Studying of Cloud Computing Scheduling

Abstract

The increasing need to provide a lot of services to users in daily life and various sectors like commercial, scientific and educational through internet led to the expansion for the use of cloud processing and virtualization technology. Moreover, cloud environment must insure on-demand service availability by optimizing cloud processing. Unprofitable resource is a type of resource misuse. The complexity of such environment can lead to resource misuse. Such resource misuse happened during job assignment of tasks on resources. Resource management is a key solution to overcome resource misuse. Resource management can be defined as the efficient way of resources organization to be able to use or utilize any resource when needed. Many algorithms are developed to optimize cloudlet distribution on virtual machines (VMs) in cloud processing environment such as First Come First Serve, Min-Max, Shortest Job First, and more. The problem of task distribution optimization is complex problem. For SJF Longer processes will cause more waiting time for cloudlets that waiting in the cloudlet queue, so it will produce low performance job distribution algorithm. So, we have to find a way to reduce waiting time in order to enhance task schedule performance, this will enhance the overall performance. To optimize cloud processing, it is necessary to optimize resource utilization including task scheduling and resource management. Many researchers discuss various algorithms to optimize both resource management and task scheduling. Task scheduling is the operation of distributing the tasks required to be processed on suitable idle virtual machines (resource) to process them conserving efficiency of cloud processing environment. To optimize scheduling operation, we must consider minimizing: Waiting time (the time between task arrival to cloud computing system and the starting time of processing - m sec),Response time (the time between starting and first output time - m sec),Run time (the time between starting and finish time - m sec),Turnaround time (the time between task arrival time and task finish processing time - m sec),Finish time (the time between starting and end processing time - m sec) and Throughput (maximizing) (average number of tasks finished per unit time task/m Sec.).

In order to validate our enhanced JSF we used a cloud simulator to simulate virtual cloud processing environment. This virtual environment consists of as many as required of virtual machines VMs to be used for processing simulation and we create a random length cloudlet generator to produce required cloudlet set for test. A final report is generated by cloud simulator include 4 (throughput, waiting time, turnaround time and finish time) parameters go compare between traditional SJF and enhanced algorithm. The proposed algorithm satisfies the above constraints that enhance cloud-processing performance. The proposed algorithm is modified Shortest Job First (SJF) scheduling algorithm where it solves the problem resource misuse by avoiding the idle VMs during the processing of all tasks. The algorithm idea based on detecting the first virtual machine that finishes all tasks assigned to it, in other words its state turns to idle. As soon as the algorithm detects first idle VM, it starts collecting all unprocessed tasks including tasks under execution. Then the algorithm redistributes all unprocessed tasks on all VMs. By this modified algorithm we achieved more significant enhancement in waiting time (measured in m Sec), turnaround time (measured in m Sec) and throughput (measured in task/m Sec.). Two of the parameters that can monitor the amount of enhancement are Throughput that enhanced by value between: 35% to 100%, Finish Time that enhanced by value between: 26% to 52%,