Multi-objective scientific workflow scheduling algorithm in multi-cloud environment for satisfying QoS requirements
Cloud computing is a high-performance distributed computing platform that integrates large-scale services. It facilitates many scientific and engineering, as well as business workflow applications. However, current workflow applications come with various Quality of Service (QoS) objectives and const...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/113146/1/113146.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Cloud computing is a high-performance distributed computing platform that integrates large-scale services. It facilitates many scientific and engineering, as well as business workflow applications. However, current workflow applications come with various Quality of Service (QoS) objectives and constraints, such as makespan, cost, reliability, resource utilization and security, which pose serious QoS management challenges with respect to satisfying the objectives under specific constraints. In addition, the cloud environment is complex, highly uncertain with chances of failures at all levels (human, software, hardware, security). Therefore, one of the major concerns of users is getting assurance of the needed QoS for their applications, especially in tight cases.
These have also led another issue in scheduling workflow for cloud computing which are minimizing workflow makespan and cost simultaneously while satisfying the reliability constraint, improving overall QoS satisfaction, as well as increasing the reliability and minimizing completion time of the scheduled process with fault-intrusion tolerance.
There are three (3) main objectives laid out in this thesis, to tackle these issues. First, to propose a multi-objective and reliability constraint handling algorithm (FR-MOS) that controls the reliability constraint by determining the reliability constraint coefficient according to the value of the resource utilization. Second, to propose a minimum-weight-based multi-objective algorithm (MOS-MWO), which is based on Particle Swarm Optimization (PSO) technique and a novel minimum weight optimization approach, that improves user’s QoS satisfaction. Third, to propose a fault-intrusion-tolerant algorithm (FITSW), which is based on both fault and intrusion-tolerant techniques, to decrease the adverse impact caused by different faults (accidental and malicious) in cloud computing
systems. All the proposed algorithms are simulated using the popular cloud simulator, Workflowsim 1.0.
Results of the experiments prove that the multi-objective and reliability constraint handling (FR-MOS) algorithm significantly minimizes the makespan by 9% and cost by 10% compared to the benchmark algorithm under the reliability constraint. This was accomplished by determining the value of the reliability constraint coefficient based on the resource utilization of each alternative and selecting the best results from various alternatives with several reliability constraints. Moreover, the improvements of different QoS metrics values achieved by using a minimum-weight-based multi-objective algorithm (MOS-MWO) for scheduling scientific workflows are better than those of the previous work which used the Pareto optimization method. MOS-MWO can thus be applied in cloud-based applications to effectively schedule workflow while achieving significant improvement in the QoS satisfaction rate (QSR) to 4.8% compared with the multi-objective scheduling algorithm (MOS). The average of different workflows objectives shows that MOS-MWO algorithm yields better makespan compared with the MOS algorithm. With the MOS-MWO algorithm, makespan is reduced by 40%, cost also reduced by 3 % and risk probability reduced by 86%. MOS-MWO increases the resource utilization by 15% than MOS, and the reliability increase by 2%. Finally, the workflow completion time of the fault-intrusion-tolerant and deadline-aware algorithm (FITSW) decreased by 15% for all datasets when compared with the previous work, and the intrusion tolerance increased due to the high success rate of workflow execution. |
|---|