Network characteristics emerging from agent interactions in balanced distributed system

Area of Science:

• Computer Science
• Network Engineering
• Distributed Systems

Background:

• Distributed computing systems (DCSs) function as complex networks with essential information exchanges and job migrations between nodes.
• Software agents in DCSs mimic social network members, cooperating and competing for resources and knowledge.

Purpose of the Study:

• To evaluate the performance of different load balancing strategies across various network structures in DCSs.
• To analyze how information distribution and network topology influence the efficiency of load balancing strategies.

Main Methods:

• Simulation of four distinct load balancing strategies on diverse network structures.
• Analysis of system steady-state achievement, indicated by constant mean job response times.

Main Results:

• All tested strategies successfully achieved even load distribution and system steady states.
• Strategy performance was demonstrably affected by network structural parameters and the spread of information.

Conclusions:

• Load balancing strategy behavior is contingent upon network topology and information dynamics.
• Findings offer insights for understanding and managing complex systems, including DCSs.