Autonomic group Computing in Peer-to-Peer Systems

点击翻译按钮获取中文摘要

The architecture of information system has been changing from the traditional client-server (CS) model to new type, peer-to-peer (P2P) ones. The P2P system is composed of peer computers, where each computer can play every role of client and server. In addition, the P2P system is scalable, i.e. a large number of computers are interconnected in various types of computers, ranging from personal computers to super computers. In addition, the P2P system is in nature open, i.e. computers autonomously join and leave the system, and the type of service and quality of service (QoS) supported by each computer are changing. Applications require the system to support various types of service requirements. In these P2P systems, it is difficult, maybe impossible to manage every component computer in a centralized control scheme as taken in traditional systems due to the scalability and openness of the P2P system. We have to develop novel concepts and architecture to design, implement, maintain, and … More operate the P2P system.In this research, we first discuss a hierarchical model for a scalable group of large number of peer processes. A group is a collection of multiple peer processes on various types of computers which are interconnected in networks like the Internet. Each peer process in a group send a message to multiple peer processes while receiving messages from multiple peer processes. Messages are required to be causally/totally delivered to processes in a group. The communication and computation overheads are O(n^*n) for the number n of processes in a group. In order to reduce the overhead, a group is composed of subgroups which are interconnected through gateway processes. Since a subgroup is smaller than the whole group, we can reduce the overhead. In addition, each peer process has to detect peer processes which support required service. We consider multimedia objects. In order to detect target processes, each peer process ask the acquaintance processes. Here, it is critical to discuss how much a peer process can trust each acquaintance process. We define the trustworthiness of each acquaintance peer in terms of QoS, and access right.This year, we discuss how to realize database applications on the P2P based architecture which we developed in the first year. In this research, database applications are realized in mobile agents. Especially, transactions can be fault-tolerant by using the mobile agents because the transactions can move to other operational computers from faulty computers.We published papers on the research in international journals like IJHPCN and JOIN and international conferences like IEEE AINA, ICPADS, ISORC, ISM, and DEXA. Less