異種クラスタの混在する分散型システム

概要

論文の詳細を見る
Kinds of distributed applications like teleconferences and telemedicines have been developed and widely available now. In the applications, multiple autonomous application processes are cooperated to achieve some objectives by communicating with each other through communication networks. The processes in the distributed system may suffer from faults, i. e. Byzantine faults[2] and stop-faults[4]. An approach towards making the system fault-tolerant is to replicate the processes in the system. There are two schemes to replicate the processes, i. e. 3tate machine[3] and primary-backup[1] approaches. In the state machine approach, the replicas can continuously service in the presence of the faults while it takes time to recover from the fault of the primary replica in the primary-backup one. However, it is expensive to replicate all processes in the system using the state machine approach. The primary-backup approach implies less over-head than the state machine one. Kinds of applications have different requirements on reliability and availability to the system. For example, all the faults have to be invisible In some applications while the processes can be rebooted and restarted in other applications if some faults occur. In this paper, each process is replicated into a collection of multiple replica', which is named a cluster. We would like to propose a system composed of dynamic clusters in each of which the replicas are replicated in kinds of replication schemes so that the requirement from the application are satisfied. Moreover, the replication scheme of the cluster changes dynamically ac-cording to the change of the system environment and requirements. In section 2, we present the replication schemes. In section 3, we present the system model. In section 4, we define a dynamic cluster. In section 5, we present inter-cluster communication.
一般社団法人情報処理学会の論文
1996-09-04

異種クラスタの混在する分散型システム

スポンサーリンク

概要

著者

関連論文

スポンサーリンク