Optimasi Mekanisme Penjadwalan Job Offloading untuk Peningkatan Kinerja Proses Eksekusi Job di Lingkungan Fog Computing Dengan Parallel Asyncrhonous Simulated Annealing

Yasin, Moch (2025) Optimasi Mekanisme Penjadwalan Job Offloading untuk Peningkatan Kinerja Proses Eksekusi Job di Lingkungan Fog Computing Dengan Parallel Asyncrhonous Simulated Annealing. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

Text 05111860010010-Doktoral.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Meningkatnya kompleksitas aplikasi seluler memerlukan daya komputasi yang substansial. Namun, kemajuan dalam teknologi baterai masih tertinggal, sehingga membatasi kemampuan perangkat seluler untuk menjalankan tugas secara lokal. Untuk mengatasi keterbatasan ini, task offloading, yang mendelegasikan tugas komputasi ke server fog atau cloud, telah muncul sebagai solusi penting. Pada kondisi dimana terdapat satu jaringan yang dipakai bersama oleh banyak perangkat, terjadi antrian yang menyebabkan konflik jaringan dan pengiriman ulang paket data yang menyebabkan penambahan delay, peningkatan konsumsi energi dan kegagalan eksekusi tugas. Beberapa penelitian terdahulu banyak focus pada penjadwalan urutan eksekusi offloading dan tempat eksekusi job. Namun demikian, hanya sedikit yang fokus pada dinamika jaringan yang menjadwalkan kapan tugas harus mulai dikirimkan ke cloud/fog. Penelitian ini mengusulkan penjadwalan yang terdiri dari dua Keputusan yaitu lokasi eksekusi job dan jadwal pengiriman data dengan satuan detik. Kami membandingkan pengaruh penjadwalan dan tanpa penjadwalan terhadap konsumsi energi pada perangkat mobile. Lebih jauh lagi, penulis mengusulkan algoritma untuk menyelesaikan permasalah ini dengan menggunakan algorithma heuristik berupa parallel simulated annealing with time arrival scheduling. Untuk meningkatkan objektifitas, algoritma PANTAS diuji sebanyak 1000 kali simulasi dengan 5 algorithma heuristik lain. Pada 5 skenario dengan kecepatan uplink 10 sampai 50mbps, didapatkan efisiensi waktu 54 persen dan efisiensi waktu -0,32 persen. Pada simulasi lanjutkan, algorithma PANTAS mampu meningkatkan rata-rata efisiensi energi hingga 10% dengan makespan 9% lebih rendah atau keandalan 9% lebih tinggi, mengungguli lima algoritma dasar lainnya dalam konsumsi energi sambil meminimalkan makespan dan memaksimalkan reliabilitas.
==================================================================================================================================
The increasing complexity of mobile applications requires substantial computational power. However, advancements in battery technology still lag behind, thereby limiting the ability of mobile devices to execute tasks locally. To overcome this limitation, task offloading, which delegates computational tasks to fog or cloud servers, has emerged as a crucial solution. In conditions where a single network is shared by many devices, queuing occurs, leading to network conflicts and the retransmission of data packets, which causes increased delay, higher energy consumption, and task execution failures. Several previous studies have focused extensively on scheduling the order of offloading execution and the job execution location. Nevertheless, few have focused on the network dynamics that schedule when a task should begin being transmitted to the cloud/fog/Fog. This research proposes a scheduling approach that consists of two decisions: the job execution location and the data transmission schedule, measured in seconds. We compare the effect of scheduling versus no scheduling on energy consumption in mobile devices. Furthermore, the authors propose an algorithm to solve this problem using a heuristic algorithm called parallel simulated annealing with time arrival scheduling (PANTAS). To enhance objectivity, the PANTAS algorithm is tested in 1000 simulations against 5 other heuristic algorithms. In 5 scenarios with uplink speeds ranging from 10 to 50mbps, a time efficiency of 54 percent and a time efficiency of -0.32 percent were obtained. In further simulations, the PANTAS algorithm was able to increase average energy efficiency by up to 10% with a makespan 9% lower or a reliability 9% higher, outperforming the five other baseline algorithms in energy consumption while minimizing makespan and maximizing reliability.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Cloud, Fog, Pembongkaran Komputasi, Keputusan, Waktu. Cloud, Fog, Computational Offloading, Decision, Time.
Subjects:	Q Science > QA Mathematics > QA76.9 Computer algorithms. Virtual Reality. Computer simulation. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.5956 Quality of service. Reliability Including network performance
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55001-(S3) PhD Thesis (Comp Science)
Depositing User:	Moch Yasin
Date Deposited:	07 Jan 2026 00:49
Last Modified:	07 Jan 2026 00:49
URI:	http://repository.its.ac.id/id/eprint/129303

Actions (login required)

View Item