Finite-state machine extended model of distributed system and quasi-synchronous method for check points

Abstract

The invention discloses a finite-state machine extended model of a distributed system and a quasi-synchronous method for check points, and solves the problems of poor accuracy and stability in the process of establishing the check points of the distributed system. In the extended model, the distributed system is taken as a set of a plurality of courses; and the finite-state machine extended model is a set consisting of the finite courses: P is equal to {P1, P2, until Pn}, wherein Pi represents the courses, i is equal to 1, 2, until n, and n is more than or equal to 2. The quasi-synchronous method is divided into two phases; in the first phase, a coordinating course collects the channel information of each course and judges whether the current state of the distributed system is in a state of global consistency; if the judgment is yes, each course respectively saves respective state and the algorithm is finished; the second phase is a verification phase; and if the current state of the distributed system is not in a state of consistency, a course losing a message is determined and the coordinating course informs a sending course losing the message to resend the lost message till all messages are received or all courses are quitted abnormally due to overtime.

Description

Technical field [0001] The invention relates to a fault tolerance method of a distributed system, in particular to a distributed system finite state machine expansion model and a checkpoint quasi-synchronization method. Background technique [0002] At the application layer, a distributed system can be regarded as composed of several independent and complex interactive processes. Such processes cooperate with each other to complete a task, complete the communication between processes through message exchange, and realize the sharing of system resources. Common faults of distributed systems mainly include: crash faults, omission faults and sequence faults, response faults and random faults. There are two solutions based on hardware and software for troubleshooting methods. The software solutions mainly include active replication and passive replication. Passive replication can use two methods: forward recovery and backward recovery: forward recovery assumes that the nature of the ...

AI Technical Summary

Problems solved by technology

Asynchronous checkpoint algorithm In the process of checkpoint establishment, the processes do not need to negotiate to establish their own local checkpoints independently, but in the recovery process, complex algorithms need to be used to restore to a system consistency checkpoint

The key to the synchronous and asynchronous checkpoint algorithms is the establishment of the consistency checkpoint, that is, the acquisition of the process state that satisfies the fact that all messages are received, and the absence of lost messages and orphan messages is a sufficient requirement for the establishment of the consistency checkpoint. However, none of the above-mentioned algorithms has an accurate mathematical description of the input and output characteristics of the process and its channel, and there is no unified criterion for judging the lost message process, which affects the accuracy and stability of the algorithm.

Images