JP2009037544A - Data base processing method, and implementation system and program therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology capable of: deriving a suitable locking form in data base processing; and setting a dynamically suitable locking form to an O/R mapping framework. <P>SOLUTION: A tracing log and a mapping table in a first storage are referred to and analyzed. An operation, a program executed based on the operation, and a tracing log of the program are associated and stored in a second storage. The operation, a program executed based on the operation, and a tracing log of the program which are thus stored are referred to. The tracing log of the operation is evaluated for each locking form and the locking form to be used by a program executing apparatus is decided. The decided locking form is set to a plurality of program executing apparatus. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a database processing technique for selecting a lock format for a database and controlling a queue of requests.

  In the prior art, when an application directly uses an API (Application Programming Interface) such as JDBC (Java Database Connectivity) to operate a database, a method in which an SQL statement is directly written in a source code by a user is mainly used. Sentences are scattered in each source, and management becomes complicated. Also, the lock format for the database must be manually set for each table by the user.

  On the other hand, the Object / Relational mapping (hereinafter referred to as O / R mapping) that has appeared is a framework that performs mapping of objects and records in a relational database table, and exclusive control in units of records can be performed by O / R mapping framework or DataBase Management. Managed by a system (hereinafter, DBMS). In general, there are two methods of performing exclusive control, the pessimistic method and the optimistic method, and the user specifies which method for each object.

  In the pessimistic method, since the lock is applied from when data is first read, the update process is always successful. Instead, because the lock acquisition time is long, if there is a transaction that accesses the same record, lock contention may occur and concurrency may be reduced.

  The Optimistic feature of acquiring a shared lock is that the exclusion range is narrow, so there is little possibility of lock contention, and concurrency is excellent. However, if there are multiple transactions that change the same data, there is a high possibility that the update will fail.

  The O / R mapping that maps a relational database record to an object treats the record as an object so that it can be handled without being aware of the record in the database. In order to operate a database record, a user operates a database record by operating an object using an API provided by the O / R mapping framework.

JP 2006-92358 A

  Since the user has to manually set the lock method for the O / R mapping framework according to the characteristics of the application, the system configuration, and the performance requirements, it takes time and effort.

  An object of the present invention is to derive an appropriate lock type and dynamically set an appropriate lock type for the O / R mapping framework. The lock type is set according to the situation (access tendency, load). Is to select dynamically.

  In addition, even if an appropriate lock method is dynamically selected, if there are many accesses to the same data and the lock collision rate is not improved, resources will be saved if lock waits frequently occur by other transactions that cannot acquire the lock. It cannot be used effectively and the overall performance of the system is degraded.

  For the purpose of the present invention, it is possible to improve the performance of the entire system by not only changing the lock method but also performing flow control (change of the number of simultaneous executions) of requests accessing the same data.

  The present invention is a computer system that selects a lock format for a database and manages database processing, analyzes a trace log and a mapping table, determines an object lock format, and sets the lock format.

  In order to achieve the above object, the computer system of the present invention creates an operation correspondence table and an object management table by analyzing a trace log and a mapping table used for failure analysis. Then, an appropriate lock format is determined from the operation correspondence table and the object management table, which are analysis results of the trace log and the mapping table, and the lock format is dynamically changed. Furthermore, the management server of the present invention controls the number of concurrent executions of the queue from the operation correspondence table and the object management table, and the scheduler efficiently allocates requests to the queue, thereby performing a process of suppressing the lock contention rate.

In the present invention, it is possible to automatically determine an appropriate lock type of exclusive control for a database by analyzing a trace mainly used in failure analysis.
In addition, the management server automatically sets an appropriate lock format for the database, so that the user can reduce the setting work for the database.
Furthermore, not only changing the lock format, but also performing flow control of requests that access the same data, it is possible to efficiently allocate requests to the application server and improve the performance of the entire system.

  A computer system according to an embodiment for managing database processing by selecting a lock format for the database will be described below.

  FIG. 1 is a diagram showing a configuration of a computer system that manages a plurality of program execution devices (application servers) according to the present embodiment. As shown in the schematic diagram of the basic configuration of FIG. 1, when receiving a request, the load balancer 101 allocates the received request to the scheduler 102 according to the rules of the load balancer. When a request is allocated to the scheduler 102 by the load balancer 101, the scheduler 102 refers to the queue table 104 and stores the request in the queue 103 in order to determine which queue the request is assigned to.

  If there is a request stored in the queue 103, the container 126 takes out the request from the queue 103 and starts processing. The container 126 acquires the trace log 115 using the tracer 114 when it is time to acquire the trace according to the processing of the request. The container 126 passes the request to the application 106.

  The application 106 operates the user object 107, and the user object 107 notifies the O / R mapping framework 108 of an operation performed from the application 106 to the user object 107.

  The O / R mapping framework 108 acquires the lock format of the user object 107 operated with reference to the mapping table 111. The O / R mapping framework 108 performs an operation according to the lock format on the reference unit 109 and the update unit 110.

  That is, when referring to the data of the database server 113, the reference unit 109 of the O / R mapping framework 108 issues a command for referring to the data to the database access unit 112. The database access unit 112 transmits a database command to the database server 113 in order to actually operate the table 124 of the database server 113. In the database server 113, the DBMS 125 interprets the received command, processes the table 124, and returns a result.

  When updating data in the database server 113, the update unit 110 of the O / R mapping framework 108 issues a command to update the data to the database access unit 112. The database access unit 112 updates the table 124 of the database server 113.

  The O / R mapping framework 108 sets the result for the user object 107 after referring to or updating the database. When the update process is executed, the O / R mapping framework 108 updates the user object 107.

  The application 106 refers to the result of the user object 107, and the application 106 returns the result to the container 126. Further, the container 126 returns the result to the transmission source of the request.

  The analysis unit 116 of the management server 100 that is a management apparatus analyzes the mapping table 111 and the trace log 115 to create an operation correspondence table 123 and an object management table 122.

  The determination unit 117 of the management server 100 includes a lock determination unit 118 and a queue determination unit 119. The lock determination unit 118 determines an appropriate lock format to be set for the user object 107 from the operation correspondence table 123 and the object management table 122. The queue determination unit 119 determines the setting of the number of simultaneous executions of the queue from the operation correspondence table 123 and the object management table 122.

  The setting unit 120 of the management server 100 refers to the server table 121 to determine the scheduler 102 to be set, and sets an appropriate lock format determined by the lock determination unit 118 for the O / R mapping framework 108. In addition, an appropriate number of concurrent executions determined by the queue determination unit 119 is set for the scheduler 102.

The scheduler 152 has a function equivalent to that of the scheduler 102. The queue 153 has the same function as the queue 103. The queue table 154 has the same function as the queue table 104. The container 176 has a function equivalent to that of the container 126. The application 156 has a function equivalent to that of the application 106. The user object 157 has the same function as the user object 107. The mapping table 161 has the same function as the mapping table 111. The O / R mapping framework 158 has functions equivalent to the O / R mapping framework 108, and includes a reference unit 159 and an update unit 160, and each has functions equivalent to the reference unit 109 and the update unit 110. . The tracer 164 has a function equivalent to that of the tracer 114, and the trace log 165 is a log output by the tracer 164.
Hereinafter, each part in FIG. 1 will be described.

  FIG. 5 is a diagram showing a flow of processing of a pessimistic / optimistic request according to the present embodiment. The process flow of the pessimistic / optimistic request of this embodiment will be described using a sequence.

  AP layer 1120 refers to container 126 and application 106 of FIG. The OR layer 1121 indicates the user object 107 and the O / R mapping framework 108 of FIG. The DB layer 1122 indicates the database server 113, DBMS 125, and table 124 of FIG.

  When the AP layer 1120 receives the request 1100, the AP layer 1120 outputs a trace log of the request start point 1101 indicating that processing of the request 1100 is started.

  When the request 1100 is passed from the AP layer to the OR layer 1121 and data is referred from the OR layer to the DB layer, the OR layer outputs a trace log of the reference start point 1102 indicating that the DB layer starts to be referenced. Further, when the reference of the DB layer is completed, the OR layer outputs a trace log of the reference completion point 1103 indicating that the reference is completed.

  When updating data from the OR layer to the DB layer, the OR layer outputs a trace log of the update start point 1104 indicating that the update of the DB layer is started. Further, when the update of the DB layer is completed, the OR layer outputs a trace log of the update completion point 1105 indicating that the update is completed. When the processing of the request 1100 is completed, the AP layer outputs a trace log of the request completion point 1106 indicating that the processing is completed.

  The time when the OR layer 1121 refers to the DB layer 1122, that is, the reference time 1107 is defined as a difference between the time when the reference completion point 1103 is acquired from the time when the reference start point 1102 is acquired. The time from when the reference of the OR layer 1121 is completed to the next update, that is, the intermediate time 1108 is defined as the difference between the time when the reference completion point 1103 is acquired and the time when the update start point is acquired. The time during which the OR layer 1121 updates the DB layer 1122, that is, the update time 1109 is defined as the difference between the time at which the update completion point 1105 is acquired from the time acquired at the update start point 1104. The waiting time 1111 is defined as the sum of the reference time 1107, the intermediate time 1108, and the update time 1109. The processing time 1110 is defined as the difference between the time when the request start point 1101 is acquired at the AP layer and the time when the request completion point 1106 is acquired. Components of the processing time 1110 include a reference time 1107, an intermediate time 1108, and an update time 1109.

  FIG. 6 is a diagram illustrating an operation example of the pessimistic lock format according to the present embodiment. The request A1200, the request B1210, and the request C1220 are examples in which the same data is accessed at the same time, and the processing execution order is the order of the request A1200, the request B1210, and the request C1220.

  A characteristic of the lock format of Pessimistic is that the lock is acquired from the start to the completion of the operation of the request database. That is, in the request A1200, a lock is acquired between the reference start point 1202 and the update completion point 1205 when accessing the database. At this time, the request B1210 tries to acquire the lock at the reference start point 1212 at the same time. However, since the request A1200 acquired the lock before the request B1210, until the reference is actually completed, that is, until the reference completion point 1213 is reached. In the meantime, the acquisition of the lock is awaited for the lock time 1230 of the request A 1200, that is, the time from the reference start point 1202 to the update completion point 1205.

  Request C1220 is awaited to be locked until request A1200 and request B1210 are unlocked, that is, the sum of the lock time of request A and the lock time of request B1210. Similarly, for subsequent requests, if a lock conflicts, the acquisition of the lock is awaited.

  The lock time 1230 is the difference between the reference start point 1202 at which access to the database is started and the update completion point 1205 at which the update is completed. When a lock collision occurs, in this case, in the request B, the lock time is the difference between the reference start point 1212 and the update completion point 1215 and the lock time 1230 of the request A. The lock waiting time 1240 is a time during which lock contention occurs and waiting for acquisition of a lock by another request. In request B, the lock waiting time corresponds to the lock time of request A. In request C, the lock waiting time corresponds to the sum of the lock time of request A and the lock time of request B.

  FIG. 7 is a diagram illustrating an operation example of the Optimistic lock format according to this embodiment. A request A 1300 and a request B 1310 indicate cases where the same data is accessed at the same time, and the processing execution order is the order of the request A 1300 and the request B 1310.

  The feature of the optimistic lock format is that a shared lock is acquired at the time of data reference, and at the time of data update, it is judged whether another request has acquired the shared lock, and the other request acquires the shared lock. If it is, the data update fails. In the failed request, the process is retried from the time of data reference.

  Request A 1300 is a reference start point 1302 and acquires a shared lock before request B 1310. Request B 1310 then acquires a shared lock at reference start point 1312. Request A 1300 fails to update data and releases the shared lock because other request B has acquired the shared lock. At this time, the AP layer outputs the update error point 1304 to the trace log, and retry from the data reference, that is, from the acquisition of the shared lock.

  When the request A1300 releases the shared lock, the request B1310 does not have any other request that holds the shared lock, so the data is actually updated and the process is completed. The request A 1300 retrying from the data reference again can acquire the shared lock after the update completion point 1315 where the request B releases the lock, and can complete the processing.

  At this time, the shared lock acquisition time acquired when the update is successful from the update error point 1304 of the request A 1300, that is, the difference time between the reference start points 1305 is defined as the longest retry time. The retry-free time is defined as the processing time when the request processing is completed without performing the retry processing. That is, the definition of the time without retry is a difference between the processing time 1316 and the wasted time 1317. The definition of the wasted time 1317 is a time when lock acquisition is wasted, including lock reacquisition, after a lock collision occurs after acquiring a shared lock.

  FIG. 8 is a diagram illustrating a flow example of the processing outline of the analysis unit 116 / determination unit 117 / setting unit 120 of the present embodiment. As shown in FIG. 8, the management server 100 refers to the server table 121 and executes the following processing of the analysis unit 116, the determination unit 117, and the setting unit 120 for all application servers (1400).

  First, the analysis unit 116 analyzes the trace log 115 and the mapping table 111 to create an operation correspondence table 123 and an object management table 122 (1401).

The lock determination unit 118 included in the determination unit 117 refers to the operation correspondence table 123 and the object management table 122 to determine an appropriate lock format for the object (1402).
The queue determination unit 119 constituting the determination unit 117 refers to the operation correspondence table 123, the object management table 122, and the queue table 104, and determines increase / decrease in the number of concurrent executions of queue request processing (1403).

Then, the setting unit 120 sets the lock format determined by the lock determination unit 118 to the O / R mapping framework 108, and informs the scheduler 102 about the increase / decrease in the number of simultaneous executions of the queue request processing determined by the queue determination unit 119. It sets for (1404). Thereafter, the process returns to the process 1401 of the analysis unit 116.
A detailed flow of the lock determination unit 118 will be described with reference to FIG. 9, and a detailed flow of the queue determination unit will be described with reference to FIG.

  FIG. 9 is a diagram illustrating a flow example of the lock determination unit 118 of the present embodiment. As shown in FIG. 9, the lock determination unit 118 determines whether the lock type of the object is pessimistic (302).

  When the object lock format is Pessimistic, (average processing time−average waiting time) × (1−collision rate) + (average processing time−average waiting time + average reference time + average intermediate time + retry waiting time) × collision The value obtained by the rate is compared with the value of the average processing time of Pessimistic (303).

  The idea of this equation is to predict the processing time in Optimistic when the lock format is Pessimistic. When the lock collision does not occur, the Optimistic processing time is predicted as (average processing time−average waiting time) because no waiting time occurs. In addition, since the waiting time does not occur for the Optimistic processing time when a lock collision occurs, there is a process that is wasted due to a lock collision and a lock reacquisition process, so (average processing time−average waiting time + (Average reference time + average intermediate time + retry waiting time)

  As a result of the comparison, if the average processing time of pessimistic is large, the lock determination unit 118 holds in the memory to instruct the O / R mapping framework to change the lock format of the object to Optimistic (304). . If the average processing time of pessimistic is equal or small, the determination process is terminated.

  On the other hand, if the lock type of the object is not pessimistic, the average processing time without retry + (average reference time + average intermediate time + average update time) × collision rate + average processing time without retry × (1−collision rate) It is determined whether the average processing time is long (306).

  When the average processing time of Optimistic is large, the lock determination unit 118 holds in the memory that the O / R mapping framework is instructed to change the lock format of the object to Pessimistic (307). If the average processing time of Optimistic is equal or small, the determination process is terminated.

  In this process, when the lock format is Optimistic, the processing time in Pseptic is predicted. That is, the pessimistic processing time when there is no lock collision does not cause a waiting time due to locking, so it is predicted as an average processing time without retry, and the processing time when a lock collision occurs is the waiting time due to locking. Is assumed to be (average reference time + average intermediate time + average update time), and the average processing time without retry + (average reference time + average intermediate time + average update time) is predicted.

  FIG. 10 is a diagram illustrating a flow example of the queue determination unit 119 according to the present embodiment. As shown in FIG. 10, the queue determination unit 119 first reads a threshold set by the user in the scheduler queue table (401). Then, it is determined whether the lock type of the object is pessimistic (402).

  If the lock type of the object is pessimistic, it is determined whether the threshold is larger than the average processing time (403). When the threshold value is larger than the average processing time, the queue determination unit 119 holds in the memory that an instruction to increase the number of simultaneous executions of the queue is issued to the scheduler (404).

  If the threshold is equal to or smaller than the average processing time, it is determined whether the threshold is larger than the average processing time without retry and smaller than the average processing time (406). The reason for the determination is that by narrowing down the queue, it is predicted that lock collisions will be reduced, the average processing time will be shortened, and it will be possible to approach the retry-free time.

  When the threshold value is larger than the average processing time without retry and smaller than the average processing time, the queue determination unit 119 holds in the memory that an instruction for reducing the number of simultaneous executions of the queue is issued to the scheduler (407). If the threshold is equal to or less than the average processing time without retry or equal to or longer than the average processing time, the determination process is terminated.

  On the other hand, if the lock format of the object is not pessimistic, it is determined whether the threshold is greater than the average processing time (408). When the threshold value is larger than the average processing time, the queue determination unit 119 holds in the memory that the instruction to increase the number of simultaneous executions of the queue is issued to the scheduler (409).

  If the threshold is equal to or smaller than the average processing time, it is determined whether the threshold is larger than the average processing time minus the average waiting time and smaller than the average processing time (410). The reason for the determination is that by narrowing down the queue, it is predicted that lock collisions are reduced, the average processing time is shortened, and the average processing time without waiting time can be approached.

  When the threshold value is larger than the average processing time-the average waiting time and smaller than the average processing time, the queue determination unit 119 holds in the memory that an instruction to reduce the number of simultaneous executions of the queue is issued to the scheduler (411). ). When the threshold does not satisfy that the average processing time is larger than the average processing time−the average waiting time and is smaller than the average processing time, the determination process is terminated.

  FIG. 11 is a diagram illustrating an example of a request flow according to the present embodiment. As shown in FIG. 11, when the load balancer 101 receives a request (501), the load balancer 101 allocates the request to the scheduler according to the rules of the load balancer itself (502).

  When there is information that can identify the user object that operates on the request, the scheduler queues the request to the queue corresponding to the operation of the request by referring to the queue table. Otherwise, the scheduler queues the request according to the queuing rule. Queue (503).

  When a request is queued in the queue, it is determined whether the maximum number of concurrent executions in the queue is smaller than the current number of concurrent executions (504). If the current number of concurrent executions in the queue is greater than or equal to the maximum number of concurrent executions, the scheduler waits for the processing of a new request until any one of the requests being concurrently executed by the queue is completed (505). If the current number of concurrent executions in the queue is less than the maximum number of concurrent executions, the scheduler queues the request in the queue and increments the current number of concurrent executions in the queue by one (506).

  The container retrieves the request from the queue (507), and the container uses the tracer to obtain a trace log indicating the request start processing (508). Then, the container passes the processing of the retrieved request to the application (509).

  The application makes a method call to the user object (510). The O / R mapping framework refers to the mapping table to know the table of the database server mapped to the user object (511).

  The container acquires a trace log indicating the start of table reference using the trace (512). The O / R mapping framework uses the reference unit to refer to the database server table via the database access unit (513). The container uses the trace to obtain a trace log indicating completion of the table reference (515).

  The container acquires a trace log indicating the start of table update using the trace (515). The O / R mapping framework uses the update unit to update the database server table via the database access unit (516). The container acquires a trace log indicating completion of table update using the trace (517). The O / R mapping framework sets the result for the user object (518).

  The user object returns the result to the application (519). The application returns the result to the container (520). The container acquires a trace log indicating completion of the request using the trace (521). The container returns the result to the transmission source of the request (522). The scheduler decreases the value of the current number of concurrent executions of the queue by one (523).

  FIG. 12 is a diagram showing the configuration of the mapping table 111 of this embodiment. As shown in FIG. 12, the mapping table 111 includes an object 601 and a lock format 602. The O / R mapping framework 108 acquires the lock format 602 of the user object 107 when accessing the database server 113 using the object 601 as a key. The analysis unit 116 is referred to in order to acquire the current lock format 602 of the user object 107 when creating the object management table 122.

  FIG. 13 is a diagram showing the configuration of the trace log 115 of this embodiment. As shown in FIG. 13, the trace log 115 includes a trace point 701, a time 702, a request serial number 703, an operation 704, an object 705, and an object ID 706.

  The trace point 701 includes a start point TRC00001 indicating the start of processing in the application server, a completion point TRC00006 indicating the completion of processing, a point TRC00002 indicating the start of data reference to the database in the O / R mapping framework, There are a point TRC00003 indicating the completion of reference, a point TRC00004 indicating the start of data update, and a point TRC00005 indicating the completion of update.

  Time 702 is the current time when the trace point 701 is acquired. Note that it is assumed that time adjustment is performed between servers constituting this system. In the request sequence number 703, a unique sequence number is assigned to one request by the tracer. By filtering with the request serial number, the flow of request processing can be confirmed.

  Operation 704 is information included in the request. For example, it refers to a URL or the like. Only trace points TRC00001 and TRC00006 can be obtained. An object 705 is an object that is mapped to database data that is actually manipulated in the processing of a request. The object ID 706 is a unique ID indicating the object 705. The object 705 and the object ID can be acquired at trace points TRC00002, TRC00003, TRC00004, and TRC00005.

  FIG. 14 is a diagram showing the configuration of the queue table 104 of this embodiment. As shown in FIG. 14, the queue table 104 includes a queue name 801, an operation 802, and a threshold value 803.

  The queue name 801 indicates the name of a queue managed by the scheduler. Operation 802 is information included in the requests queued in the queue. The threshold value 803 is information used by the queue determination unit and input by the user.

  FIG. 15 is a diagram showing the configuration of the operation correspondence table 123 of this embodiment. As shown in FIG. 15, the operation correspondence table 123 includes operations 901 and objects 902.

  Operation 901 is information included in the request. An object 902 is an object that is mapped to database data that is actually manipulated in the processing of a request. A plurality of objects may be manipulated for one operation.

  FIG. 16 is a diagram showing the configuration of the object management table 122 of this embodiment. As shown in FIG. 16, the object management table 122 includes an object 1001, a lock format 1002, an average processing time 1003, an average reference time 1004, an average intermediate time 1005, an average update time 1006, an average waiting time 1007, and a lock collision rate 1008. Is done.

  An object 1001 is an object that is mapped to database data that is actually manipulated in the processing of a request. The lock format 1002 is a lock format 602 acquired from the queue table using the object 1001 as a key.

  The average processing time 1003 is an average value of processing times up to the present time for the same object. The average reference time 1004 is an average value of reference times up to the present time for the same object. The average intermediate time 1005 is an average value of the intermediate time up to the present in the same object. The average update time 1006 is an average value of update times up to the present time for the same object.

  The average waiting time 1007 is an average value of waiting times up to the present time for the same object. The lock collision rate 1008 is a lock collision rate for the same object. The average processing time without retry 1009 is the average time when a request is processed without causing a lock collision when the lock type is Opsmistic in the same object, and indicates the time to be retried when a lock collision occurs. It does not include time.

  FIG. 17 is a diagram showing the configuration of the server table 121 of this embodiment. As shown in FIG. 17, the server table 1703 includes a server name 1701 and a scheduler 1702. The server name 1701 indicates a server group managed by the management server 100. When the setting unit 120 sets the lock format or the queue concurrent execution number, the scheduler 102 to be set is acquired using the server name 1701 as a key.

  The information processing apparatus according to this embodiment includes the application server 105 in FIG. 2, the management server 100 in FIG. 3, and the database server 113.

  FIG. 2 is a diagram illustrating a hardware configuration of the application server 105 according to the present embodiment. 2, the application server 105 includes a memory 202, a CPU 201, a storage device 203, and a communication interface 204. In the memory 202, a container 126, a queue table 104, a scheduler 102, a queue 103, an application 106, a mapping table 111, a user object 107, a database access unit 112, a tracer 114, an O / R mapping framework 108, a reference unit 109, an update The units 110 are stored, and each is executed by the CPU 201. A trace log 115 is stored in the storage device 203 and loaded into the memory 202 as necessary.

  FIG. 3 is a diagram illustrating a hardware configuration of the management server 100 according to the present embodiment. In FIG. 3, the management server 100 includes a memory 233, a CPU 242, and a communication interface 231. The memory 233 stores a determination unit 117, a lock determination unit 118, a queue determination unit 119, an analysis unit 116, a setting unit 120, an operation correspondence table 123, an object management table 122, and a server table 121, each of which is executed by the CPU 242. Is done.

  In the present embodiment, a program for causing a computer to function as each processing unit such as the analysis unit 116 is recorded on a recording medium such as a CD-ROM and stored in a magnetic disk or the like, and then loaded into a memory and executed. Shall. The recording medium for recording the program may be a recording medium other than the CD-ROM. Further, the program may be used by installing it from a recording medium in a computer, or the program may be used by accessing the recording medium through a network.

  FIG. 4 is a diagram showing a hardware configuration of the database server 113 of this embodiment. In FIG. 4, the database server 113 includes a memory 263, a CPU 261, a storage device 264, and a communication interface 262. A DBMS 125 is stored in the memory 263 and is executed by the CPU 261. The storage device 264 stores a database table 124 and is loaded into the memory 263 as necessary.

  In the prior art, the lock format is manually set for the O / R mapping framework according to the application characteristics, system configuration, and performance requirements, and the lock format is a definition file for the O / R mapping framework. It was necessary for the user to describe the settings manually.

  In this embodiment, since the management server automatically determines and sets an appropriate lock format, it is not necessary for the user to manually write a definition file, and the setting work of the O / R mapping framework can be reduced.

  As described above, according to the database processing apparatus of the present embodiment, the lock format of the object is determined by analyzing the trace log and the mapping table, and the lock format is set. It is possible to dynamically set an appropriate lock format for the O / R mapping framework.

  Further, according to the database processing apparatus of this embodiment, the trace log and the mapping table are analyzed to determine the increase / decrease in the number of simultaneous executions of queue request processing, and the number of simultaneous executions is set. It is possible to improve.

It is a figure which shows the structure of the system which manages the some application server of this embodiment. It is a figure which shows the hardware constitutions of the application server 105 of this embodiment. It is a figure which shows the hardware constitutions of the management server 100 of this embodiment. It is a figure which shows the hardware constitutions of the database server 113 of this embodiment. It is a figure which shows the flow of a process of the pessimistic / optimistic request of this embodiment. It is a figure which shows the operation example of the pessimistic lock format of this embodiment. It is a figure which shows the operation example of the Optimistic lock format of this embodiment. It is a figure which shows the example of a flow of the process outline | summary of the analysis part 116 / determination part 117 / setting part 120 of this embodiment. It is a figure which shows the example of a flow of the lock determination part 118 of this embodiment. It is a figure which shows the example of a flow of the queue determination part 119 of this embodiment. It is a figure which shows the example of the request flow of this embodiment. It is a figure which shows the structure of the mapping table 111 of this embodiment. It is a figure which shows the structure of the trace log 115 of this embodiment. It is a figure which shows the structure of the queue table 104 of this embodiment. It is a figure which shows the structure of the operation corresponding | compatible table 123 of this embodiment. It is a figure which shows the structure of the object management table 122 of this embodiment. It is a figure which shows the structure of the server table 121 of this embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 100 ... Management server, 101 ... Load balancer, 102 ... Scheduler, 103 ... Queue, 104 ... Queue table, 105 ... Application server, 106 ... Application, 107 ... User object, 108 ... O / R mapping framework, 109 ... Reference part , 110 ... update section, 111 ... mapping table, 112 ... database access section, 113 ... database server, 114 ... tracer, 115 ... trace log, 116 ... analysis section, 117 ... determination section, 118 ... lock determination section, 119 ... queue Judgment unit, 120 ... setting unit, 121 ... server table, 122 ... object management table, 123 ... operation correspondence table, 124 ... table, 125 ... DBMS, 126 ... container, 152 ... scheduler, 153 ... queue 154 ... Queue table, 156 ... Application, 157 ... User object, 158 ... O / R mapping framework, 159 ... Reference unit, 160 ... Update unit, 161 ... Mapping table, 162 ... Database access unit, 164 ... Tracer, 165 ... Trace log, 176 ... Container, 201 ... CPU, 202 ... Memory, 203 ... Storage device, 204 ... Communication interface, 231 ... Communication interface, 233 ... Memory, 242 ... CPU, 261 ... CPU, 262 ... Communication interface, 263 ... Memory, 264 ... Storage device, 1100 ... Request, 1101 ... Request start point, 1102 ... Reference start point, 1103 ... Reference completion point, 1104 ... Update start point, 1105 ... Update completion point, 11 6 ... Request completion point 1107 ... Reference time 1108 ... Intermediate time 1109 ... Update time 1110 ... Processing time 1111 ... Wait time 1120 ... AP layer 1121 ... OR layer 1122 ... DB layer 1200 ... Request A 1202 ... Reference start point, 1205 ... Update completion point, 1210 ... Request B, 1212 ... Reference start point, 1213 ... Reference completion point, 1215 ... Update completion point, 1220 ... Request C, 1230 ... Lock time, 1240 ... Waiting for lock Time, 1300 ... Request A, 1302 ... Reference start point, 1304 ... Update error point, 1305 ... Reference start point, 1310 ... Request B, 1312 ... Reference start point, 1315 ... Update completion point, 1316 ... Processing time, 1317 ... Waste Time, 601 ... Object, 602 ... Lock type, 701 ... Trace point, 702 ... Time, 703 ... Request serial number, 704 ... Operation, 705 ... Object, 706 ... Object ID, 801 ... Queue name, 802 ... Operation, 803 ... Threshold 901 ... Operation, 902 ... Object, 1001 ... Object, 1002 ... Lock type, 1003 ... Average processing time, 1004 ... Average reference time, 1005 ... Average intermediate time, 1006 ... Average update time, 1007 ... Average waiting time, 1008 ... Lock collision rate, 1009 ... Average processing time without retry, 1701 ... Server name, 1702 ... Scheduler, 1703 ... Server table.

Claims (10)

The database device manages the database,
A plurality of program execution devices (application servers) have a first storage device, access the database device by executing a program based on the accepted operation, and realize a service,
The management device has a second storage device, manages the plurality of program execution devices,
The program execution device stores a trace log of the program including the operation and a mapping table storing a lock format used when accessing the database used by the program in the first storage device,
The management device analyzes the trace log and the mapping table from the first storage device, and associates the operation, a program executed based on the operation, and the trace log of the program. And storing the operation in the second storage device, referring to the stored operation, a program executed based on the operation, and a trace log of the program, and evaluating the trace log for the operation for each lock type. And determining the lock format used by the program execution device and setting the determined lock format to the plurality of program execution devices.   The management device compares the average processing time in the first lock format set as the object lock format with the average processing time in the second lock format not set as the object lock format, and locks the object The database processing method according to claim 1, wherein format determination is performed.   3. The database according to claim 2, wherein the management device calculates an average processing time in a second lock format not set as an object lock format according to a predetermined formula and compares the average processing times. Processing method.   4. The database processing method according to claim 2, wherein the lock format in which the management device performs the determination is a pessimistic lock format and an optimistic lock format.   When the first lock format set as the object lock format is the pessimistic lock format, the management device performs an average process in the Optimistic lock format that is the second lock format not set as the object lock format. The time is calculated according to the formula of (average processing time−average waiting time) × (1−collision rate) + (average processing time−average waiting time + average reference time + average intermediate time + retry waiting time) × collision rate. 5. The database processing method according to claim 4, wherein average processing times are compared.   When the first lock format set as the object lock format is the Optimistic lock format, the management device performs an average process in the pessimistic lock format that is the second lock format not set as the object lock format. The time is calculated according to the following formula: average processing time without retry + (average reference time + average intermediate time + average update time) × collision rate + average processing time without retry × (1−collision rate) The database processing method according to claim 4, wherein the database processing method is performed.   The management device reads the operation correspondence table, the object management table, and the queue table from the storage device to determine increase / decrease in the simultaneous execution number of the queue request processing, and the management device determines the simultaneous execution number of the queue request processing. The database processing method according to claim 1, wherein the increase / decrease is transmitted to the application server by the communication device to set the number of simultaneous executions of request processing.   8. The database processing method according to claim 7, wherein the management device determines an increase or decrease in the number of simultaneous executions of request processing by comparing an average processing time in a set lock format with a predetermined threshold value. The database device manages the database,
A plurality of program execution devices (application servers) have a first storage device, access the database device by executing a program based on the accepted operation, and realize a service,
The management device includes a second storage device, manages the plurality of program execution devices,
The program execution device stores a trace log of the program including the operation and a mapping table storing a lock format used when accessing the database used by the program in the first storage device,
The management device analyzes the trace log and the mapping table from the first storage device, and associates the operation, a program executed based on the operation, and the trace log of the program. And storing the operation in the second storage device, referring to the stored operation, a program executed based on the operation, and a trace log of the program, and evaluating the trace log for the operation for each lock type. And determining the lock format used in the program execution device and setting the determined lock format in the plurality of program execution devices. The database device manages the database,
A plurality of program execution devices (application servers) have a first storage device, access the database device by executing a program based on the accepted operation, and realize a service,
The management device includes a second storage device, manages the plurality of program execution devices,
The program execution device stores a trace log of the program including the operation and a mapping table storing a lock format used when accessing the database used by the program in the first storage device,
The management device analyzes the trace log and the mapping table from the first storage device, and associates the operation, a program executed based on the operation, and the trace log of the program. And storing the operation in the second storage device, referring to the stored operation, a program executed based on the operation, and a trace log of the program, and evaluating the trace log for the operation for each lock type. A program for causing a computer to execute a database processing method for determining the lock format used in the program execution device and setting the determined lock format in the plurality of program execution devices.

Images