JPS63238652A - Operation managing system for distributed processing system - Google Patents

Abstract

PURPOSE:To prevent a drop of a throughput of the whole system by monitoring a utilizing state of shared resources, in the owning node of its shared resources. CONSTITUTION:A control table 2 for recording users 4, 5 for utilizing shared resources 1 in accordance with a node, and its utilization time, and a resources monitoring mechanism 3 for monitoring the contents of this control table 2 are prepared in advance. In this state, when there is a person whose requests to utilize the shared resources 1, full names of its users 4, 5 are recorded in the control table 2, and, when there is a user who is being recorded and also when there is a person who requests to utilize the shared resources 1, its full name and the waiting time are recorded as its idle waiting user in the control table 2, and by the monitoring mechanism 3, the control table 2 is checked at a prescribed time interval, and when the waiting time has exceeded a prescribed time, a warning is given to the idle waiting user. In such a way, a drop of a through-put of the whole system can be prevented.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an operation management method for a distributed processing system having a plurality of commonly used shared resources.

[Conventional technology]

°In conventional distributed processing systems that have multiple shared resources that are commonly used, in order to detect when the system is in a deadlock state, it is necessary to provide only one management mechanism for all shared resources. , whose management mechanism uses a method of monitoring the usage status of shared resources by all users.

[Problem that the invention seeks to solve]

In the conventional operation management method for distributed processing systems as described above, the resources of the entire distributed processing system are centrally monitored by a single management mechanism, so the central processing unit (host)
However, in a horizontal distributed processing system with multiple hosts, if a node to which a management mechanism is assigned encounters a system failure, the entire system will stop. In addition to the user information transmitted on the system network, information for operating the management 8% disaster is concentrated on that node, which causes a reduction in the throughput of the entire distributed processing system. be.

An object of the present invention is to distribute the resource management load, prevent the entire system from stopping, and eliminate the drawbacks of the conventional operation management method for distributed processing systems as described above. An object of the present invention is to provide an operation management method for a distributed processing system that can prevent a decrease in throughput.

[Means for solving problems]

The operation management method for a distributed processing system of the present invention includes a control table that records the usage status of the shared resource and a resource monitoring mechanism in each node that has the resource shared in the system.

In other words, the operation management method of the distributed processing system of the present invention is as follows:
An operation management method for a distributed processing system having a plurality of nodes having a shared resource, the control table including a control table for recording the names of users who use the shared resource and their usage times corresponding to the nodes; A resource monitoring mechanism is prepared to monitor the contents, and when there is a person who requests the use of the shared resource, the name of that user is recorded in the control table, and when there is a user in the record, the When there is a person who requests the use of the shared resource, his/her name and waiting time are recorded in the control table as a waiting user, and the control table is checked at regular intervals by the monitoring mechanism. The system includes the step of giving a warning to the waiting user when the waiting time exceeds a predetermined value.

〔Example〕

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a configuration diagram showing the configuration of nodes holding shared resources in a distributed processing system to which an embodiment of the present invention is applied, and FIG. 2 shows a distributed processing system having the nodes shown in FIG. FIG. 2 is an explanatory diagram showing a state in which resources are used.

In FIG. 1, 1 is a shared resource, 2 is a control table for recording the usage status of the shared resource, and 3 is a resource monitoring mechanism. In FIG. 2, 2 is a control table, 4 is a user who is using the shared resource, and 5 is a user who is waiting for the shared resource to become available.

In FIGS. 1 and 2, when no one uses the shared resource 1, the control table 2 is completely blank because no information is written therein, and the resource monitoring mechanism 3 updates the control table 2 with the contents at regular intervals. The check does not raise any warning signals. This situation! When the user 4 requests the use of the shared resource 1, the name of the user 4 is recorded in the control table 2, and the shared resource 1 becomes in use. Furthermore, after that, when another user 5 requests the use of the shared resource 1, the name of the user (waiting user) 5 and "°0"' as the waiting time are recorded in the control table 2. User 5 becomes a waiting user. After this time, the status of the control table 2 is checked by the resource monitoring mechanism 3 at fixed time intervals, and if there are consecutive waiting users 5 on the control table 2,
A fixed value is added to the waiting time. If the idle waiting time of the shared resource 1 exceeds the capacity of the distributed processing system, the resource monitoring 8% system 3 issues a warning to the idle waiting user 5.

As described above, a distributed processing system distributes the resource management load by monitoring the usage status of shared resources at the node that owns the resource, so if one of the nodes that owns the shared resource fails on the system network, However, only the part related to the shared resource will stop operating, the remaining shared resources will remain available as before, and the system will continue to operate. Since a situation where a possible deadlock does not occur, only the originally necessary user information is sent back and forth in the distributed processing system, thereby suppressing a decrease in throughput.

〔Effect of the invention〕

As explained above, by using the operation management method for a distributed processing system of the present invention, it is possible to prevent the entire system from stopping even if a failure occurs in some nodes, especially in a horizontal distributed processing system. This has the effect of making it possible to perform the following tasks, and it also has the effect of preventing a significant decrease in the throughput of the entire system.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram showing the configuration of a shared resource holding node according to an embodiment of the present invention, and FIG. 2 is a state of the control table when the distributed processing system shown in FIG. 1 is operating and using shared resources. FIG. 1... Shared resource, 2... Control table 5,... Resource monitoring mechanism, 4.5... User. Z system Junsu 4 Otsu figure

Claims (1)

[Claims] An operation management method for a distributed processing system having a plurality of nodes having a shared resource, the control table including a control table for recording the names of users who use the shared resource and their usage times corresponding to the nodes; A resource monitoring mechanism is prepared to monitor the contents, and when there is a person who requests the use of the shared resource, the name of that user is recorded in the control table, and when there is a user in the record, the When there is a person who requests the use of the shared resource, his/her name and waiting time are recorded in the control table as a waiting user, and the control table is checked at regular intervals by the monitoring mechanism. 1. An operation management method for a distributed processing system, comprising: issuing a warning to the idle waiting user when the waiting time exceeds a predetermined value.