Source-linked AI summary
Efficient Optimal Algorithm of Task Scheduling in Cloud Computing Environment
Dr. Amit Agarwal, Saloni Jain
TL;DR
Cloud scheduling must improve execution and resource use in an environment with diverse tasks and virtual machines. The paper proposes a generalized priority algorithm that ranks tasks by size and VMs by MIPS, and reports better performance than FCFS and Round Robin in CloudSim experiments. The study is limited by its use of a limited number of tasks and identifies broader testing as future work.
Problem
Cloud computing faces scheduling challenges, and effective scheduling is important for executing computational tasks and improving resource utilization.
Method
The paper proposes generalized priority scheduling, comparing it with FCFS and Round Robin while ranking tasks by size and VMs by MIPS.
Results
The generalized priority policy is reported to perform better than FCFS and Round Robin scheduling.
Takeaways & Limitations
The reported comparison supports using task size and VM MIPS as the priority factors for the studied cloud-scheduling setting.
Takeaways & Limitations
The paper reports that its generalized priority algorithm was developed with limited tasks and proposes testing more tasks and reducing execution time in future work.
Abstract
from arXiv · showhide
Cloud computing is an emerging technology in distributed computing which facilitates pay per model as per user demand and requirement.Cloud consist of a collection of virtual machine which includes both computational and storage facility. The primary aim of cloud computing is to provide efficient access to remote and geographically distributed resources. Cloud is developing day by day and faces many challenges, one of them is scheduling. Scheduling refers to a set of policies to control the order of work to be performed by a computer system. A good scheduler adapts its scheduling strategy according to the changing environment and the type of task. In this research paper we presented a Generalized Priority algorithm for efficient execution of task and comparison with FCFS and Round Robin Scheduling. Algorithm should be tested in cloud Sim toolkit and result shows that it gives better performance compared to other traditional scheduling algorithm.
1. INTRODUCTION
Cloud computing provides subscription-based services through distributed infrastructure, platforms, and software, but its continued development faces challenges in effective task scheduling. Scheduling policies control work order and support resource utilization and system throughput.
- Cloud computing delivers IaaS, PaaS, and SaaS through subscription-based, pay-as-you-use services.IaaS provides virtual-machine infrastructure, PaaS provides an application development platform, and SaaS provides software applications.
- Scheduling is a major cloud-computing challenge that influences effective execution.
- Scheduling policies control work order while aiming for high performance, system throughput, and maximum resource utilization.The paper compares Time Shared, Space Shared, and generalized priority algorithms.
2. RELATED WORK
Prior work applies simulation, load balancing, optimization, priority scheduling, and cost constraints to improve cloud task execution and resource management. These approaches address performance, QoS, resource utilization, cost, and deadline-related objectives.
- CloudSim supports testing task-scheduling approaches and policies before applying them in real systems.
- Load-balancing research uses two-level scheduling to meet user requirements and improve resource utilization.The cited work also implements QoS-based meta-scheduling and backfilling for job dispatch.
- Genetic simulated annealing incorporates completion time, bandwidth, cost, distance, and reliability as QoS requirements.Annealing is applied after selection, crossover, and mutation to improve local search ability.
- Hierarchical scheduling targets service-level agreement response times by executing high-priority deadline-based jobs first.Priority jobs are spawned from remaining jobs using estimated completion time.
- Other studies prioritize tasks using activity-based costing, minimize workflow cost under deadlines, or balance VM requests by processing power.These approaches respectively address cost and performance measurement, execution cost and time, and weighted VM allocation.
- CloudSim-based comparisons evaluate FCFS, priority scheduling, and shortest-job-first policies under different conditions.
3. PROPOSED FRAME WORK AND METHODOLOGY
The proposed framework treats cloud scheduling as resource discovery, resource selection, and task submission, then compares FCFS, Round Robin, and a generalized priority algorithm. The priority strategy ranks tasks by size and VMs by MIPS, and is reported to outperform the two traditional policies.
- Proposed framework: Cloud resource scheduling is organized into resource discovery and filtering, resource selection, and task submission.The broker discovers resource status, selects a target using task and resource parameters, and submits the task.
- Compared algorithms: The study compares First Come First Serve, Round Robin, and generalized priority scheduling.
- First come first serve: FCFS selects a resource with the smallest waiting queue but is non-preemptive, so short tasks behind long tasks may wait.
- Round Robin: Round Robin gives queued jobs equal execution turns and returns unfinished jobs to the queue for later turns.The cited passage notes that heavy loads can make Round Robin take a long time to complete.
- Generalized priority algorithm: The generalized priority algorithm ranks tasks by size and virtual machines by MIPS value.The highest-size task receives the highest rank, while the highest-MIPS VM receives the highest rank.
- Generalized priority algorithm: The proposed priority policy is reported to perform better than FCFS and Round Robin scheduling.
- Generalized priority algorithm: In the example, VMs are ordered by computational power, with the 1000-MIPS VM receiving first preference.
Algorithm-
The algorithm creates and organizes virtual machines and cloudlets according to computational capacity, maintains VM allocation information, assigns tasks to suitable VMs, and updates available resources.
- Algorithm: The algorithm stores suitable virtual machines in a VM list while traversing adjacent unvisited vertices.
- Algorithm: VMs are created across datacenters according to host computational power, cost, processor speed, memory, and storage.
- Algorithm: Cloudlet lengths are allocated according to computational power and bound to VMs using length and MIPS.
- Algorithm: The VM load balancer maintains an index table of VMs and their current allocation status.
- Algorithm: The highest-length cloudlet is assigned to the highest-MIPS virtual machine.
- Algorithm: The datacenter broker sends each request to the identified VM and updates available resources afterward.
4. EXPERIMENT AND EVALUATION
The experiment verifies the scheduling algorithms in CloudSim using configured virtual machines and workload tasks, comparing FCFS, Round Robin, and Generalized Priority.
- Experimental Setup: The algorithms were tested with CloudSim 3.0.3 on an Intel Core i5 2.6 GHz computer running Windows 7.CloudSim supports modeling data centers, hosts, virtual machines, scheduling, and resource provisioning policies.
- Experimental Setup: The experiment created five virtual machines with 512 MB RAM and specified MIPS capacities.The reported VM configurations include MIPS values of 250, 1000, 250, 500, and 250, with another listing of 1000, 500, 250, 250, and 250.
- Workload: Twelve Cloudlet tasks were created with lengths ranging from 10000 to 20000.The task-length sequence contains both 10000 and 20000 values.
- Compared Algorithms: The evaluation compares FCFS, Round Robin, and Generalized Priority algorithms.These three algorithms are listed as the evaluated scheduling approaches.
FCFS VS ROUND ROBIN VS GENERALIZED
This section presents a comparison of FCFS, Round Robin, and Generalized Priority using execution time, data center ID, and virtual machine information.
- Algorithm Comparison: The comparison includes FCFS and Round Robin scheduling.The section labels identify both traditional scheduling algorithms.
- Evaluation Measures: Reported fields include execution time, data center ID, and virtual machine assignment.These fields are repeated for the compared algorithms.
5. CONCLUSION
The paper evaluates a Generalized Priority Algorithm against FCFS and Round Robin for cloud task scheduling and reports that the proposed algorithm is more efficient.
- Conclusion: The proposed scheduling algorithm is the Generalized Priority Algorithm, or GPA.The paper creates FCFS and Round Robin algorithms for comparison.
- Conclusion: The experiment varies the number of virtual machines and workload traces.These settings are used to compare GPA with FCFS and Round Robin.
- Conclusion: The reported result is that GPA is more efficient than FCFS and Round Robin.The passage states this as the outcome of the experiment.
6. FUTURE WORK
Future work extends the generalized priority algorithm beyond its current limited-task setting, examines execution time reduction, and applies it to grid environments.
- Future Work: The current algorithm is described as operating with limited task workloads.The authors state that future work will consider more tasks.
- Future Work: Future work will investigate whether execution time can be reduced.The passage identifies reducing execution time as a planned direction.
- Future Work: The authors plan to develop the algorithm for grid environments and observe time differences between cloud and grid.The intended comparison concerns execution time in the two environments.