JPH07120305B2 - Database processing method by multiprocessor - Google Patents

Description

[Detailed Description of the Invention] [Outline] The local processing for centering the resource management of the database on the local memory and the shared processing centering on the shared memory are dynamically changed according to the access status. Regarding a database processing method by a multiprocessor, the target of shared processing is for each resource, which is a unit of the database, for the purpose of efficiently ensuring the integrity and improving the access performance according to the access status to the database resource. Control information for each processor module is processed symmetrically according to the setting information of the database management information storage unit and the database management information storage unit that stores whether it becomes a target of local processing or shared processing or non-processing. In any of the local processing that processes symmetrically It is configured to include an access management unit that performs the processing, and a shared / local change control unit that switches shared processing management and local processing management.

[Industrial application field]

The present invention relates to a database processing method by a multiprocessor that dynamically changes a local process for resource management of a database mainly in a local memory and a shared process mainly in a shared memory according to an access situation. Regarding

The multiprocessor system is roughly classified into a tightly coupled multiprocessor system and a loosely coupled multiprocessor system. Further, there are various types such as a composite of these and an intermediate one between the tightly coupled and the loosely coupled. The system configuration has been put to practical use.

On the other hand, in order to realize a database management system on one processor, there is a certain limit to the response, so it is considered to be processed by multiple processors. However, the integrity for maintaining data in a correct state is always considered. Warranty processing time tends to increase. In order to improve processing performance, it is necessary to make this integrity guarantee efficiently.

[Conventional technology]

FIG. 7 shows an example of a conventional database processing system by a multiprocessor.

A processor module is shown in FIGS.
Reference numeral 11 is a processing device having a CPU and a local memory 25, respectively. Each processor module 11 is bus-coupled to enable inter-processor communication.

Access to the resource 22 of the database is normally managed in a unit called a granule, that is, a unit of a certain data set in the database. This unit may be, for example, a file level or a record level.

In access to this resource 22, integrity guarantee for maintaining data integrity such as prevention of occurrence of data inconsistency due to access competition and recovery at the time of failure must be performed.

Conventionally, in a system such as that shown in FIG. 7 (a), for the data stored in the database entity storage unit 13 of the secondary storage, the resource 22
However, each processor module 11 is statically determined in advance or dynamically determined according to the load condition, and each processor module 11 locally guarantees the integrity of each resource 22.

On the other hand, in the system as shown in FIG. 7B, the resource 22 read from the database entity storage unit 13 is placed in the shared memory 10 accessible by each processor module 11 to perform processing such as exclusive control and buffer control. Has been performed in common in the entire system, that is, each processor module 11 performs symmetry, thereby realizing integrity guarantee as shared processing.

[Problems to be Solved by the Invention]

In the conventional processing method, the integrity guarantee is either performed by a local process closed in each processor module 11 or a shared process common to the entire system. It was never done as a shared process.

In the method of managing the resource 22 by the local processing as shown in FIG. 7 (a), when the resource 22 of one processor module 11 is desired to be accessed by another processor module 11, the inter-processor communication enables exclusive control. A lock request and a buffer access request are issued to the management source processor module 11. For the resource 22 that has a large communication overhead for that purpose and is likely to be accessed from multiple processor modules 11,
There was a problem that the processing performance was extremely poor.

In addition, when performing optimal allocation for a certain resource 22,
Since humans had to determine the processor module 11 to be allocated with reference to the access statistical data and the like, there was a problem that tuning was difficult.

On the other hand, in the method of managing the resource 22 by the common processing as shown in FIG. 7B, even when only one specific processor module 11 makes a large access to a certain resource 22, the system is always operated. Since it is necessary to manage exclusive control and buffer control in consideration of the whole,
There is a problem that the processing cost becomes large and efficient access control cannot be performed.

An object of the present invention is to solve the above-mentioned problems, to efficiently perform integrity guarantee according to the access status to the resource of the database, and to improve the access performance.

[Means for Solving the Problems]

 FIG. 1 shows a configuration example of the present invention.

In FIG. 1, 10 is a shared memory, 11-1, 11-2, ... Are processor modules (PM), 12 is each processor module 1
1 is a high-speed bus connecting the shared memory 10, 13 is a database entity storage unit, 20 is a database management information storage unit, 21 is a lock information storage unit that stores exclusive control information, 22 is a resource that is a unit of the database, 23 is an access management unit that manages access to the resource 22, 25 is a shared / local change control unit, 25 is a local memory of each processor module, and 26 is a local lock information storage that stores exclusive control information in each processor module. Represents a part.

Each of the processor modules 11-1, 11-2, ... Has a local memory 25 and is bus-coupled to the shared memory 10. Further, in this system, each processor module has a database entity storage unit 13 and an I / O bus 14
, And can be equally accessed from each processor module. In addition, two buses are 1
You may comprise by one.

The resource 22 of the database is managed based on the management information set in the database management information storage unit 20. The database management information storage unit 20 may be arranged in the shared memory 10 or the local memory 25.

According to the present invention, the database management information storage unit 20 is provided with a shared / local display unit that indicates whether the resource 22 to be accessed is a target of shared processing or a target of local processing. ing. The database management information storage unit 20 may be in the local memory 25.

Each of the processor modules 11-1, 11-2, ... Has the same symmetrical configuration, and has an access management unit 23 and a shared / local change control unit 24.

The access management unit 23 uses the resources 22 on the shared memory 10.
And the management function of the resource 22 on the local memory 25.

In shared processing for these resources 22, resource 22
Each processor module symmetrically performs the integrity guarantee processing for. Here, the symmetric means that control is performed in the same processing sequence regardless of which processor module accesses the resource 22.

In the local processing, the integrity guarantee processing for the resource 22 is performed by the specific processor module to which the resource 22 is assigned. That is, the processing between the processor module that manages the resource 22 and the other processor modules is non-symmetrical.

Whether to control access to the resource 22 as a shared process or a local process is determined by referring to the shared / local display unit of the database management information storage unit 20.

The sharing / local change control unit 24 manages the access status of each resource 22, and when the access frequency is unevenly distributed to a specific processor module, the resource 22 is targeted for local processing by the specific processor module and accessed. When the frequency is not unevenly distributed to a specific processor module, change control is performed so that the resource is shared. That is, by monitoring the access status of each resource 22, the database management information storage unit 20 can be shared / used.
Update the local display to control the change of resource management from shared to local or from local to shared.

[Operation] FIG. 2 is an explanatory view of the operation of the present invention.

According to the present invention, for resource management of a database, local processing centered on a local memory and shared processing centered on a shared memory are dynamically changed according to an access situation by sharing / local change control. It is the one.

In the case of local processing, exclusive control and buffer control for accessing a resource are performed by a specific processor module that manages the resource. When access to a resource is unevenly distributed in a specific processor module, most access control can be speeded up by performing exclusive control and buffer control locally.

In the case of shared processing, integrity assurance is performed for the entire system, so the processing may be slower than in the case of local processing for resources managed by itself by a specific processor. However, for resources that are uniformly accessed by each processor without unevenly distributing access to specific processor modules, avoiding concentration of load on specific processor modules and requesting access to specific processors. It is possible to prevent an increase in the communication cost of, and improve the throughput of the entire system.

According to the present invention, as shown in FIG. 2, when the access is unevenly distributed to a specific processor module, the local processing management is performed, and when the uneven distribution tendency is resolved, the shared processing management is dynamically performed. ， It will be possible to perform optimal access management according to the access status of resources.

〔Example〕

FIG. 3 is a system configuration example according to an embodiment of the present invention, FIG.
FIG. 5 is an explanatory view of a processing outline according to one embodiment of the present invention, FIG. 5 is an explanatory view of management data used in one embodiment of the present invention, and FIG. 6 is a shared / local change control according to one embodiment of the present invention. An example of the processing will be shown.

The present invention is implemented, for example, on a multiprocessor system as shown in FIG. In FIG. 3, SSU is a shared memory device, SCA is an I / O interface adapter, DK is a disk storage device, DLP is a data link processor, and DRC is a driver / receiver card for line control.

Each processor module 11 has one or a plurality of CPUs and a local memory 25. I / O to the local memory 25 is performed in the same way as a cache memory.
Can be reduced. In addition, each processor module 11
Under the control of a bus controller (not shown), the shared memory device SSU can be accessed via the high-speed bus 12.

The high-speed bus 12 and the I / O bus 14 are multiplexed for speeding up bus operations and for troubleshooting. Of course, it may be a single layer.

Each processor module 11 is configured such that the processing of one processor module 11 can be performed by another processor module 11 instead. Therefore, when the processor module 11 fails, the failed device can be separated from the system and the remaining processor modules 11 can continue the service.

The substance of the database is stored in a disk storage device DK such as an optical disk or a magnetic disk, and can be accessed from any processor module 11 via the I / O bus 14 and the adapter SCA.

As a system similar to such a system, Nikkei BP
Published by Nikkei Electronics No.461, November 28, 1988
Although various systems are known, such as those described in "Japanese issue", P110 to P115, the present invention is not limited to these, and the present invention can be realized on various multiprocessor architectures in which shared memories are bus-coupled. It is possible.

The processing of the access request to the database from the terminal etc.
This is performed as shown in FIG.

FIG. 4 (a) is a flow of processing from inputting a message to starting the application program APP, and FIG. 4 (b) is
The flow of processing of a database management part is shown. In the following description, ~ corresponds to ~ shown in FIG.

An arbitrary processor module 11 performs input processing of a message such as a database search request. For each terminal that inputs a message, the processor module 11 in charge of the input processing may be determined in advance.

Application program AP to be executed according to the message
Messages are distributed to the application program APP of each processor module 11 according to the database accessed by P and the load status of each processor module 11.

As a result, the execution of the application program APP to which the message has been distributed is started. The application program APP is
An access request in the database operation language is issued to the database management unit if necessary.

In the database management unit, the access management unit 23 performs exclusive control for access requests and resources, that is, lock control for avoiding access conflict, according to the interface with the application program APP. This exclusive control differs depending on whether the accessed resource is a shared processing target or a local processing target.
Effective lock management for the entire system or for a specific processor module.

~ Similar to exclusive control, buffer control is performed by selectively using shared memory and local memory depending on whether the resource is a shared processing target or a local processing target. If necessary, access to secondary memory is also made.
Then, the access result is the application program APP of the request source.
To notify.

Apart from the above access processing, the shared / local change control unit 24 manages / maintains the access status of each processor module. For each resource, it also collects information for determining whether to perform shared processing or local processing.

For each resource, check whether the access frequency is concentrated on a specific processor module. Then, it is determined whether to perform local processing management or shared processing management.

When it becomes necessary to change the shared or local management form, the management information and resources are moved between the shared memory and the local memory, and the shared / local change processing is performed.
The shared / local change control unit 24 performs the above-described processing periodically at predetermined intervals or at another appropriate trigger.

As the management data used for managing the shared processing and the local processing, there is data as shown in FIG. 5, for example.

The space management table 30 shown in FIG. 5 is a table holding space management information for managing the database. In this embodiment, one processor module can have a plurality of address spaces, and a kernel (core) unit common to each space manages the space and the database. When space is abnormal, mutual backup is possible. The space management table 30 is the database attribute table 31.
With a pointer to.

The database attribute table 31 has the following information for each resource.

(A) Resource name: This is a name added to a unit for managing sharing / localization of each resource of the database. By this name, the resource 22 of the database actually arranged can be identified. That is, the data managed by the shared buffer or the buffer in the local memory can be dealt with via the buffer management information 32.

(B): Shared / local display: This is a display with a flag or the like indicating whether the corresponding resource 22 is shared-type control or local (ubiquitous) type control. According to this display, whether to allocate the resource 22 or its management information to the shared memory is determined.

(C) Access status pointer: Points to the database access status table 33.

(D) Database allocation information: In the case of a local type, it has an identifier of a processor module that guarantees its integrity.

The database access status table 33 has the following information indicating how many times each processor module has accessed that resource in order to manage the access status for each resource.

(A) PM-ID: This is an identifier of a processor module that has accessed the resource directly or indirectly by an access request.

(B) Reference count / update count: The reference or update access count for each processor module. The control table for management data shown in FIG. 5 is arranged in the shared memory or the local memory in the processor module.

When the resource is of a local type, the database attribute table 31 and the database access status table 33 exist in the local memory of the processor module that performs the local control.

If the resource is shared, the database attribute table 31 is in the shared memory, but the database access status table 33
Are arranged on the shared memory and the local memory in each processor module, and the processor module updates only the access status of the processor module. On the shared memory, access status data is collected from each processor module at a certain time interval.

The processing of database access with local integrity guarantee and database access with shared integrity guarantee are individually known in the distributed database management system, etc., and detailed description of the processing contents is omitted. To do.

In particular, the shared / local change control processing related to the present invention is
It is started at a predetermined cycle, and for example, the processing shown in FIG.
Is done like.

For each resource, determine whether it is fixed as shared or local. If the operator or system administrator fixedly specifies the resource type regardless of the access status, sharing / local modification is not performed.

The access status of each processor module is collected and registered in the database access status table on the shared memory.

At present, whether the resource of interest is a shared type or a local type is determined by the database attribute table. In case of shared type,
Move to processing. If local type, move to processing.

Check the access status result to see if the number of accesses in one processor module is extremely higher than that in another processor module. If the access is not unevenly distributed, the processing for the resource ends.

A processor module that guarantees the integrity of a resource that is determined to be a local type is a processor module that has an extremely large number of accesses.

At this point, the database access status table is cleared and these management information will be transferred to the processor module that guarantees the integrity in the future.

As for buffer management, shift to local processing management,
The latestness of the database is managed by the processor module that is determined to be local type.

In order to improve the access efficiency of the database, the database buffer for this resource is moved to the buffer provided in the local memory of the processor module that performs local processing.

Currently, in the case of the local type, the processor module that performs the local processing registers the access statuses thereof based on the access request from each processor module. Referring to the result, if the access count of the processor module that is performing local processing is less than any of the access counts (access request counts) of other processor modules, it is determined that the resource is shared, and the resource is shared. Execute the processes to migrate.

At this point, the database access status table is cleared, and in the future, these pieces of management information are moved from the local memory of the processor module that guarantees the integrity to the shared memory.

Shared processing management is also used to manage the database freshness by buffer management. That is, each processor module manages the resources symmetrically.

The database buffer for this resource is moved to the shared memory to improve the reference efficiency by other processor modules. It should be noted that all shared resources do not necessarily have to be processed on the shared memory. For example, after locking, a copy thereof can be moved to the local memory for processing.

It goes without saying that the condition of resource management shift from shared to local or from local to shared in the above embodiment is an example, and various condition settings can be made. For example, there are cases in which people are forced to make decisions, and schedules are changed systematically.

〔The invention's effect〕

As described above, according to the present invention, a database
Shared-type integrity guarantee or local-type integrity guarantee is automatically selected according to the access status of resources. Optimal allocation of resource management aims to reduce processing costs such as high-speed access and reduction of communication overhead. It will be possible.

[Brief description of drawings]

FIG. 1 is a configuration example of the present invention, FIG. 2 is an operation explanatory diagram of the present invention, FIG. 3 is a system configuration example according to an embodiment of the present invention, and FIG. 4 is an outline of processing according to an embodiment of the present invention. Explanatory diagram, FIG. 5 is an explanatory diagram of management data used in one embodiment of the present invention, FIG. 6 is an example of shared / local change control processing according to one embodiment of the present invention, and FIG. Here is an example: In the figure, 10 is a shared memory, 11-1, 11-2, ... Are processor modules, 12 is a high-speed bus, 13 is a database entity storage unit, 14
Is an I / O bus, 20 is a database management information storage unit, 21 is a lock information storage unit, 22 is a resource, 23 is an access management unit, 24 is a shared / local change control unit, 25 is a local memory, and 26 is local lock information. Represents a storage unit.

Front Page Continuation (72) Inventor Tomohiro Hayashi 1015 Kamiodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa, Fujitsu Limited (72) Inventor Kazuhiko Saito 1015 Kamedota, Nakahara-ku, Kawasaki, Kanagawa Prefecture, Fujitsu Limited (72) Inventor Masayoshi Shimonori 1015 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture, Fujitsu Limited

Claims (1)

[Claims] 1. A database processing system by a multiprocessor system comprising a plurality of processor modules (11-i) each having a local memory (25) and a shared memory (10) bus-coupled to each of these processor modules. In database, a database management information storage unit (20) that stores, for each resource, which is a unit of the database, whether to be a target of shared processing or a target of local processing, and the database management information for the access request to the database. According to the setting information in the storage unit, the access control unit performs access control by either shared processing which is processed symmetrically in each processor module or local processing which is processed non-symmetrically in each processor module (2
3), the access status is managed for each resource that is a unit of the database, and when the access frequency is unevenly distributed to a specific processor module, the resource is targeted for local processing by the specific processor module and the access frequency is A database processing method by a multiprocessor, characterized by comprising a shared / local change control unit (24) for sharing the resource when the resource is not unevenly distributed to a specific processor module.