Ciptaningtyas, Henning Titi (2024) Multi-objective Task Scheduling using Opposition-Based Learning and Nature Inspired Algorithms with Deadline Constraints in Cloud Resource Provisioning System. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 05111960010006-Dissertation.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
05111960010006-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (4MB) | Request a copy |
Abstract
Cloud computing is a model that enables network-based access to a collection of configurable computing resources. Scalability is a prominent advantage of cloud computing, enabling users to manage resources per user's requirements. To fulfill the requirements of both cloud providers and consumers, it is critical to implement a Cloud Resource Provisioning System (RPS) to provision multipurpose cloud resources. Metaheuristic algorithms, categorized as Evolutionary, Swarm Intelligence, and Physics-Based Algorithms, are designed to manage cloud task scheduling problems by generating near-optimal solutions within a reasonable time limit. This research presents an alternative approach to optimizing the cloud task scheduling by considering multiple objectives simultaneously (multiobjective). These objectives include makespan, throughput, and resource utilization, all within predetermined time constraints. This research stage is divided into two phases: dataset construction and cloud task scheduling optimization algorithm implementation. Dataset construction includes preprocessing real-world data (SDSC Blue Horizon logs) and creating synthetic datasets (simple random datasets and stratified random datasets). Evolutionary algorithms (Genetic Algorithm and a combination of Genetic Algorithm and Artificial Neural Network) and swarm intelligence algorithms (Particle Swarm Optimization, Squirrel Search Algorithm, and a combination of Squirrel Search Algorithm and Opposition Based Learning) are the optimization algorithms that have been implemented. The algorithms are executed in a simulation environment using CloudSIM 4.0. The cloud tasks in this experimentare independent. The performance metrics compared are makespan, average start time, average finish time, average execution time, total wait time, total scheduling length, throughput, resource utilization, energy consumption, and imbalance degree. The proposed algorithm achieves highly effective task scheduling outcomes in terms of preventing premature convergence and accomplishing multiple objectives. In the implementation of the evolutionary algorithm, GA-ANN outperforms GA when the quantity of cloud tasks is substantial and comprises a diverse range of task lengths. GA-ANN is beneficial for overcoming imbalance degrees and speeding up cloud simulation. In the implementation of the swarm intelligence algorithm, the SSA-OBL algorithm exhibits enhanced performance in terms of makespan and energy consumption. However, compared to the SSA algorithm, it demonstrates less efficiency across other parameters. Nonetheless, this SSA-OBL shows a gradual increase in value, indicating that this algorithm is adequate for predicting cloud performance as an increasing number of tasks are executed. The dissertation outputs are two international conference articles and two international journal articles: "Survey on Task Scheduling Methods in Cloud RPS System" - 23rd International Seminar on Intelligent Technology and Its Applications (ISITIA) 2022; "A Systematic Literature Review of Genetic Algorithm-Based Approaches for Cloud Task Scheduling” - 14th International Conference on Information & Communication Technology and Systems (ICTS) 2023); “Multi-objective Task Scheduling Algorithm in Cloud Computing Using Improved Squirrel Search Algorithm” - International Journal of Intelligent Engineering & Systems (IJIES) Vol.17, No.1, 2024 Vol.17, No.1, 2024; and "Addressing Data Imbalance in Cloud Provisioning: Enhancing Performance through Genetic Algorithms and Artificial Neural Networks" - Journal of Wireless Mobile Networks, Ubiquitous Computing, and Dependable Applications (JoWUA) 2023 (draft).
=================================================================================================================================
Komputasi berbasis awan (cloud computing) adalah model yang memungkinkan akses sekumpulan sumber daya komputasi yang dapat dikonfigurasi bersama melalui jaringan. Keunggulan yang menonjol dari penggunaan cloud computing adalah skalabilitas (scalability) yang memungkinkan pengguna mengatur sumber daya (resource) sesuai kebutuhan. Untuk memastikan ekspektasi pengguna cloud dan penyedia cloud terpenuhi, maka diperlukan penyediaan sumber daya cloud multi-tujuan menggunakan Cloud Resource Provisining System (RPS). Algoritma metaheuristik (yang terbagi menjadi Algoritma Evolusioner, Kecerdasan Kawanan, dan Berbasis Fisika) adalah algoritma yang memberikan solusi yang hampir optimal dalam waktu yang wajar untuk menangani permasalahan cloud task scheduling. Penelitian ini berusaha memberikan alternatif solusi optimasi cloud task scheduling dengan memperhatikan lebih dari satu tujuan (multiobjective), yaitu makespan, throughput, dan resource utilization dalam batasan tenggat waktu yang ditentukan. Tahapan penelitian ini terbagi menjadi 2 bagian, yaitu pembuatan dataset dan implementasi algoritma optimasi cloud task scheduling. Pembuatan dataset meliputi preprocessing data realworld (log SDSC Blue Horizon) dan pembuatan dataset sintetik (simple random dataset dan stratified random dataset). Algoritma optimasi yang diimplementasikan adalah algoritma evolusioner (Genetic Algorithm dan kombinasi Genetic Algorithm – Artificial Neural Network) serta kecerdasan kawanan (Particle Swarm Optimization, Squirrel Search Algorithm, dan kombinasi vi Squirrel Search Algorithm-Opposition Based Learning). Algoritma dijalankan di lingkungan simulasi menggunakan CloudSIM 4.0. Tugas cloud (cloud task) dalam percobaan ini adalah tugas independen. Metrik kinerja yang dibandingkan adalah makespan, average start time, average finish time, average execution time, total wait time, total scheduling length, throughput, resource utilization, energy consumption, dan imbalance degree. Hasil penjadwalan tugas dari algoritma yang diusulkan cukup efektif menghindari premature convergence dan memenuhi multi-tujuan. Pada implementasi algoritma evolusioner, GA-ANN mengungguli GA saat kuantitas dan variasi cloud task berjumlah banyak dan mempunyai variasi panjang task dengan rentang yang jauh. GA-ANN bermanfaat untuk mengatasi imbalance degree dan mempercepat simulasi cloud. Pada implementasi algoritma kecerdasan kawanan, algoritma SSA-OBL menunjukkan kinerja makespan dan energy consumption yang unggul, namun kurang efektif dibandingkan dengan algoritma SSA untuk parameter lain. Namun, kombinasi SSA-OBL ini menunjukkan pertambahan nilai secara gradual yang menjadikan algoritma ini cukup bagus untuk memprediksi kinerja cloud saat load task yang dijalankan semakin banyak. Luaran yang dihasilkan dari disertasi ini adalah publikasi 2 artikel seminar internasional (“Survey on Task Scheduling Methods in Cloud RPS System” - 23rd International Seminar on Intelligent Technology and Its Applications (ISITIA) 2022 dan “A Systematic Literature Review of Genetic Algorithm-Based Approaches for Cloud Task Scheduling” - 14th International Conference on Information & Communication Technology and System (ICTS) 2023), publikasi 1 artikel jurnal internasional (“Multi-objective Task Scheduling Algorithm in Cloud Computing Using Improved Squirrel Search Algorithm” - International Journal of Intelligent Engineering & Systems (IJIES) Vol.17, No.1, 2024 Vol.17, No.1, 2024), dan 1 draft artikel jurnal internasional (“Addressing Data Imbalance in Cloud Provisioning: Enhancing Performance through Genetic Algorithms and Artificial Neural Networks” - Journal of Wireless Mobile Networks, Ubiquitous Computing, and Dependable Applications (JoWUA) 2023).
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | Cloud Computing, Cloud Provisioning, Task Scheduling |
| Subjects: | T Technology > T Technology (General) > T57.62 Simulation |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55001-(S3) PhD Thesis |
| Depositing User: | Henning Titi Ciptaningtyas |
| Date Deposited: | 07 Feb 2024 07:01 |
| Last Modified: | 07 Feb 2024 07:01 |
| URI: | http://repository.its.ac.id/id/eprint/106507 |
Actions (login required)
 |
View Item |