling algorithms resulting in maximization of resource utilization. The former phase of task ranking is aided by improved AHP with substantial usage of fuzzy clustering followed by an enhanced CUCMCA (Chimp Updated and Cauchy Mutated Coot Algorithm) algorithm for optimal resource allocation of cloud applications. The contributed model promises leveraged performance of 32% for memory usage, 33.5% for execution time, 29% for makespan and 18% for communication cost over pre-existing conventional models considered.


INTRODUCTION
Cloud computing is a flexible strategy for exchanging distributed services and resources with the user wherever and anytime they require [1]. Because of its scalable user pool of services and resources, cloud computing has become more popular in recent years. This is because it allows users to freely control their consumption and only pay for the cloud resources they actually utilize [10] [28]. Google Compute Engine, Rackspace Cloud, and Amazon EC2 are just a few of the commercial cloud computing systems that have recently entered the market [2] [15]. Moreover, for hosting and delivering software solutions for many industrial applications, cloud computing has established itself as a dependable, affordable, and scalable service option [14] [12]. Still, a cloud must have sufficient capacity to meet peak user demand in order to uphold user expectations for QoS [19] [16].
Resource usage, server power consumption, and storage all play significant roles in the cloud environment, making resource management essential. While provisioning and allocating cloud resources, availability was frequently used as the determining factor without considering the other essential factors like resource utilization or the server's thermal properties [9]. A cloud system's ability to manage its resources autonomously and adaptively based on the workload changes is known as autonomous RM [20]. Elastic resource management encompasses a wide range of processes, including balanced virtual machine and application scheduling, server over/under-load control utilizing VM migration, etc. [3] [17]. One or more physical machine resources, including CPU, memory, I/O, and network bandwidth, may be overloaded due to the increased load of virtual machine operations [5] [8].
The scheduling solution was optimized in terms of each metric specified in the QoS model using a QoS-driven CESS method [4]. It has been suggested to use an OP-MLB system, which combines a number of algorithms that cooperate to provide effective resource management for cloud environments [6] [7]. With more effective storage and V/F scaling improvement, the EARU model significantly reduces LLC disappointments and thus more effectively utilizes asset. In terms of CPU utilization, preparation time, and energy output, it also achieves preferred execution to the board's current asset management plan [11]. Greedy method named MESA is recommended due to the high computing complexity of addressing the MINLP problem [18] in order to arrive at the best solution [13]. Also, advanced metaheuristic models are in need to proceed with optimal scheduling process. www.ijacsa.thesai.org  Initially, modified AHP process is introduced for ranking the task.
 A hybrid optimization model, namely, CUCMCA method for optimal scheduling of task with appropriate allocation of resources to execute the task.
 The proposed method is implemented using CloudSim simulator.
The work progresses initiating with intense literature review in Section II paving path for improvisation issues addressed with modified AHP based task ranking and hybrid optimal resource allocation provided in Section III followed by proposed CUCMCA Algorithm in Section IV formulates results and discussions in Section V Section VI contributes to conclusion of work with a wide overview of work.

II.
RELATED WORK Several Researchers have contributed innumerable solutions addressing issues of scheduling and resource allocation. Despite, leaving few coins unturned that are addressed in our work with enhancements improvising performance.
In 2020, Mahdi Abbasi et al. [1] presented two approaches, XCS and BCM-XCS, depending on XCS -LCS, to manage the network's edge power consumption and lessen workload delay. The outcomes of this tests show that BCM-XCS is superior to the standard XCS-based approach. The workloads were distributed using the suggested approaches in a way that both the communication and processing delay among cloud and fog nodes were kept to a minimum. Additionally, the suggested approaches can recharge the reusable batteries utilized at the network's edge 18% faster than the existing technique.
In 2020, Yunliang Chen et al. [2] suggested a detailed QoS model to evaluate the performance level of data center clouds. To improve the cumulative QoS and sojourn time of all activities, an enhanced CESS algorithm was created. According to experimental findings, this approach outperforms the baseline algorithm in terms of accumulative QoS as well as sojourn duration by up to 56.1% and 25.4%, correspondingly. The algorithm's duration only increases linearly as more Cloud data centres and workloads are added. This technique constantly develops scheduling solutions with acceptable QoS without compromising sojourn time when the arrival rate as well as service rate ratio are kept constant.
In 2022, M. Hasan Jamal et al. [6] suggested a HAWDA and HASRA depending on thermal profiling considering outlet temperature detection. In order to reduce the peak output temperatures, HAWDA distributed workload on servers in a thermally efficient manner, while HASRA optimized server positioning in thermal hotspot areas. To evaluate the effectiveness of HAWDA against the TASA and GRANITE methods, performance comparison is done. Results showed that HAWDA, which reduces peak outlet temperature, achieved average peak server utilization comparable to GRANITE as well as TASA without adding additional load to the cooling system, with or without server relocation.
In 2022, M. Hasan Jamal et al. [6] suggested a HAWDA and HASRA depending on thermal profiling considering outlet temperature detection. In order to reduce the peak output temperatures, HAWDA distributed workload on servers in a thermally efficient manner, while HASRA optimized server positioning in thermal hotspot areas. To evaluate the effectiveness of HAWDA against the TASA and GRANITE methods, performance comparison is done. Results showed that HAWDA, which reduces peak outlet temperature, achieved average peak server utilization comparable to GRANITE as well as TASA without adding additional load to the cooling system, with or without server relocation.
In 2020, Ali Asghari et al. [7] developed a new architecture made up of many cooperating agents that took into account all aspects of TS and resource provisioning and managed the QoS offered to users. The integrated model that was suggested included all processes for TS and resource provisioning, and its many components help with managing user applications and making better use of cloud resources. This framework performs effectively with concurrent dependent activities, which have a challenging scheduling procedure due to the dependency of their subtasks.
In 2021, A. M. Senthil Kumar et al. [8] suggested a new task allocation method employing BAT and CSO method. The BAT algorithm aids the CSO algorithm in overcoming a pre convergence problem. The suggested HGCSBAT algorithm's performance was assessed and contrasted with that of the well-known CSO & BAT methods. In regards to availability & throughput, HGCSBAT performs better than that of the BAT, Cat Swarm Optimization, & Genetic algorithms. Traditional work scheduling algorithms features and limitations are given in Table I. In 2023 K.Pradeep, Sharma, and Jishnu [29] proposed an intense review on Task scheduling parameters which shed light on various strategies that make way for efficient scheduling with fault tolerant approaches in Fog computing. Their work emphasizes on tuning QoS parameters for improved results.

III.
MODIFIED AHP BASED TASK RANKING AND HYBRID OPTIMAL SCHEDULING WITH APPROPRIATE RESOURCE ALLOCATION Fig. 1 depicts the scenarios of Task Scheduling where user requests assemble at various virtual machines pertaining to a physical machine. These VMs promise improved performance by scheduling the tasks using our proposed approach. Scheduling of Tasks is performed in two phases i.e., Task ranking using FCM clustering followed by CUCMCA optimization for enhanced results.

A. System Model
Considering a data center with M servers include servers types were given to it. The proposed workload scheduling and resource management is progressed with two different steps: 1) Task ranking: This phase handles the process for ranking the tasks as per their priorities by first identifying priorities and generating Task queues pertaining to priority groups.
2) Optimal scheduling: Optimal scheduling phase deals with the assignment of corresponding resources according to the constraints resulting in better performance. www.ijacsa.thesai.org

B. Task Ranking
This is the first stage, where the ranking of task was done by using modified AHP (Analytical Hierarchy Process). Conventional AHP steps for ranking purpose are as follows:  Implement the Saaty preference table [23]. Given below is Saaty preference table which offers point scale including descriptors (Table II). First condition is slightly more significant than second condition 5 First condition is rather more significant than second condition. 7 Obviously, the first condition is more significant than the second one. 9 The first condition is unquestionably more significant than the  Afterwards each column summation and each column normalization and weighted sum is computed [24].
 Based on the Saaty preference table, AHP ranking is performed.
According to recommended concept, modified AHP process is followed for ranking purpose. Modified AHP steps are as follows:  Instead of using the Saaty preference table, in modified AHP, FCM (Fuzzy c-Means Clustering) is followed on the basis of 1-9 clustering for implementing the table.
A data collection is divided into N clusters using the FCM data clustering method, with each cluster having some of the data points in the dataset.
 Based on the FCM, AHP ranking is performed for task scheduling. Modified AHP based ranking is shown in Fig. 2.

C. Optimal Task Scheduling and Resource Management
In this step, the process of optimal scheduling is done by selecting the optimal PM and VM to execute the tasks. During this process, certain constraints are evaluated to ensure the scheduling in a precise manner. Also, the constrains involve in the assignment of corresponding resources  to execute the tasks involved.
Constraints involved in this optimal scheduling process are as follows: www.ijacsa.thesai.org

( )
Solution encoding: The selection process will be decided by the proposed optimization algorithm where the solution including both PM and VM set from which the model selected corresponding PM and VM to execute the respective task. Fig. 3 gives solution encoding of proposed CUCMCA method. These weights are calculated using chaotic cubic map function. (4)

Cubic map:
The cubic map is one of the maps that are most frequently used to generate chaotic sequences in many applications. In eq. (5), cubic map is defined.
IV. CHIMP UPDATED CAUCHY MUTATED COOT ALGORITHM (CUCMCA) OPTIMIZATION Hybrid optimizations are a new class of optimization methods that we develop to solve the optimization issue with more convergence efficiency. For hybrid optimizations, two or more algorithms must have been used for the same optimization. In this paper, we hybridized two algorithms named coot and chimp optimization. The solution update is done by this hybridized algorithm. Here the random number is estimated by using Tent map function. Also, Cauchy mutation is introduced in our proposed concept. Our proposed hybrid algorithm concept is given below: Small water birds called Coots [21] were the rail family members. In coot optimization, with the formula (6), a small area is used to produce the population at random, where Coot pos(i) represents the coot position, g represents the variables count, and ub,lb represents the upper as well as lower bound of search space.
Random motion to this direction and that direction: Different areas of the search space are explored by coot migration. This movement will let the algorithm escape the local optimal if it becomes stuck in the local optimal. Coot's new position is determined in eq. (7).
Where, K 2 represents the random number which is calculated using tent map function according to proposed Tent map: Tent chaotic map is also refereed as the logistic map represents particular chaotic effects. The following equation (8) gives the definition of this map: Chain movement: It is possible to construct chain movement by using the average position of 2 coots. The formula (9) is used to calculate the coot's new position, where ( ) represents the second coot.
( ) Adjusting position in accordance with the group leaders: To carry out this movement, a system is deployed based on the formula (10) to choose the leader, where i represents current coot index, D represents leader index, and lc represents the leader count.
According to proposed model, position update is done by hybridizing Coot and Chimp position [22] which is specified in eq. (12) to eq. (16).

Proposed update equation:
( ) Leader movement: Leaders need to change their location with respect to the objective in order to move the group toward an objective (the ideal region). It is advised to update leader position using formula (17), where OBest represents the best position, J 3, J 4 represents the random number.
Here, ( ) represents maximum iteration, and R represents the current iteration.
Cauchy mutation: Cauchy mutation is also employed in this algorithm to produce the solution. Due to its wider search range, Cauchy mutation has a significant ability to seek globally. This ensures the high convergence rate. Below Algorithm 1 gives the pseudocode of suggested CUCMCA model:

B. Dataset Description
This is a trace of the workloads running on eight Google Borg compute clusters for the month of May 2019 [25]. The trace describes every job submission, scheduling decision, and resource usage data for the jobs that ran in those clusters.
It builds on the May 2011 trace of one cluster, which has enabled a wide range of research on advancing the state-ofthe-art for cluster schedulers and cloud computing, and has been used to generate hundreds of analyses and studies.
Since 2011, machines and software have evolved, workloads have changed, and the importance of workload variance has become even clearer. The new trace allows researchers to explore these changes.
The new dataset includes additional data, including:  CPU usage information histograms for each 5 minute period, not just a point sample;  Information about allow sets (shared resource reservations used by jobs); and  Job-parent information for master/worker relationships such as MapReduce jobs.
Just like the last trace, these new ones focus on resource requests and usage, and contain no information about end users, their data, or access patterns to storage systems and other services."

C. Evaluation of Communication Cost
The proposed CUCMCA method is compared to extant systems in terms of communication cost. Fig. 4

D. Evaluation on Execution Time
In this section, the analysis on Execution Time is examined for varied VMs. The analysis on developed CUCMCA method over BES, AOAM, MFO, HSO, EHO, Chimp and COOT for varied VMs is exposed in Fig. 5. In order to improve the system's performance, the execution time should be reduced. The developed approach holds minimal execution time of 8.24 at the VM 50; whereas, the traditional models holds the highest execution time for BES (39.53), AOAM (47.92), MFO (49.67), HSO (62.18), EHO (18.65), Chimp (48.85) and COOT (53.67), respectively. Moreover, it is observed that, the suggested method has accomplished better outcomes at 30th VM than at 10th and 20th VMs. As a result, the superiority of the suggested CUCMCA work is proved. www.ijacsa.thesai.org

E. Evaluation of Fitness
The Fitness analysis of the proposed CUCMCA classifier is computed over the existing classifiers and the graphical illustration is represented in Fig. 6

F. Evaluation of Makespan Time
The Makespan using suggested CUCMCA method is analyzed over BES, AOAM, MFO, HSO, EHO, Chimp and COOT for varied VMs as shown in Fig. 7. Low makespan is required for enhanced system performance. The suggested method successfully achieves our goal, whose Makespan values exceed the traditional approaches. Moreover , respectively. This analysis shows that the developed CUCMCA approach makes the system more robust at scheduling workloads and managing resource than the conventional approaches.

G. Analysis on Memory Utilization
The performance of adopted CUCMCA model regarding Memory Utilization is displayed in the Fig. 8. The memory utilization should be minimal for better system performance. In this manner, it is observed that the adopted model achieves least memory utilization when compared to models like BES, AOAM, MFO, HSO, EHO, Chimp, and COOT, respectively. Particularly, incredible outcomes for both proposed and current approaches have been obtained for all measures at the 10th, 30th, and 50th VM. Nevertheless, the developed approach has delivered more determinative outcomes than distinct strategies for every VM. For instance, at 50th VM, the memory utilization of suggested approach is 0.04, which is better than the values obtained for existing schemes like BES is 0.05, AOAM is 0.06, EHO is 1.2, COOT is 1.1 and HSO is 0.07, respectively. As a consequence, this assessment proves that the proposed CUCMCA model is better to make an efficient workload scheduling when compared to other conventional models.

H. Convergence Analysis
The convergence study of the proposed CUCMCA work is contrasted to the traditional methods like BES, AOAM, MFO, HSO, EHO, Chimp and COOT are shown in Fig. 9. In addition, it was examined by altering the iterations to 5, 10, 15, 20 and 25, respectively. When comparing the suggested method to other extant schemes, the findings show that the developed model has the lowest error rate. Here, the COOT algorithm has exhibited the worst performance in the initial (0th) iteration. From iteration 5 through iteration 25, the proposed approach and other current classifiers have lower error rates. Nevertheless, at the last 25th iteration, the adopted approach recorded the lowest error rate of (~) 1.0. Thus, recommended strategy resulted in a slightly lower error rate www.ijacsa.thesai.org than BES, AOAM, MFO, HSO, EHO, Chimp and COOT. Therefore, the proposed CUCMCA strategy is appropriate for the workload scheduling and resource management. The improvisation in the performance of proposed logic shows the impact of proposed hybrid algorithm in enhancing the convergence rate and speed.  Table III represents the statistical analysis with respect

VI. CONCLUSION AND FUTURE SCOPE
Cloud system must be able to manage its resources autonomously and adaptively in response to the changes in workload needs with its vast computing power and flexibility for ever evolving challenges. Despite credibility features, it faces many challenges such as Scheduling, Security, and Energy Management etc. Among aforementioned issues the one of concern that owes to be improved is Task scheduling, intending the maximization of user-favoured application QoS parameters our proposed hybrid algorithm performs task scheduling and allocates resources efficiently in cloud computing environments. Our work considers Google cloud (May 2019) workload traces as input using modified AHP to rank the task via FCM. Furthermore, the optimal task scheduling and resource allocation are done by the developed Chimp Updated and Cauchy Mutated Coot Algorithm (CUCMCA). Deliberately the outcome promises improved results in comparison to the existing conventional BES, AOAM, MFO, HSO, EHO, Chimp, COOT models with respect to makespan time, execution time, communication cost and memory utilization gives improved results. As a scope for further enhancement classifying user tasks prior to scheduling promises improved results in terms of QoS metrics and extends scope for better resource allocation.