Efficient implementation of hierarchical resource control for multi-agent systems

Abstract

Development of the World Wide Web makes it possible for multiple computers to work together in order to solve problems and make the most efficient use of resources. A distributed system is composed of such computers which are separately located and connected with each other through a network. One paradigm for computation using distributed systems is the multi-agent systems, in which many autonomous agents interact with each other to solve problems. The agents in a multi-agent system may be distributed on different computers (or nodes), where each computer owns its resources. Although the resources in a multi-agent system are connected by a network through which mobile agents can migrate for accessing sufficient resources, how to share these independently owned resources in both an effective and an efficient way is not fully understood. A key challenge in multi-agent systems is how to account for and control the resources which are located on individual nodes. The CyberOrgs model offers one approach to manage resources among competitive or collaborative agents by organizing computations and resources in a hierarchy. A cyberorg encapsulates agents and resources in a boundary and distributes the resources available to it within this boundary. A cyberorg contained in another cyberorg has a contract with the outer cyberorg, according to which it receives resources that it may use. A cyberorg also encapsulates an amount of the eCash, which is the currency for purchasing resources from its host cyberorg. Therefore, cyberorgs have a hierarchical structure in which resources are delivered to computations by a process where resources flow down from the root to the leaves of the hierarchy and the eCash flows up from the leaves toward the root. However, the hierarchical structure of the CyberOrgs model presents challenges in scalability. As a result, efficiency is an important concern in the implementation of CyberOrgs. In this thesis, an efficient implementation of the CyberOrgs model is described. System design, APIs of the implementation, example applications, experimental results, and future directions are presented.