Deadlock detection method based on side tracing for distributed system

Abstract

The invention discloses a deadlock detection method based on side tracing for a distributed system. The deadlock detection method comprises the following steps: S1, initially judging whether deadlock has occurred in the distributed system according to the state of the distributed system by each node in the distributed system, and initially judging the nodes performing deadlock as initiating nodes for deadlock detection to begin to carry out deadlock detection; S2, sending probe messages to the nodes depending on the initiating nodes by the initiating nodes; S3, receiving the non-initiating nodes of the probe messages and transmitting the probe messages to the nodes depending on the non-initiating nodes; S4, after all the probe messages of the nodes depending on the non-initiating nodes are received by the non-initiating nodes, sending the dependency messages of the non-initiating nodes to the initiating nodes by the non-initiating nodes; and S5, judging whether deadlock occurs according to dependency among the nodes after the dependency messages of all the non-initiating nodes are received by the initiating nodes. Compared with an existing method, the quantity of transmitted messages in the deadlock detection process can be reduced.

Description

technical field [0001] The invention relates to the field of program deadlock detection. More specifically, it relates to a deadlock detection method for distributed systems based on edge tracing. Background technique [0002] With the increasing scale of big data and the continuous development of data processing technology, the original stand-alone system cannot meet the current fast and efficient data processing needs. Therefore, more and more commercial systems are beginning to transform to a distributed architecture. Among them, deadlock detection in distributed systems has become an important research topic in parallel program error detection. [0003] The definition of deadlock is that two or more processes in the system appear to be blocked while waiting for each other during execution. Usually, some nodes in a distributed system more or less need to rely on shared resources on other nodes. The reason for the dependency between distributed nodes is that the process...

AI Technical Summary

Problems solved by technology

[0009] The purpose of the present invention is to provide a distributed system deadlock detection method based on edge tracking, which mainly solves the problems of too many messages and too large message size during deadlock detection, so as to reduce deadlocks compared with existing methods The number of messages passed during detection

Images