Optimized efficient job scheduling resource (OEJSR) approach using cuckoo and grey wolf job optimization to enhance resource search in cloud environment

Abstract

Cloud computing is a technique for developing systems that rely on dynamic sharing to enable the integration of several systems to provide services. For the purpose of completing their intended work, cloud users may literally access the cloud resources over the internet. The tasks that cloud users submit and the time it takes to do them determine the effectiveness and efficiency of cloud computing services. Task scheduling is essential to improving a cloud system’s efficacy and performance since it optimizes resource allocation and utilization. In this context, cloud computing offers a variety of advantages, including cost savings, security, mobility, flexibility, disaster recovery, quality control, automated software updates, and sustainability. Therefore, the requirement to control resource allocation has increased along with the number of cloud users. However, cloud task scheduling requires a quick and intelligent algorithm that can identify available resources and plan out tasks that different people desire. Therefore, a quick, effective work scheduling method is needed for improved resource allocation and scheduling. Grey Wolf Job Optimization (GWO) and Cuckoo Search Optimization (CSO) are used for the Optimized Efficient Job Scheduling Resource (OEJSR). The "grey wolf optimization" (GWO) ensemble with OEJSR has provided the best resource allocation models. The prior research was compared using computation time, make span, iteration-based performance, fitness, and success rate. Studies demonstrate the superiority of the suggested approach.

Keywords:

• Cloud computing
• resource allocation
• scheduling
• cuckoo
• grey wolf

REVIEWING EDITOR:

• Swadesh Kumar Singh, Gokaraju Rangaraju Institute of Engineering and Technology, India

SUBJECTS:

• computer science (general)
• computer engineering
• computing & IT security
• internet & Multimedia - Computing & IT
• legal, ethical & social aspects of IT

1. Introduction

Software as a service (SaaS), platform as a service (PaaS), and infrastructure as a service (IaaS) are three examples of the services that cloud computing offers as part of its Service Oriented Architecture (SOA) (Choudhary & Vithayathil, 2015). Under the pay-per-use model, users may employ network services to access cloud resources. As a result, providers are working to manage Service Level Agreement (SLA) violations while optimising resources in accordance with client expectations. Managing unforeseen workloads becomes more challenging due to the increasing demand for cloud services. Allocating resources in accordance with workload is a challenge that affects Quality of Service (QoS) metrics. It thus requires a precise strategy that can support flexibility and scalability. Therefore, allocating the task to the resources appropriately will aid in their availability and utilization.

Networks for cloud computing, where all users have constant access to servers, are considered to be among the most important and well-liked distributed approaches (Moosavi et al., 2023). Resource management is one of the main issues with cloud settings. Distributing computation, storage, and power to a collection of applications is part of managing cloud resources to satisfy the demands of cloud users and cloud service providers (Harvinder et al., 2020). Customers who use the cloud want guarantees on the security of their data, the affordability of their transactions, the ability to preserve their privacy and other factors. Conversely, cloud service providers seek to maximize resource utilization, boost resource availability, and reduce energy usage (Singh et al., 2020).

The current literature review demonstrates that the growing demand for cloud services has made the issue of cloud resource management more challenging (Moosavi et al., 2023). Therefore, the research in this field inspired the creation of a unique method for resource matching and throughput utilisation. Furthermore, before assigning a job to a virtual machine (VM), the resource allocation strategy should be able to determine its precise needs. Such a plan will control resource availability, make it easier to prevent SLA breaches, and manage job deadlines (Harvinder et al., 2020). Thus, the scheduling issue in cloud computing may be solved by creating a common approach.

From the standpoint of quality of service, we have looked at a number of research projects pertaining to resource management. The proposal includes the following methodical steps to meet appropriate resources: demand identification criteria, initial matching generation, matching criteria that satisfy QoS requirements, selection of appropriate solutions, resource usability through appropriate distribution, performance analysis, and SLA management.

1.1. Problem statement

The process tasks must be planned in the cloud effectively, which requires taking into account both the tasks’ end dates and the dependencies of the process. Usually, the task scheduling approach is simple for independent tasks. However, because of the inter-dependencies between its activities, scheduling a workflow may be challenging (Yu & Wu, 2023). It is vital to provide enough resources for every work subsequent to its scheduling and prioritisation. Depending on how long a project takes to finish, each one has different processing requirements (Qiang, 2017). However, resource efficiency might drastically decline if load balancing is not adequately taken into account (Kun et al., 2011). Because of this, cloud service providers need to give priority to accurately and on-time scheduled work as well as sufficient resource supply (Torres et al., 2023). Due to their ease of construction, rule-based task scheduling algorithms make up the bulk of cloud computing algorithms. When it comes to the preparation of multidimensional tasks, rule-based algorithms perform poorly. In addition, scheduling and resource allocation are linked to quality of service (QoS) and may impact cloud service providers’ earnings in the long run. It is widely acknowledged that resource scheduling is among the most significant issues facing cloud computing, and researchers have access to a broad range of choices for this purpose.

1.2. Proposed model

The cloud computing work scheduling issue may be resolved using heuristic-based solutions or enumeration techniques (Kun et al., 2011). The scheduling of resources in cloud computing is an example of this kind of issue. Enumeration procedures have no place in the context of cloud computing since they generate all potential combinations of schedules for work before choosing the most productive one. Cloud computing makes this impossible. This approach is time-consuming, thus it shouldn’t be used in a cloud computing environment where a lot of work has to be done.

The main goal of this research is to quickly distribute all available cloud resources in line with the needs of cloud users’ jobs after conducting a thorough search and resource discovery. Here, we start by allocating the solution at random according to the quantity of tasks and cloud nodes. We use optimisation methods to change the answer after calculating fitness. The OEJSR approach employs the Cuckoo Search Optimization (CSO) process, Grey Wolf Job Optimisation (GWO) algorithm is used to distribute the resources discovered to be accessible. A list of the most significant improvements that this research has made is shown below.

1. To make cloud work scheduling easier, the Optimized Efficient work Scheduling Resource (OEJSR) was created. To aid in the scheduling process, we develop a multiobjective function for fitness in this.

2. To identify the optimal solution and mitigate some of the drawbacks of the individual Grey Wolf Job Optimization (GWO) and Cuckoo Search Optimization (CSO) procedures, the Cuckoo Search Optimisation (CSO) and GWO methods are hybridised.

3. This approach maintains quality standards and timeline limits while cutting down on the overall time it takes to complete the work.

The following is the ongoing work. Significant studies are included in Section 2. The suggested strategy is described in the third section. Section four talks about the outcomes of the simulation and section five comes with conclusion.

2. Literature review

2.1. Background

One of the most amazing innovations that has drawn attention from engineers all around the world is cloud computing. Although cloud computing offers numerous advantages, there are also many risks associated with it that should not be disregarded in terms of security, scalability, latency delay, and so on. A successful adoption of cloud computing in an organisation requires appropriate planning and understanding with the opportunities, risks, vulnerabilities, and possible solutions that may arise (Tripathi et al., 2020; Masadeh et al., 2019). Determining the optimal layout principles to increase cloud security has thus become crucial for all cloud jobs. Because of a writing audit, we are investigating and surveying the most important information security and organisation security bets on cloud frameworks in this examination. Virtualization adds the product to the organisational structure, which may damage security when improperly created and transmitted. This is evident from a deeper look at the virtual environments that many organisations have pushed and marketed as the solution to current security challenges (Than & Thein, 2020; Erdei & Toka, 2023; Srimoyee et al., 2020). Prior research has shown that cloud networks have certain security, latency, and delay problems. These problems persist in this study subject, along with a few additional problems.

In order to satisfy the correspondence needs of the photoelectric hybrids network architecture, an analysis in light of big data and computing cloud innovation is recommended. This study’s primary objective is to examine innovations in cloud computing and big data. It avoids using data structures and other methods, explores specialised features, uses topological optical connections, and ultimately creates an exploration approach for large data and cloud computing. According to the findings of the exploratory study, the optical links have loads of 60, 50, and 20, respectively. At the moment when loads of optical links start to become less substantial, Hub B reaches the routes of the six goal nodes. Through the use of optical connections and the ability to regulate the scope of neighbouring connections, the photovoltaic hybrid network structure enables new forms of communication. It is possible to look at the problems with the photoelectric hybrid structure of networks in terms of big data innovation and cloud computing.

By consolidating allocation choices for the best applicant and optimising communication latencies, Masoud et al. (Harvinder & Anshu, 2017) introduced an energy-efficient resource allocation technique that has lowered energy usage and enhanced performance. An optimisation technique for resource allocation—especially for educational purposes—was proposed by Fernando et al. (Harvinder et al., 2019a). It has cut costs and allocated duty dynamically (Harvinder et al., 2019a).

A resource provisioning approach, was introduced by Singh et al. (Harvinder et al., 2019b) while using an optimiation mechanism for suitable resource mapping. For workload identification and categorization, the K-means clustering algorithm and PSO are used; this reduces cost, time, and energy while allocating the appropriate resource (Harvinder et al., 2019b).

A VM allocation approach was published by Zhang et al. (Masoud & Ronak, 2016). It used a resource reservation-based system that effectively carried out VM-to-PM mapping. It used resources more efficiently and saved more energy (Masoud & Ronak, 2016).

In order to get cost-effective outcomes, Allah et al. (Fernando et al., 2016) proposed a resource allocation strategy that hybridises genetic and cuckoo search algorithms. Cuckoo algorithm using genetic operators were used in a rigorous approach to identify the appropriate solution (Fernando et al., 2016).

A “Swarm Intelligence Based Prediction Approach” (SIBPA) for controlling the accuracy and reaction time of cloud resources was described by Kholidy (Sukhpal et al., 2018). It enhanced the dynamic behaviour of customer requests and the throughput utilisation of CPU, memory, and disc storage (Sukhpal et al., 2018). An energy-aware use of resources method that effectively controls workload and power consumption was presented by Than et al. (2020). It lowers data centre power consumption and assesses energy utilisation in resource allocation (Javadpour et al., 2023).

Using a dynamic thresholding technique, Bhattacherjee et al. (Zhang et al., 2019) presented VM placement and migration strategy. Markov chain has been used to forecast needs in the future by using past data. It lowers the energy use of data centre resources (Zhang et al., 2019).

The cost-based job scheduling (CJS) technique is suggested by Mansouri et al. (Min et al., 2019) as a means of resource optimisation. It identified the cost of calculation and communication and responded in an efficient and economical way (Min et al., 2019). Profit and availability of resources aware methods that controlled accuracy and complexity were introduced by Reshmi et al. (Kholidy, 2020). It grouped related jobs and used a hybrid cuckoos search algorithm to allocate the best resource (Kholidy, 2020).

The "Modified Dragonfly Algorithm," which maximises virtual machine placement to handle resource distribution problems, was suggested by Tripathi et al. (Than & Thein, 2020). The dynamic deployment of virtual machines results in improved resource utilization (Than & Thein, 2020).

Resource utilization increased as an alternate method of resource delivery (Erdei & Toka, 2023). Nevertheless, admission control techniques were absent from the multi-tiered cloud systems that were detailed in Srimoyee et al. (2020) based on Y. According to Ling et al. (Mansouri & Javidi, 2020), the concept of cloud virtualization technology allows users to rent out computer resources to companies or private individuals. The fault-tolerance strategy should be used to manage cloud concerns since they reduce performance (Masadeh et al., 2019). According to Tripathi et al. (2020) speculates that the issue may have been caused by malfunctioning hardware, a damaged virtual machine, network congestion, or a malfunctioning programme. The exploration and exploitation stages of GWO are integrated with the crossover and mutation processes of GA in a proposed Hybrid GGWO method optimisation (Reshmi & Poongodi, 2020). This research shows that GGWO outperforms other optimization algorithms with regard to of processing time and the root mean squared error (RMSE) due to its excellent accuracy.

2.2. Contribution to the research work

The optimise scheduling and resource allocation, increase virtual machine availability. Additionally, dynamic workload identification is needed to allocate resources appropriately based on client demand. Thus, excellent matching is required for dynamic task allocation and scheduling. Furthermore, it was challenging to calculate cost consumption and measure cloud resource utilisation. Therefore, the objective function’s design is to assess resource use and cost. The optimisation parameters that are often used to calculate cost and utilisation are also helpful in determining makespan and energy usage (Singh et al., 2020; Yu & Wu, 2023). Consequently, by satisfying QoS requirements, the defined objective function satisfies the requirements of the proposed task.

3. Proposed system

This section presents the methodically stated suggested QoS based Optimized Efficient Job Scheduling Resource (OEJSR) algorithm. The Cuckoo and Grey Wolf Job Optimisation approach has been used for scheduling and resource allocation issues. The process of negotiating has highlighted the best fit. The OEJSR was developed methodically: First, identify the task requirements. Next, start the negotiation process. Fourth, generate feasible solutions. Fifth, evaluate the solutions based on QoS validation. Sixth, design the mapping based on the solutions. Seventh, allocate resources and schedule them. Eighth, execute the task and release the resources.

The suggested OEJSR code functions in this way: The input of user tasks initiates the algorithmic procedure. Subsequently, the jobs are verified by the system analyzer and approved against the competence list. When a resource is idle or free, it is first checked for availability and added to a list. The negotiating process was started by many ant colony systems depending on how many resources were available in the availability list. It looks at every potential outcome to provide an initial answer. Solutions are still being generated by the process up to the termination criteria. For assessment, these solutions are kept in a temporary list. Additionally, a QoS-based evaluation finds an acceptable option and forwards the others to the evaporation process.

These standardised solutions are able to do tasks in a timely and cost-effective manner. The mapping was carried out using the system’s optimum solution. Moreover, virtual mapping has been followed in the deployment of VMs. An optimum schedule is defined as the best outcome among all possible options. Then, with the ideal workload allocated to the resources, the task execution procedure began. After a job is completed, resources are made available. Until users complete their duties and the process is allocated, this will continue.

3.1. Scheduling of job in cloud computing

You may access cloud-based assets from a variety of websites. They will not be placed in the same area. The user-supplied query could be related to several jobs, It may need a distinct set of resources to finish. Jobs are assigned to the suitable assets by the task scheduler. The cloud-based work scheduling architecture is shown in Figure 1.

Figure 1. Architecture of job scheduling in cloud.

"U1, U2,… Un" in Figure 1 denotes the users; "W1, W2,… Wn" denotes the user-submitted works or queries; "J1, J2,… Jm" indicates the tasks required to finish every task or query; and "R1, R2,… Rp" indicates the resources required to finish each work or query. Take a look at this example: When a person searches for jobs, they might come across a variety of professions. Every assignment will need a different set of resources to complete. Anywhere in the cloud might be where the resources are kept. The task must be assigned to the appropriate resource by the job scheduler. Appropriate work scheduling in a cloud computation environment requires a good method. A new resource finding and allocation strategy is presented and shown in Figure 2 for scheduling cloud-based workloads. Based on questions provided by the user, the Optimised Efficient Job Scheduling Resource (OEJSR) algorithm produces first answers. Finally, each solution is assessed using the Grey Wolf Job Optimisation (GWO) algorithms and Cuckoo Search Optimisation (CSO). The optimal solutions are finally obtained using the OEJSR method. Once OEJSR has completed its search and identified the best resource to carry out the assigned work, task scheduling is done using the optimal solution at the end.

Figure 2. Proposed diagram of OEJSR method.

As seen in Figure 2, the hybrid OEJSR algorithm’s early solutions are first developed using the user-supplied queries. The algorithms Grey Wolf Job Optimisation (GWO) and Cuckoo Search Optimisation (CSO) are then used to process the top OEJSR solutions.

The first answers for processing the combined OEJSR algorithm are produced based on the user’s queries. When a user makes a query to have a job accomplished, it is finished according to the user’s instructions. The query will do many tasks, each requiring a distinct set of resources. It must be acknowledged that no resource is able to do every duty at once. While several resources are capable of doing the same work, they will vary in terms of efficacy and efficiency. Table 1 displays the execution time as well as resource quality for a selection of tasks.

Table 1. Time as well as resource quality samples for completing the task.

	R1	R2	R3	–	Rp
J1	(8,3)	–	(4,1)	–	–
J2	–	(6,1)	–	–	–
J3	–	–	–	–	–
J4	(3,1)	–	–	–	–
	–	–	–	–	–
	–	–	–	–	–
JM	(2,1)	–	–	–	–

Table 1 explains the following: Job J1 may be completed by either resource, R1 or R3, however, R1 needs nine seconds and an average of four, whereas R3 needs five seconds along with a value of two. As the resource R2 is has an overall score of 2, it will take seven seconds to finish job J2. The right tools will also be used to complete the remaining tasks. Table 1 is used to construct initial solutions according to the tasks’ execution capability that the user has supplied. Examine the following initiatives or queries that were filed by U1, U2, and UN: $$\begin{array}{c}W1=\{J_1,J_2,J_4\}\\ W2=\{J_1,J_2,J_3\}\\ W3=\{J_1,J_3,J_4,J_M)\end{array}$$

The equation that was recently presented states that the jobs {J1, J2, and J4} must be completed to respond to the user’s first inquiry; similarly, the jobs {J1, J2 and J3} must be completed to respond to the user’s second question; and finally, the jobs {J1, J3, J4 and JM} must be completed in order to answer the user’s Nth query. A result that was generated by using these questions submitted by users.

Fitness evaluation. Every generated solution is assessed according to how well the resource fits the job at hand. A person’s degree of physical fitness is determined by taking into consideration a number of elements, such as the overall amount of work that has to be done, the time needed to complete the jobs, and the quality of the tools used to carry out the jobs.

3.2. Cuckoo search algorithm

The cuckoo search (CS) is a population-based metaheuristic probabilistic global search strategy. Cuckoo eggs are used as a stand-in for potential solutions in computer science algorithms. Three rules serve as examples of the CS algorithm and are as follows: Each cuckoo deposits an egg in a randomly selected nest, one at a time. The next generation will inherit the cosiest nest full with the finest eggs, or solutions. Every host has a probability of finding at least a single alien egg and the number of host nests that are now accessible is limited. If a host discovers a cuckoo egg, it may destroy it or give it to an invasive species rather than store it in its own nest. $$\text{Pa of n x}\left(\mathrm{t}+1\right)$$

To make things easier, the final hypothesis may be approximated by the proportion of Pa of n nests replaced with new nests containing random answers. There may be an inverse relationship between an objective function and the fitness of an optimising solution. It is possible to build several fitness categories, much as with other evolutionary algorithms. Cuckoo eggs are a symbol for new ideas. The objective is to swap out inferior nest solutions with novel options (cuckoos). To come up with new approaches for x(t + 1) cuckoo i, a Levy flight is used. $$x_i^{(t+1)}=x_i^t+\alpha \oplus \mathit{\text{Levy}}(\lambda )$$

α(α > 0) symbolises a scaling size step and parameter should be related in some way to the size of the problem the algorithm is trying to solve. It is feasible to set to 1 or another constant in the great majority of cases. A "product" is the result of merging entries one by one. A distribution where the standard deviation is always variable and the mean is infinite is used to determine the random step’s duration $$\mathit{\text{Levy}}(u)=t^{-\lambda }$$

0 <λ < 3, In this case, the cuckoo’s sequence of steps—or hops—resembles an arbitrary walking process with a heavy tail that follows an exponential step-length distribution.

Algorithm 1:

 

Cuckoo search Algorithm Initialization $$\begin{array}{c}\mathit{\text{Objective}}->\mathit{\text{function}}[f(x),x={(x_1,x_2,\dots .x_d)}^T\\ \mathit{\text{Generation}}->t=1\\ \mathit{\text{Initial}}-\mathit{\text{population}}-\mathit{\text{of}}-\mathit{\text{host}}-\mathit{\text{nests}}->x_i(i=1,2,\dots n)\end{array}$$

While (t < (Maximum Generation) or (stop criterion))

Get a cuckoo (say i) randomly by Levy flights

Evaluate fitness for cuckoo F

Choose a nest among n (say j) randomly

If (Fi > Fj) then

Replace j by the new solution

End if

Abandon a function (Pa) of worse nests and build new ones

Keep the best solutions(or nests with quality solution)

Rank the solution and find the current best

Update the generation number t = t + 1

End While

3.3. Grey Wolf Optimization algorithm

The swarm-based algorithm known as GWO was developed at the University of California, Berkeley. Fortunately, its interactions with others in the wild is modelled like that of grey wolf packs in their native habitat. Wolves’ tracking and hunting of their prey is an example of their pursuit of the best possible outcome. In their natural environment, grey wolves tend to gather in packs. Typically, wolf packs consist of five to twelve members. Furthermore, to facilitate hunting, Four groups are formed from the wolves in the pack according to their social status. The initials of the groupings are as follows: The Alpha (α), a canine leader that may be either male or female, is in charge of setting the group’s rules on waking up, hunting, and sleeping. The second level of wolves is called beta (β), and it consists of both male and female wolves that help the other wolves in the pack make decisions and make decisions as a group. The third rank, Delta (δ), is in charge of many significant duties, pack leader, including caretaker, sentinel, and hunter. Reaching Omega (ω) is the ultimate and most challenging stage. In the hierarchical model, this level serves as a convenient target and is subject to the orders of superiors, although being the weakest.

The Grey Wolf Optimization (GWO) algorithm’s mathematical model. In addition to having a similar hunting approach, grey wolves also have a social hierarchy. It is divided into four tiers.

• Level 1: Alpha (α): This level regulates decision-making (e.g. where to shoot, when to wake up, and where to sleep).

• Level 2: Beta (β): The contender has the best chance of taking over as the next leader and replacing the wolf. serves as a consultant or adviser for

• Level 3: Delta (δ): The wolves are informed of which ones have now gained their regard. They are trying to find wolves named X. They carry out a range of duties, such as those of scouts, guardians, seniors, group caretakers, and predators.

• Level 4: Omega (ω): They thought that 1/4 equal wolves were the thinnest wolves. They take on the role of the critic.

3.4. Mathematical model and algorithm

The following is an explanation of each of the three GWO elements that pertain to the model derived from mathematics. Hunting, encirclement, and attachment were a few actions that may take place in this circumstance. The computation under surrounding behaviour is represented by the formulae below, which indicate surrounding behaviour. $$\begin{array}{c}D=C.\mathit{\text{Xp}}(t)-X(t)\\ X(t+1)=X(t0-A.D\end{array}$$ where Xp. indicates the prey’s position vector, X symbolizes the grey wolf’s location vector, t indicates the repetition that already exists, and A and C imply two fixed matrix vectors. Below is a description of both of the vectors (A and C): $$\begin{array}{c}A=2d.r_1-d\\ C=2d.r_2\end{array}$$

A value decline from 2 to 0 may happen throughout iterations, and the vectors R1 and R2 are arbitrary. Second, grey wolves are able to identify their prey and pursue it; throughout this process, they often follow the dominant wolf. Additionally, the beta version plus estuary may take part in shooting with a restricted scope. Even though beta and delta wolves are well informed about potential food sources, alpha wolves are often regarded as the better choice. As a result, as shown by the computations below, the locations of the remaining wolves, including the last one, will be updated using the positions of the significant, beta, and delta wolves $$\begin{array}{c}x(t+1)=\frac{1}{3}X1+\frac{1}{3}X2+\frac{1}{3}X3\\ X1=X\alpha (t)-A_{1}D\alpha \\ X2=X\beta (t)-A_{2}D\beta \\ X3=X\delta (t)-A_{1}D\delta \\ D\alpha =\left|C1.X\alpha -X\right|\\ D\beta =\left|C2.X\beta -X\right|\\ D\delta =\left|C3.X\delta -X\right|\end{array}$$ where the Eqs represent C1, C2, and C3. It shows how variations in alpha, beta, as well as delta parameters over time impact the search wolf’s position movement throughout the search space. The last location would be an accidental point inside a circle made up of the search space’s alpha, beta, as well as delta values, which is quite notable. To put it simply, the other wolves randomly report their whereabouts near the kill, while the alpha, beta, as well as delta wolves determine the location of the target.

After one algorithm iteration, the remaining wolves in the packet will likewise pay heed to the top three ideal sites for the alpha, beta, and delta wolves, as shown in Figure 3. The answer (position) with the highest accurate classification is an alpha, which is followed by a beta and a delta. The classification method is qualified and validated at the conclusion of the process iteration, and the classifier’s accuracy is targeted for every subsection (solution) of the scenario medium. $$\mathit{\text{Fitness}}-\mathit{\text{function}}=W_1\ast \mathit{\text{Accuracy}}+W_2\ast \frac{1}{\mathit{\text{Number}}-\mathit{\text{of}}-\mathit{\text{selected }}\mathrm{\text{Re}}\mathit{\text{sources}}}$$

Figure 3. Makespan analysis for OEJSR method with existing system.

Each resource subset has a list of resources. The part of the set with the fewest structures will be selected if two subsets are identical in accuracy but vary in the whole of their attributes. Moreover, the criteria of (W1) and (W2) in the computation above are programmable, with the exception that (W1) is boosted by W2, the opposite Sum of Designated Structures, and by Correctness.

4. Experiment results discussion

This article examines and evaluates the recommended work scheduling technique’s efficacy. A workstation with an i7 CPU operating at 2.42 GHz, 16 GB of RAM, and a 64-bit edition of Windows 11 was used to evaluate the recommended method for scheduling tasks with Java (jdk 1.8) and CloudSim.

4.1. Makespan

Figure 3 will provide an example. The proposed OEJSR approach is evaluated against various job scheduling techniques in terms The duration required to complete the given tasks and the degree of fitness attained by altering the total number of iterations as well as the initial response given in reaction to the three different inputs.

Table 2, Makespan’s comparative examination of the OEJSR approach with other methods. The outcome demonstrates that the OEJSR method has done better than the other approaches in every way. For instance, the OEJSR approach took just 95.03 s to reply after 20 iterations, but the Makespan of the other methods, such as BAT, WBAT, Firefly, and BLA, were 111.48, 107.49, 103.46, and 99.26 s, respectively. Comparably, the OEJSR method possesses a Makespan of 94.72 s for 100 iterations, compared to 119.65, 109.48, 104.97, and 101.89 s for the other known approaches, such as BAT, WBAT, Firefly, and BLA.

Table 2. Makespan analysis using the current system using the OEJSR technique.

Iterations of the Jobs	BAT-Algorithm (Javadpour et al., 2023)	WBAT- Algorithm (Srimoyee et al., 2020)	Firefly-Algorithm (Tripathi et al., 2020)	Bees-Life-Algorithm (Masadeh et al., 2019)	OEJSR
20	111.48	107.49	103.46	99.26	95.03
40	114.86	108.39	102.87	100.45	96.26
60	113.89	107.16	103.97	99.57	95.73
80	115.75	108.95	104.54	101.57	94.16
100	119.65	109.48	104.97	101.89	94.72

4.2. Computation time

Figure 4 shows the OEJSR methodology’s calculation time analysis in comparison to other approaches. It is evident from the statistics that the OEJSR method has fared better than the other approaches in every way. For instance, the OEJSR method responded in 216.65 ms after 20 iterations, but the computation times of the other methods now in use, such as BAT, WBAT, Firefly, and BLA, are 417.76, 336.75, 302.54, and 263.76 ms, respectively. Comparably, the OEJSR approach computes 251.54 ms for 100 iterations, compared to 458.17 ms for BAT, WBAT, Firefly, and BLA, 392.87 ms for 3 of those methods, 59.27 ms for 59.27 ms, and 289.43 ms for the other known techniques.

Figure 4. Computation time analysis for OEJSR method with existing system.

It is evident from the statistics that the OEJSR method has fared better than the other approaches in every way. For instance, the OEJSR method took just 7.268 s to reply after 20 iterations, but the fitness values of the other methods now in use, such as BAT, WBAT, Firefly, and BLA, are 13.764, 11.437, 10.873, and 8.517 s, respectively. Comparably, the OEJSR approach has a Fitness of 8.248 seconds for 100 iterations, compared to 13.378, 13.427, 11.864, and 10.817 s for the other known strategies, such as BAT, WBAT, Firefly, and BLA.

The OEJSR methodology’s success rate is shown alongside the existing approaches. The results show that the recommended method outperforms the alternatives in every way. For instance, the OEJSR approach has a success rate of 95.63% after 20 iterations, compared to 85.64%, 87.84%, 90.16%, and 93.15% for the other known approaches, such as BAT, WBAT, Firefly, and BLA. Comparably, after 100 iterations, the suggested approach has a 97.17% success rate, compared to 86.93%, 89.72%, 92.34%, and 94.89% for the three known methods—BAT, WBAT, Firefly, and BLA. This demonstrates the improved performance and better success rate of the OEJSR approach.

5. Conclusion

Allocating cloud resources is a real-time issue that may be effectively resolved to lower execution costs and increase resource use. When resource allocation complies with demand constraints, consumers’ expectations may be fulfilled in terms of resource utilization. Task scheduling is an NP-hard issue where improper matching results in a drop in performance and a breach of the service level agreement. Managing unforeseen workloads becomes more challenging due to the increasing demand for cloud services. Allocating resources by workload is a challenge that affects the Quality of Service metrics. It thus requires a precise strategy that can support flexibility and scalability. Therefore, allocating the right amount of work to each resource will also aid in its availability and use. In this study, we suggest a hybrid approach to task scheduling that brings together the benefits of the Grey Wolf job optimisation algorithm with the Cuckoo search technique. We have assessed the performance of these algorithms using many measures, and the combination of them produces the best results. To enable the OEJSR algorithm to perform as intended, the freshly created preliminary solutions are fed into it. The OEJSR algorithm optimises work scheduling by considering the quality of presently available resources as well as execution time. After that, the previously scheduled tasks are completed, and the result relevant to the question submitted by the customer is shown to them. The suggested OEJSR technique is assessed and contrasted with other comparable cloud resource allocation efforts. The assessment considers both the overall time required to do the assigned activities and the achieved level of fitness. The equipment was inadequate, but the suggested approach outperformed the rival algorithms in terms of general health and the time required to complete the tasks given. To highlight OEJSR's success, a number of pertinent indicators are collected and compared with those of comparable efforts. To determine if the suggested approach can be used to tackle a range of complex, real-world optimisation issues, further investigation is needed.

Author contributions

Resources: R.V.S.S.S.Nagini, Prasanthi Gottumukkala. Software: Sanjeev Kumar Shah. Writing – original draft: R.V.S.S.S.Nagini. Writing – review & editing: Prasanthi Gottumukkala, Navdeep Singh, Sanjeev Kumar Shah.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

Yes, on your request data will be provided.

Additional information

Funding

This Research received no external funding.