JP2856150B2 - Transaction history recording system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a technology for recording transaction history data for failure recovery in an online transaction processing system, and more particularly to a high-speed recording of transaction history data in a large-scale online transaction processing system requiring high reliability. Technology for making

[0002]

2. Description of the Related Art Conventionally, in this type of online transaction processing system, a physical I / O operation to an external storage device at the time of recording transaction history data for failure recovery is the largest bottleneck in transaction processing performance. Has become. Therefore, in order to improve the throughput performance and the turnaround performance of the online transaction processing system, how to speed up the recording of the transaction history data is the most important issue.

As a conventional technique for speeding up the recording of transaction history data, a method of mounting a plurality of external storage devices and recording transaction history data in parallel is generally known.

FIG. 7 is a diagram for conceptually explaining such a conventional technique, and shows the relationship between the number of physical devices in a transaction history file and the recording time of transaction history data.

In FIG. 7, the time required for physically writing one transaction to the transaction history file is 100 [ms], and the number of transactions arriving per second is 50 [messages / second] (the average transaction arrival interval is 20 [ms] ]).

The vertical direction indicates the passage of time, and arrows 711 to 714 and arrows 741 to 744 indicate the writing period (100 [m
s]). Further, arrows 721 to 724 and arrows 751 to 754 indicate transaction arrival times, and arrows 731 to 734 and arrows 761 to 764 indicate completion times of the respective transaction processes.

As shown in FIG. 7A, when the number of transaction history files is "1", the time required for recording four transactions is represented by the following equation (1).

100 × 4 = 400 [ms] (1)

On the other hand, as shown in FIG. 7B, when the number of transaction history files is "2", 4
The time required to record a transaction is represented by the following equation (2).

100 × 2 + 20 = 220 [ms] (2)

Therefore, the recording time per transaction is represented by the following equation (3), and the recording time of apparent transaction history data is reduced.

220/4 = 55 [ms] (3)

As another conventional technique for speeding up the recording of transaction history data, for example, Japanese Patent Application Laid-Open No. Hei 2-176949 discloses a journal on a semiconductor memory device which can be accessed faster than an external memory device. A method of providing a data buffer has been proposed.

FIG. 8 is a diagram for explaining the outline of the conventional method described in the above-mentioned Japanese Patent Application Laid-Open No. 2-176949.

Referring to FIG. 8, in this conventional method,
In addition to a journal file (transaction history file) 84 on an external storage device represented by a magnetic disk device, a journal data buffer (transaction history buffer) 88 using a semiconductor storage device is provided.

When a transaction is received by the online processing unit 82 in the CPU 81, the transaction history data is sequentially recorded in a transaction history buffer 88. Then, when the remaining amount of the transaction history buffer 88 falls below a certain value, the transfer control means 87 in the control unit 86 transfers the transaction history data recorded in the transaction history buffer 88 to the transaction history file 84.

As a result, the number of physical I / O operations on the transaction history file 84 can be reduced, and transaction history data can be recorded at high speed.

[0018]

However, in the above-mentioned conventional method of mounting a plurality of generally-known transaction history files, the expansion of hardware resources such as the addition of an external storage device in order to improve the transaction processing performance is required. In addition, since the system configuration becomes large-scale or complicated, there is a problem that it is very costly to realize high transaction processing performance.

In the conventional method described in Japanese Patent Application Laid-Open No. 2-176949, high speed is realized at the expense of reliability, so that it cannot be applied to an online transaction processing system requiring high reliability. There is a problem.

That is, in the conventional method described in the above-mentioned Japanese Patent Application Laid-Open No. 2-176949, transaction history data is recorded in a buffer on a semiconductor storage device, and the transaction history data is recorded in a transaction history file in which non-volatility is guaranteed. Since recording is not performed until the buffer becomes full, transaction history data that has not been written to the transaction history file will be lost when a severe system failure such as a system down occurs.

Accordingly, the present invention has been made in view of the above-described problems, and has been developed in an online transaction processing system requiring high performance and high reliability.
An object of the present invention is to provide a transaction history recording system capable of speeding up the recording of transaction history data without lowering the reliability and realizing this at low cost.

[0022]

In order to achieve the above-mentioned object, a transaction history recording system according to the present invention comprises:
An online transaction processing system including one or a plurality of transaction history files (15 in FIG. 1) and recovering transaction data using transaction history data stored in the transaction history file (15 in FIG. 1) when a failure occurs. ,
In addition to the conventional configuration of the transaction history file (15 in FIG. 1) and the online processing unit (11 in FIG. 1), the transaction is stored in a storage area (shared memory (16 in FIG. 1)) that can be shared by a plurality of processes. A transaction concatenation means (12 in FIG. 1) having a storage table (17 in FIG. 1) for linking transaction data received simultaneously from a plurality of terminals with a pointer chain in the transaction storage table (17 in FIG. 1); The transaction data linked in a chain on the storage table (17 in FIG. 1) is stored in a transaction history file (FIG. 1) by one physical write operation.
(15) in FIG. 1).

Further, the transaction history recording system of the present invention monitors the operation of the history recording means (13 in FIG. 1), and if all the transaction history files (15 in FIG. 1) are being written, the transaction linking means. A connection suppressing means (14 in FIG. 1) for stopping the connection operation of transaction data by (12 in FIG. 1) is further provided.

With the above-described configuration, according to the transaction history recording system of the present invention, the online processing unit (11 in FIG. 1) monitors arrival of transactions from a plurality of terminals. When the transaction arrives,
All the transaction data that has arrived is stored in the transaction storage table (17 in FIG. 1) on the shared memory (16 in FIG. 1), and the transaction coupling means (12 in FIG. 1) is notified of the occurrence of a new transaction.

Transaction connection means (12 in FIG. 1)
Connects the newly arrived transaction data on the transaction storage table (17 in FIG. 1) by a pointer chain, and requests the history recording means (13 in FIG. 1) to write the chained transaction data.

The history recording means (13 in FIG. 1) receives a write request from the transaction concatenation means (12 in FIG. 1), and performs a physical write operation on all chained transaction data in one transaction history file (FIG. 1). Record in 1) 15). When the writing is completed, the history recording unit (13 in FIG. 1) notifies the transaction linking unit (12 in FIG. 1) of the writing result.

The connection suppressing means (14 in FIG. 1) monitors the operation of the history recording means (13 in FIG. 1). If all of the mounted transaction history files (15 in FIG. 1) are being written. The transaction linking means (Fig. 1)
The connecting operation according to 12) is stopped. When writing of one or more transaction history files among the mounted transaction history files (15 in FIG. 1) is completed, the connection operation by the transaction connection means (12 in FIG. 1) is restarted.

With the above operation, the transaction history file (15 in FIG. 1) can be obtained by one physical write operation for a plurality of transactions arriving during the writing operation of the transaction history file (15 in FIG. 1).
Can be completed. This allows
The number of physical I / O operations to the external storage device when the transaction load increases is greatly reduced, and the throughput performance and turnaround performance under a high transaction load can be improved.

Further, when all of the mounted transaction history files (15 in FIG. 1) are being written, the connection operation by the transaction connection means (12 in FIG. 1) is stopped, so that a newly arrived transaction is unnecessary. And the physical I / O to the transaction history file (15 in FIG. 1) according to the write state of the transaction history file (15 in FIG. 1).
The number of operations can be optimized.

[0030]

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

FIG. 1 is a block diagram showing the overall configuration of the embodiment of the present invention. FIG. 2 is a diagram for explaining the operation of the embodiment of the present invention.

Referring to FIG. 1 and FIG. 2, the online processing unit 11 holds a plurality of communication paths 21 one-to-one corresponding to terminals and a path management table 22.

The path management table 22 requests the operating system (OS) to monitor an event for each communication path.

The transaction storage table 17 is
It is implemented on a storage area (shared memory 16) that can be referred to by a plurality of processes. The transaction storage table 17 holds a terminal record corresponding to the path management table 22 and control information for managing a pointer chain.

The history recording means 13 includes a write buffer 23 for physically writing the transaction history file 15.

Next, the operation of the embodiment of the present invention will be described.

Referring to FIG. 2, the online processing unit 11
Monitors the arrival of a transaction using the path management table 22. For the communication path where the transaction arrived from the terminal, the operating system (O
By S), a data reception event is notified to the event area of the path management table 22.

The online processing section 11 notifies the transaction concatenating means 12 of the arrival of a new transaction, searches the path management table 22 for each communication path in which the data reception event is set, and searches the transaction storage table 17 The received data is stored in the terminal record corresponding to the communication path.

The transaction connection means 12 receives the notification of the data reception event and connects the transaction data.

FIG. 3 is a flowchart for explaining the operation of connecting transaction data by the transaction connection means 12.

Referring to FIGS. 2 and 3, first, the address of the chain head information storage area is set as an initial value of the next terminal position storage address (step 301).

Next, the communication path in which the reception event has occurred is searched from the path management table 22 (step 30).
2) It is confirmed whether or not there is a found communication path (step 303).

If the communication path is found (step 303), the terminal record corresponding to the found communication path is recognized, and the start address of this terminal record is set in the area indicated by the next terminal position storage address. (Step 30
4). Here, since the address of the chain head information storage area is set as the initial value of the next terminal position storage address in step 301, the start address of the current terminal record is stored in the chain head information storage area.

Next, the next terminal location storage address is changed to the address of the chain information storage area of the current terminal record (step 305), and the processing is repeated from step 302. By doing so, the next step 30
In the process of No. 4, the start address of the terminal record corresponding to the communication path searched next is stored in the chain information storage area of the current terminal record.

While the communication path in which the reception event has occurred can be searched, the above-described steps 302 to 3 are executed.
When the process of step 05 is repeated and there are no more communication paths to be searched (step 303), the chain terminal symbol is stored in the area indicated by the next terminal position storage address (step 306).

By the above operation, starting from the chain head information storage area, a pointer chain connecting the chain information storage areas of the terminal records at which the transaction has arrived is constructed.

Referring to FIG. 2, when the connection operation is completed as described above, the transaction connection means 12
Notifies the history recording means 13 of the address of the chain head information storage area, and requests writing of transaction history data.

Upon receipt of the transaction history data write request, the history recording means 13 reconfigures the transaction data chained by the procedure shown in FIG.

FIG. 4 is a flowchart for explaining the operation of reconstructing the chain-linked transaction data by the history recording means 13.

Referring to FIGS. 2 and 4, first, the write buffer 23 is initialized (step 401).
Next, the terminal record pointed to by the address stored in the chain head information storage area is set as an initial value of the terminal record being processed (step 402).

Then, after transferring the contents of the received data storage area of the terminal record being processed to an empty area of the writing buffer 23 (step 403), it is determined whether or not the chain information of the terminal record being processed = chain end symbol. (Step 404), and if it is not a chain terminal symbol, the process proceeds to step 405 to recognize the terminal record indicated by the address stored in the chain information storage area of the terminal record being processed as a new terminal record being processed. Then, the processing from step 403 is repeated.

In step 404, if a chain terminator is detected in the chain information storage area of the terminal record being processed, the reconstruction of the chained transaction data is terminated.

Referring to FIG. 2, the history recording means 13
The plurality of transaction data reconstructed on the write buffer 23 is stored in the transaction history file 1.
5 is recorded by one physical write operation.
When the recording of the transaction history data is completed, the history recording unit 13 notifies the transaction connection unit 12 of the write result. Further, the transaction connection unit 12 returns the notified write result to the online processing unit 11.

Further, the connection suppressing means 14 monitors the operation state of the history recording means 13, and performs a physical write operation on all the transaction history files 15 implemented in the system by the history recording means 13. , The connection operation by the transaction connection means 12 is stopped. Thereafter, when the writing operation for one or more transaction history files 15 is completed, the connection suppressing unit 14 restarts the connecting operation by the transaction connecting unit 12.

[0055]

Next, in order to more specifically describe the above-described embodiment of the present invention, an embodiment of the present invention will be described in detail with reference to the drawings.

[0056]

Embodiment 1 First, a first embodiment of the present invention will be described.

FIG. 5 is a diagram for explaining a first embodiment of the present invention. In this embodiment, UNIX
It is assumed that a transaction history recording system is constructed using an operating system (OS).

Referring to FIG. 5, the online processing unit 11 and the transaction connection unit 1 shown in FIGS.
The respective mechanisms of the online processing unit 11a and the transaction connection unit 12a having the same functions as those of No. 2 are mounted on one transaction processing process (# 1) 51a.

In the case of an operating system (OS) such as UNIX, the maximum number of communication paths per process is limited by the specifications of the OS. Enables application to large-scale online transaction processing systems. In the case of FIG. 5, by connecting two processes (transaction processing processes (# 1) 51a and (# 2) 51b), connection with about 500 terminals is enabled.

With the multiplexing of the online processing units 11a and 11b and the transaction connection units 12a and 12b, the transaction storage table 1 on the shared memory
7a and 17b are also distributed and implemented by the number of processes.

The history recording means 13 shown in FIGS. 1 and 2
The history recording unit 13a having the same function as that described above is mounted on an independent transaction history recording process (# 1) 53a, and is associated one-to-one with the transaction history file 15a using a magnetic disk or the like. In the case of FIG. 5, two sets of the history recording unit and the transaction history file (the history recording unit 13a and the transaction history file 15a, the history recording unit 13b and the transaction history file 15b) are mounted, and the transaction processing process (# 1) is performed. ) 51a, (# 2) 51b and transaction history recording process (# 1) 53a, (# 2) 53b
By using the message queue 52 as a communication method between the transaction history data and the conventional technology, a method of reducing the average recording time of transaction history data is applied in the same manner as in the above-described conventional technique.

As a component corresponding to the connection suppressing means 14 shown in FIGS. 1 and 2, in the present embodiment, the response receiving message queue 54a is one-to-one associated with the transaction connecting means 12a and 12b. , 54b.

Next, the operation of the first embodiment of the present invention will be described.

Referring to FIG. 5, the transaction processing processes (# 1) 51a and (# 2) 51b monitor the transactions arriving from the terminals connected thereto, and store the received data in the transaction storage tables 17a and 17a.
7b. At the same time, according to the procedure shown in FIG.
b, a pointer chain is formed, and a transaction history data recording request is transmitted to the request transmission message queue 52. Transaction processing process (#
1) When 51a and (# 2) 51b transmit a transaction history data recording request, they perform a read operation on the respective response receiving message queues 54a and 54b. As a result, the transaction processing processes (# 1) 51a and (# 2) 51b perform the transaction processing until the transaction history data recording operation in the transaction history recording processes (# 1) 53a and (# 2) 53b is completed. The data concatenation operation will be blocked.

On the other hand, the transaction processing process (#
1) The transaction history data recording request transmitted by 51a and (# 2) 51b is a first-in first-out (F
Message queue 5 for request transmission in accordance with the principle of
2 is stored.

The transaction history recording process (#
1) 53a and (# 2) 53b are each performing a read operation from the same request transmission message queue 52. The recording request of the transaction history data arriving at the request transmission message queue 52 is received by one of the transaction history processes (# 1) 53a and (# 2) 53b, and the transactions are collectively processed according to the procedure shown in FIG. History file 15a or 1
5b.

When the writing of the transaction history data is completed, the transaction history recording process (#
1) 53a or (# 2) 53b writes to the response receiving message queue 54a or 54b corresponding to the transaction processing process (# 1) 51a or (# 2) 51b that is the source of the transaction history data recording request. Send the resulting response.

[0068]

Embodiment 2 Next, a second embodiment of the present invention will be described.

FIG. 6 is a diagram for explaining a second embodiment of the present invention.

Referring to FIG. 6, in the present embodiment,
Transaction connection means 12 shown in FIGS. 1 and 2
Are composed of a two-stage connecting unit of intra-process connecting means 12a 'and 12b' and inter-process connecting means 61.

Further, the transaction processing processes (# 1) 51a and (# 2) 51b and the transaction history recording processes (# 1) 53a and (# 2) 53b which used the message queue in the first embodiment described above. Is changed to a connection-type communication method (socket), and communication is performed via the secondary connection process 65.

Further, as a modification of the transaction storing table 17 shown in FIGS. 1 and 2, a recording request linking table 62 is provided on a shared memory.

In addition, as the connection suppressing means 14 shown in FIGS.
Then, the file status table 63 on the shared memory is mounted. Note that the writable device counter 64 is a counting semaphore whose initial value is the number of mounted transaction history files.

The transaction history recording process (#
1) When performing a physical write operation on the magnetic disk, the 53a and (# 2) 53b lock the writable device counter 64 (subtract "1") and change the file status in the file status table 63 to "BUSY". Change to When the writing to the magnetic disk is completed, the file status in the file status table 63 is returned to “IDLE”, and the writable device counter 6 is returned.
4 is unlocked ("1" is added). By performing this control, the writable device counter 6 is written while all the mounted transaction history files are being written.
The semaphore value of 4 is “0”.

Transaction processing process (# 1) 5
1a and (# 2) 51b respectively link the received transaction data with a pointer chain, and then issue a transaction history data recording request in the secondary connection process 6.
5 is transmitted. The secondary connection process 65 includes a transaction processing process (# 1) 51a, (# 2) 51
Prior to monitoring the arrival of the transaction history data recording request from b, the lock operation of the writable device counter 64 is performed.

Here, when all the transaction history files 15a and 15b are being written, the processing of the secondary connection process 65 is blocked, but the transaction processing processes (# 1) 51a and (# 2) during this time as well. 5
1b, the process of not transmitting the transaction history data recording request can connect the received transaction data with the pointer chain, and transmit the transaction history data recording request to the secondary connection process 65.

When the writing of one or more transaction history files 15a and 15b is completed, the secondary connection process 65 can continue the processing. Therefore, after unlocking the writable device counter 64, the transaction processing process ( (# 1) 51a, (# 2) 51b
It monitors the arrival of transaction history data recording requests from. The secondary linking process 65 uses the record request linking table 62 to link the pointers of the transaction history data recording requests according to the procedure shown in FIG. 3, and in the file status table 63, the transaction history with the file status “IDLE”. A chain connection recording request is transmitted to the recording process (# 1) 53a or (# 2) 53b.

Transaction history recording process (#
1) 53a and (# 2) 53b reconstruct a connected transaction by following the recording request connection chain and the transaction connection chain in accordance with the procedure shown in FIG. 4, and perform the transaction by one physical write operation. Recording can be performed in the history files 15a and 15b.

When the writing is completed, the transaction history recording processes (# 1) 53a and (# 2) 53b notify the secondary connection process 65 of the writing result. Further, the secondary connection process 65 performs the respective transaction processing processes (# 1) 5 of the transmission source according to the recording request connection chain
1a and (# 2) 51b are notified of the write result.

[0080]

As described above, according to the present invention,
Without breaking the principle of recording transaction history data on a magnetic disk that is guaranteed to be non-volatile,
Since the number of the most time-consuming physical write operations can be reduced, a plurality of chained transaction data can be recorded in a transaction history file by one physical write operation, so that the transaction can be performed without lowering the reliability. It is possible to speed up the recording of the history data.

Further, according to the present invention, unlike the conventional method of mounting a plurality of transaction history files, it is not necessary to add an external storage device. Therefore, the conventional method of simply mounting a plurality of transaction history files is not required. In comparison, the transaction processing performance can be improved at a lower cost.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment of the present invention.

FIG. 2 is a diagram for explaining an operation of the exemplary embodiment of the present invention.

FIG. 3 is a flowchart for explaining a transaction data connection operation by a transaction connection unit according to the embodiment of the present invention.

FIG. 4 is a flowchart for explaining a reconfiguration operation of chain-linked transaction data by the history recording unit according to the embodiment of the present invention.

FIG. 5 is a diagram for explaining a first embodiment of the present invention.

FIG. 6 is a diagram for explaining a second embodiment of the present invention.

FIG. 7 is a diagram conceptually illustrating an example of a conventional technique (showing the relationship between the number of physical devices in a transaction history file and the recording time of transaction history data).

FIG. 8 shows a conventional method (Japanese Patent Laid-Open No. 2-176949).
It is a figure for explaining an outline of.

[Explanation of symbols]

11, 11a, 11b Online processing unit 12, 12a, 12b Transaction connection means 12a ', 12b' In-process connection means 13, 13a, 13b History recording means 14 Connection suppression means 15, 15a, 15b Transaction history file 16 Shared memory 17, 17a, 17b Transaction storage table 21 Communication path 22 Path management table 23 Write buffer 51a, 51b Transaction processing process # 1,
# 2 52 Request message queue 53a, 53b Transaction history recording process #
1, # 2 54a, 54b Message queue for response reception 61 Inter-process connection means 62 Recording request connection table 63 File status table 64 Writable device counter 65 Secondary connection process 81 CPU 82 Online processing unit 84 Journal file (transaction history file) 86 control unit 87 transfer control means 88 journal data buffer (transaction history buffer)

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-62-237540 (JP, A) JP-A-5-127966 (JP, A) JP-A-7-44440 (JP, A) JP-A-2- 176949 (JP, A) JP-A-3-20839 (JP, A) JP-A-4-205632 (JP, A) JP-A-4-209044 (JP, A) JP-A-4-260946 (JP, A) JP-A-5-313977 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) G06F 12/00

Claims (3)

(57) [Claims] 1. An online transaction processing system comprising one or a plurality of transaction history files and recovering transaction data using transaction history data stored in the transaction history file when a failure occurs. A transaction linking means for linking transaction data received simultaneously from a plurality of terminals with a pointer chain in the transaction storage table; and includes a history recording unit configured by times of physical write operations write to the transaction history file, and allows sharing the predetermined storage area among a plurality of processes
And the transaction concatenation means comprises a plurality of transactions.
In-process connection means for the
Consists of a two-stage connecting unit of inter-process connecting means
Transaction history record systems that are composed, characterized in that the. 2. A system according to claim 1, further comprising a connection suppressing unit that monitors an operation of said history recording unit and stops a transaction data connecting operation by said transaction connecting unit when all of said one or a plurality of transaction history files are being written. The transaction history recording system according to claim 1, wherein 3. The method according to claim 1, wherein the connection suppressing means uses a counting semaphore.
3. The transaction history recording system according to claim 2 , wherein the transaction history recording system is configured as follows .