Construction of Cooperative Distributed Systems

点击翻译按钮获取中文摘要

The purpose of this research was to investigate the method for constructing "cooperative" distributed systems with high extensibility and flexibility. The technical points for realizing cooperative distributed systems are three-fold : (1) a virtual memory management mechanism that facilitates close interaction among processes, even if they reside in separate address spaces, (2) a communication mechanism that enables cooperation of processes without spoiling their individual functionalities and security, (3) a mechanism for changing the action (or specification) of processes during their execution. The first point was achieved by introducing a novel concept for address space management called "versioned single-level store" on a linear 64-bit address spaces. The versioned single-level store integrates the management of secondary storage and virtual memory, thereby providing user programs a uniform view of whole storage of a computer system. The single-level store eliminates the operational differences between the volatile data (those on the primary storage) and persistent data (those on the secondary storage). The uniform view of volatile and persistent data helps processes to cooperate, which has been difficult on conventional systems because they provides little support for sharing data and control among separate address spaces. The second and third point was realized by the "trigger" mechanism. The trigger system notifies users of an occurrence of specific events in the operating system, such as read and write on a specific file in the file system, delivery of a specific message from/to a process, occurrence of virtual memory protection, and so on. By revealing these events that conventional systems have hidden from the user programs, user programs gained the ability of monitoring the activities in the system, and change their own behavior according to the change of the outer environment.