Deadline guarantee for scientific workflow using dynamic scheduling algorithms on IaaS clouds

• Main Author: Alqaisy, Sarah Abdulrahman Shukur
• Format: Thesis
• Language: English
• Published: 2018
• Online Access: http://psasir.upm.edu.my/id/eprint/69013/1/FSKTM%202018%2046%20-%20IR.pdf; http://psasir.upm.edu.my/id/eprint/69013/

Many important scientific applications can be expressed as workflows, which describe the relationship between individual computational tasks and their input and output data in a declarative way. This enables workflows to be automatically adapted to run across different environments. For complex workflows, abstraction also helps scientists to express their workflows at a higher level without being concerned about the details of how individual jobs are invoked or how data is transferred between jobs. Also, large-scale applications expressed as the scientific workflows that are often grouped into ensembles of interrelated workflows ( J. Vöckler, G. Juve & G. B. Berriman, 2011),( M. Malawski, G. Juve, E. Deelman & J. Nabrzyski ,2015). Normally, commercial Cloud computing is rapidly becoming the target platform on which to preform scientific computation. Typically, the commercial Cloud services charge on the basis of the number of hours the resources (such as CPU, network bandwidth and amount of storage) are used. This charging model is referred to as pay-per use. The flexibility inherent in the elastic Cloud model, while powerful computing results in inefficient usage and high costs when inadequate scheduling and provisioning decisions are made. The problem of scientific workflow scheduling in Clouds requires an alternative scheduling approach in mapping tasks to resources while fulfilling the deadline in the workflow of Cloud computing. In this project, our objective achieved better performance in term of success rate when compared to existing scheduling algorithms. In terms of solving the problems we efficiently designed the workflows scheduling on dynamically provisioned Cloud resources, while reducing the computation complexity. Specifically, we enhance two scheduling algorithms Proportional Deadline Constrained (PDC) and Deadline Constrained Critical Path (DCCP) that address the workflow scheduling problem in Infrastructure as a Service (IaaS) Cloud. We will conduct a simulation experiment of scientific workflow algorithms with both of the algorithms mentioned above. In addition, the simulation will be managed under deadline constraints on Infrastructure as a Service (IaaS) Clouds. The performance of the scientific workflow will be based on measuring Success Rate (SR) and Throughput. Finally, we expected the scheduling algorithms PDC and DCCP to be improve the Cloud resource usage high efficiency in the IaaS Cloud with efficient scheduling for scientific workflow.