PHD Research Proposal for Task Scheduling in Cloud Computing

Cloud computing [1] provides the resources for computing tasks on demand over the Internet in the remote Cloud. This computing has merits such as high computing power, low services cost, better performance, scalability, accessibility as well as availability. The data centers are massively scalable and can be ubiquitously accessed from any device, anywhere at any time all over the world. The efficient usage of the Cloud computing resources is based on the resource scheduling and task allocation. Task scheduling [2] is an essential and most important part in a cloud computing environment since it mainly focuses to enhance the efficient utilization of resources and hence a reduction in task completion time. It also considers other parameters in scheduling algorithms such as task completion cost. The main aim of the task scheduling algorithm is to improve the performance and quality of service and also maintaining the efficiency among the tasks and reduce the cost. Proper scheduling of the task plays a crucial role in the utilization of the resources. There are several types of task scheduling methods that works based on the priority of the task. They are:

Preemptive scheduling

In this type, the scheduling process takes place based on the priority of the task that is the currently running task gets interrupted when a new task arrives since the running task has a priority less than the new task. In this time the low priority suspends its execution, and high priority process starts its execution. It maximizes CPU utilization and throughput, and minimizes turnaround time, waiting time and response time.

Non-Preemptive scheduling

In this type, the scheduling process considers the arrival of the task and not the time that is the execution of the task takes place in the first come first serve manner. It means that the task completes its execution without waiting for the newly arriving task.

There are few more scheduling based on the usage of the preemptive and non-preemptive scheduling

First Come First Serve (FCFS)

  • In First Come First Serve process the oldest task in the queue is executed first at first as a result the waiting time is long. It schedules purely based on non-preemptive scheduling.

Shortest job first (SJF)

  • In First Come First Serve process the oldest task in the queue is executed first at first as a result the waiting time is long. It schedules purely based on non-preemptive scheduling.

Priority scheduling

  • In priority scheduling, the priority is assigned to all the task, and this priority is based on the CPU, memory usage or based on the choice of users. Here, the scheduling is based on the priority of the task. It also schedules using either the preemptive or non-preemptive scheduling.

Round Robin (RR) scheduling

  • The Round Robin scheduling is similar to the FCFS scheduling with the time-slicing and preemption technique. In this scheduling, the first task begins its execution based on FCFS and completes its process within its allocated time. Similarly, the next task is executed within the time, and the previous task waits for its turn to continue its execution.

There are specific task scheduling strategy for the effective utilization of the resources which are described as follows:

Several existing systems propose the task scheduling using the container, priority, and heuristic algorithm such as greedy algorithm, Bandwidth-Aware divisible Task Scheduling (BATS), Modified Analytic Hierarchy Process (MAHP), Longest Expected Processing Time preemption (LEPT), and divide and conquer algorithm. Here one of the existing algorithms uses the container-based task scheduling algorithms by using the containers in task processing. This approach either accepts or rejects the incoming task, and the accepted tasks are completed within the delay constraint. However, both the precedence of the task and load balancing should be considered in the workflow task. Similarly, a virtual machine power efficiency-aware greedy scheduling algorithm (VPEGS) [3] is used in few existing systems for estimating the task energy by considering factors, including task, resource demands, VM power efficiency, and server workload before scheduling tasks greedily. In some of the research work, the scheduling is based on the priority of the task. Here, the Deadline-Aware Priority Scheduling (DAPS) [4] is modeled for minimizing the average makespan and maximize resource utilization under deadline constraint. This model sorts the tasks in ascending order based on length priority and labeling the VM’s state as successful that achieves the deadline constraint and then mapping the tasks to the suitable VM takes minimum processing time. A few research works use the priority based scheduling [5] along with the combination of several algorithms such as MAHP, BATS, LEPT, divide-and-conquer for performing the effective task scheduling and resource allocation. Here, each task is processed before task scheduling using the MAHP algorithm and then based on the priorities of the virtual machine the tasks are mapped to them using the bipartite graphs.

In the dynamic scheduling, based on the user’s budget the resources are located and released several times, this increase time consumption. Hence, it results in the violation of the SLA and minimizes the service provider’s performance. Placing the resources in the distributed data center involves the exchange of the information; thus, Most of the researcher models focus their task scheduling in the single data center. However, it is necessary to consider the data transfer while performing resource scheduling. Thus, it is an essential one to define specific parameters before performing the task scheduling in the multiple data centers. During task scheduling in the various data centers, both the time and the cost are essential to include. Though most of the researchers focus on the energy, time and cost; they fail to contribute to the remaining metrics, for instance, storing, processing, and scheduling over multiple resources. Furthermore, few researchers perform the migration for balancing the overloaded resources; however, execution of migration increases the cost, power consumption, and makespan. Therefore, it is necessary to implement the effective resource allocation and task scheduling that helps to maintain the resources in the balanced state.

Reference:

  • [1] Ashraf, Imran, “An overview of service models of cloud computing,” International Journal of Multidisciplinary and Current Research, Vol.2, No.1, pp.779-783, 2014.

  • [2] Salot, Pinal, “A survey of various scheduling algorithm in cloud computing environment,” International Journal of Research in Engineering and Technology, Vol.2, No.2 pp.131-135, 2013.

  • [3] Lin, Weiwei, Wentai Wu, and James Z. Wang, “A heuristic task scheduling algorithm for heterogeneous virtual clusters,” Scientific Programming, 2016.

  • [4] Alworafi, Mokhtar A., and Suresha Mallappa, “An Enhanced Task Scheduling in Cloud Computing Based on Deadline-Aware Model,” International Journal of Grid and High Performance Computing (IJGHPC), Vol.10, No.1, pp.31-53, 2018.

  • [5] Gawali, Mahendra Bhatu, and Subhash K. Shinde, “Task scheduling and resource allocation in cloud computing using a heuristic approach,” Journal of Cloud Computing, Vol.7, No.1, pp.4, 2018.

Leave Comment

Your email address will not be published. Required fields are marked *

clear formSubmit