Master’s Thesis at the College of Information Technology Discusses Adaptive Load Balancing in Distributed Servers

Duhaa Fadill Abbas
The Department of Information Networking at the College of Information Technology conducted the discussion of a Master’s thesis submitted by Ameer Akram Mousa Mohammed, entitled “Adaptive Load Balancing and Global Memory Coordination for Parallel File Processing in Distributed Servers.” The thesis was discussed on Sunday, December 21, 2025, under the supervision of Professor Dr. Mahdi Saleh Naama, in the presence of a number of faculty members and field specialists.

The thesis highlighted that load balancing is a fundamental technique in distributed systems and network infrastructures, designed to efficiently distribute incoming workloads or network traffic across multiple servers or resources through dynamic task allocation. Load balancing reduces response time and prevents any single server from becoming a bottleneck.
Furthermore, fault tolerance is an essential requirement in distributed systems, as failures in nodes, networks, or internal processes can lead to significant data loss, service interruptions, and overall system performance degradation. Traditional fault-tolerance methods provide only partial solutions and often require complex error recovery models and delayed responses while ensuring cost-effective resource utilization.

The proposed system in this thesis is based on the principle of relative load balancing (RLB) among server groups handling client requests, addressing the limitations of conventional static and dynamic load balancing methods such as Round Robin and Least Connection. Requests are divided proportionally according to the capacities of root server groups, thereby creating a more robust fault-tolerant system based on the principle of checkpoint replication.
The proposed architecture consists of a Master Server (control server) and Slave Servers (subordinate servers). The Backboned Server collects network information and identifies the best servers for processing requests, while multiple Root Servers share in handling client requests and returning results, ensuring efficient and reliable distributed processing.