JPH0266647A - Fault information informing system for distributed type data base system - Google Patents

Abstract

PURPOSE:To accurately and effectively carry out the process of a distributed type data base by positively informing the system states between both sites with each other to accurately grasp the system states with each other and producing a process procedure to the data process request in a distributed type data base system based on the latest system state information. CONSTITUTION:The resource state control/transmission-reception means 21-26 controls the resource states (the working states of DBMS 31-36, the load states of distributed type information process control systems 11-16, etc.) of sites (a)-(f) and also perform the transmission and reception of the state information. In the case a fault occurs at the server side, for example, the resource state control/transmission-reception means of a relevant site transmits the fault information to each site. At the same time, the inter-site connection matrices of all systems are held at the sites (a)-(f) respectively. Thus the state information on the faulty site are transmitted among those sites with each other, and the latest state information on each site is grasped.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a failure information notification method for a distributed database system, and in particular to generating accurate and efficient processing procedures even when a failure occurs. This paper relates to a failure information notification method for a distributed database system that allows for

[Prior art]

Conventional database systems seek to integrate mutually related data into one, that is, to physically centralize the data. This centralized database system makes it easy to control data and relatively simple to operate the system, but there are problems in that a failure in one part of the system affects the entire system and that communication costs from remote locations increase. be.

Furthermore, distributed database systems composed of multiple computer systems are currently attracting attention. This distributed database system improves the reliability of the entire system and allows for flexible configuration, but the control of distributed data and system management become complicated.

In addition, in a distributed database system, the processing requester (
When a process from a client (hereinafter referred to as a client) to a process requesting side (hereinafter referred to as a server) is completed, the server needs to perform a process of releasing resources related to the client.

However, if some kind of failure occurs on the server side, the process of releasing client resources will fail.

In this case, the server sends [
A method is adopted in which a false report is made that ``resource release processing has completed successfully'', and when the next processing request from the client comes to the server, a report is sent to the client that ``resource reservation failed.'' has been done. This method is not unique in distributed database systems, but is common in complex subsystems.

Regarding the conventional failure information notification method, for example, ``Performance of the File Eye Leftry System for Database Installation and
Distributed Networks, Unisli W.C., National Computer Conference, 1976, pp. 577-58.
7 (Performance of file)
directory systems for
databases in 5tar and
distributed networks, W
ESLEY Iil, C11U, National
Computer Conference, 1976,
pp.

577-587)".

[Problem to be solved by the invention]

In the above conventional technology, if some kind of failure occurs on the server side and the process of releasing resources for the client fails, the process requester on the client side (e.g., user application, transaction, etc.) that received the false completion report It has the effect of shortening the response time to
By the time the next processing request is issued, the failure on the server side may have been recovered, and it is expected that there will be no report of "resource processing failure."

However, this is used in an information processing system within a single computer system, and there are problems when using it in a distributed processing system.

For example, in distributed database processing, etc., heavy preprocessing is required to convert a global access view of the database to a local access view on each server on the client side. This conversion also knows the status of each server that must be accessed.

Set the procedure for accessing each server.

Therefore, when performing this conversion, the latest state of each server must be used.

In this way, when performing distributed database processing, the method of forwarding false information to the client side will cause the server to not know whether the failure has been recovered or not, and the client's global to local Since preprocessing, which is a burdensome conversion, and processing for issuing access processing requests to the database are performed, these processings, communication processing, etc. may be wasted.

The purpose of the present invention is to improve such problems, eliminate unnecessary processing requests even when a failure occurs, and accurately generate appropriate processing procedures such as distributed database processing procedures. Another object of the present invention is to provide a fault information notification method for a distributed database system that can efficiently execute distributed database processing.

[Means to solve the problem]

In order to achieve the above object, the distributed database failure information notification method of the present invention provides means for transmitting and receiving system status information including failure status information between sites in a distributed database system composed of a plurality of sites.
resource status management/transmission/reception means) and means for managing the system status information (resource status management/transmission/reception means)
The sites actively notify each other's system status, accurately grasp the system status of other sites, and process procedures for data processing requests in the distributed database system based on the latest system status information. It is characterized by generating.

[Effect]

In the present invention, the distributed information processing management system in the computer system of each site constituting one distributed database system is provided with resource status management/transmission/reception means.

This resource status management/transmission/reception means determines the status of various resources for database processing, such as the operating status of the database management system (hereinafter referred to as DBMS) at the site, the load status of the distributed information processing management system at the site, etc. In addition, as a response to a processing request from another site or as an interruption to another site, the relevant site resource status is transmitted to the other site, and the resource status transmitted from the other site is received and managed.

This allows us to accurately understand the system status of each site,
Appropriate processing procedures can be generated even if a failure occurs at each site.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 is a configuration diagram of a distributed database system in an embodiment of the present invention.

First, let's talk about the configuration.

In this embodiment, in order to simplify the explanation, it is assumed that databases distributed among the sites h a to f are managed.

Site a = f in this embodiment is computer system 1 to 6.
and databases 41-46, and each site is connected by communication lines 100-109. Further, the computer systems 1 to 6 are distributed information processing management systems 11 to 16 and database management systems (DB), respectively.
MS) 31-36. Further, the distributed information processing management system (distributed database management system) 11 to 16 includes resource state management/transmission/reception means 21 to 26, respectively.
Equipped with.

The resource status management/transmission/reception means 21 to 26 manage the resource status of each site a"f (the operating status of the DBMSs 31 to 36, the load status of the distributed information processing management systems 11 to 16, etc.), and transmit and receive status information. For example, if a failure occurs on the server side, the resource status management/sending/receiving means of the relevant site will send failure information to each site.Also, in order to grasp accurate status information of each site, this In this embodiment, each site a = f maintains a connection matrix between sites in the entire system (see Figure 3), and by this, the status information of the site where a failure has occurred is transferred between the sites, and the latest information of each site is Understand status information.

Next, a method for transferring system status information between sites will be described.

FIG. 3 is a side view of a connection matrix between sites in an embodiment of the present invention.

In this embodiment, a connection matrix between sites (a-f) shown in FIG. 3 is held at each site. This is a connection matrix between sites in the entire system, and based on this, system status information of sites where a failure has occurred is transferred to each other,
Understand the latest status information of other sites.

For example, if some problem occurs at site a in FIG. 2 and the system status information of that site is to be transferred to each site, site a scans the connection matrix and selects sites that are adjacent to site a, In other words, find a site where the row of a is 1. As a result, site b and site C are obtained.

Next, communication lines 101 and 102 are connected to these sites, respectively.
The system status information si and the current timestamp information t of site a are broadcasted via . Site b and site e subsequently transfer the received information to related sites based on the connection matrix S.

In other words, site b has communication lines 103, 105°106
, the system status information S□ and time stamp information t1 of site a are broadcast to site C, site d, and site f. Similarly, at site e, system status information S and time stamp information L□ are sent to communication line 1.
04, 107, and 108 to site C, site d, and site f.

If there is a conflict in the transferred information from other sites, such as site C, site d, and site f, check whether the time stamp information is new or old on that site, invalidate the old system status information, and use the old system Reverse the route through which the status information was transferred and transfer the new system status information to each site on that route. Note that the route can be easily determined from adjacent matrices.

For example, if the information from site e is new, it is transferred from site C to site b via communication line 105.

At each site that receives this system state information, if a conflict as described above occurs, it checks the time stamp information and scans the adjacency matrix to confirm that the next site to be transferred is not site a.

As a result, if the site to be transferred to the primary site is site a, the process of transferring the system status information S of site a is terminated, and in other cases, the same process as above is continued. Note that if the timestamp information t is equal and there is no connection from the site to another site 1-, the transfer process of the system status information S1 of site a is terminated at that point, and the connection to the other site is terminated. If there is, the system status information transfer process continues.

With this method, each site a = f transmits its own system status every time the system status changes due to a failure, etc., so that each site can understand each other's system status and at least update the latest information of nearby sites. status information is maintained at each site.

Next, the operation of the distributed database system of this embodiment will be described.

FIG. 1 is a flowchart showing a failure information notification method of a distributed database system according to an embodiment of the present invention.

In this embodiment, when a transaction occurs at site a, for example, in distributed information processing at each site, a distributed information processing management system (distributed database management system (distributed DBMS)) 11 is called (1000).

As a result, the distributed DBMS 11 analyzes the processing contents of the transaction (1001), and if the syntax is correct, the resource state management/transmission/reception means 21 of the distributed DBMS 11 analyzes the processing content of the transaction based on the connection matrix shown in FIG. It performs transfer processing, checks the status information of other sites that it knows and maintains, and obtains the system status of the site where the database entity accessed by the transaction is located.

Note that the processing content of this transaction is that the user application program at the site is actually started, a distributed database access command is issued, and the distributed DBMS II processes it, but in this example, For the sake of simplicity, we will consider this command and transaction to be the same.

Next, create a processing procedure for the transaction based on the acquired system status information of other sites (10
02).

For example, if the transaction needs to access the database 46 of site f, the following routes (1) to (
There are multiple access paths including 8).

(1) Communication line 101 → site b → communication line 106 → site f (2) Communication port H101 → site b → communication line 105 → site C → communication line 109 → site f (3) Communication line Lo
t→Site B→Communication line 105→Site C→Communication line 1
07 → Site e → Communication episode! 108 → To site 1 f (4) Communication line 102 → Site e → Communication line 108 → Site f (5) Communication line 102 → Site e → Communication line 107 → Site C → Communication line 109 → Site f (6) Communication line 10
1 → Site b → Communication line 103 → Site d → Communication line 1
04 → Site e → Communication line 108 → Site f (7) Communication line 101 → Site b → Communication line 103 → Site d → Communication line 104 → Site Ie → Communication line 107
→ Site C → Communication line 109 → Site f (8) Communication line 102 → Site e → Communication line 104 → Site d → Communication line 103 → Site b → Communication line 106 → Site f Among the multiple access paths generated in this way The distributed DBMSll-16 at each site selects the shortest access path from the above. This selection is based on each site's knowledge of the accurate system status of other sites (at least the latest status information of nearby sites).

For example, in the processing procedure generated at site a, site e
When it is determined that it is optimal to transfer the processing request to site f (1003, 1004), the processing procedure is transferred to site e via communication line 102, and site e transfers the processing request to site f. A distributed information processing management system 14 examines the route to site f (10
05).

In other words, the processing procedure generated by site a is
Since the line status between site e and site f may have been formed using old information, site e uses the system status information managed by its own site's resource status management/transmission/reception means 24 again to check the validity of the processing procedure. Check the gender.

As a result (lOO6), if the processing procedure is valid, the processing procedure will be transferred to site f as it is.1
007), and if invalid, the site e reconfigures the processing procedure transfer route to site f in consideration of the processing load 1008), and transfers it to the next site 1- (100
9).

The site to which the processing procedure has been transferred in this way performs the process of determining an effective route in the same way as site e, and transfers the processing procedure to site f (1010).

Furthermore, site f, which has received the processing procedure, executes it (1011), generates an optimal path to site a (1012), and transfers the processing result to the next site (1011).
013).

The site that receives the processing result checks the transfer route and then transfers the processing result to site a (10
14).

In this way, by managing and using system status information between sites, even if a server failure occurs, it is possible to eliminate waste and efficiently perform distributed database processing.

〔Effect of the invention〕

According to the present invention, at least between sites that are close to each other,
The latest system status is accurately known, and the generation of distributed database processing procedures is never wasted.

As a result, even when a failure occurs, an appropriate processing procedure can be generated and distributed database processing can be executed accurately and efficiently.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing a failure information notification method of a distributed database system according to an embodiment of the present invention, FIG. 2 is a configuration diagram of a distributed database system according to an embodiment of the present invention, and FIG. FIG. 3 is a side view of a connection matrix between sites in one embodiment. 1 to 6: Computer system, 11 to 16: Distributed information processing management system (distributed database management system (distributed DB
MS)), 21-26: Resource status management/transmission/reception means, 31-36: Database management system (DBMS)
, 41-46: Database (DB), 100-109
: Communication line. Figure (Part 2) Figure

Claims (1)

[Claims] 1. In a distributed database system composed of multiple sites, the sites are provided with means for transmitting and receiving system status information including failure status information, and means for managing the system status information, and the systems are 1. A failure information notification system for a distributed database system, characterized in that the systems notify each other, grasp the system status, and generate a processing procedure for a data processing request in the distributed database system based on the system status.