JP3531469B2 - Transaction control method and its relay device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a transaction parallel technique for realizing load distribution.

[0002]

2. Description of the Related Art A conventional online processing system consisting of a CPU processing unit having a single CPU is migrated to a plurality of Cs.
When transaction load is distributed to PUs, if the business process from the terminal is divided into multiple transactions, each transaction will be processed by another CPU, and transaction order control and processing information transfer between CPUs will be performed. Therefore, shared memory was required.

Further, in order to distribute the load of transactions from a terminal to a plurality of CPUs, Japanese Patent Laid-Open No. 5-2 is known.
Additional information is required in addition to the transaction, such as setting of parameter information of distribution destination like 74261, addition of control information data for load distribution like Japanese Patent Laid-Open No. 7-221754, and conventional online processing and terminals. The treatment could not be used as is.

[0004]

In order to realize the load distribution of transactions among a plurality of CPUs in a conventional online processing system, it is necessary to control the order of transactions among the CPUs, which causes overhead and A CPU processor with shared memory was needed.

Further, when the transaction used in the conventional online processing system of the CPU processing unit having a single CPU is migrated to distribute the load of the transaction from the terminal to a plurality of CPUs, the parameter information of the distribution destination is used. And the control data for load balancing had to be added.

Further, since the transaction parallel processing is a technology premised on a plurality of CPU processing devices and a plurality of online processings, the influence of competition of branch transactions to a plurality of online independent terminals is as follows. It was not considered in the prior art.

An object of the present invention is to eliminate the overhead due to order control of transactions among a plurality of CPUs,
In addition to speeding up the transaction processing of the entire online processing system, the failure is localized only to the business processing in progress when the failure occurs, and it is used in the conventional online processing system consisting of a CPU processing device having a single CPU. The purpose of this is to realize transaction parallelization without affecting the terminal processing and online processing that were previously used.

In order to solve the above problems, a transaction control method in a relay device for relaying a CPU processing device having a plurality of CPUs and a terminal of the present invention is the CP control method described above.
In response to the reception of a block request transaction indicating the beginning of a branch transaction from U, the block request is suspended if the in-progress state corresponding to another CPU is the branching state, and the block request is not in the branching state. Sent C
The WIP state corresponding to the PU and in the branch, in response to said receiving branch response last transaction indicating the end of the branch transaction from the terminal, the WIP initial state corresponding to the CPU that sent the closure request .

Further, the load of online processing is obtained from each CPU through the communication means, so that the CP having a small load can be obtained.
U is specified, and the transaction is distributed to each business processing unit.

In order to minimize the influence on the terminal in the case of a CPU processing device failure, a communication failure with the CPU processing apparatus, an online processing failure, etc., the in-process status of the terminal is managed. is there.

Further, when there is a branch transaction to a terminal independently generated by a plurality of CPUs, a plurality of CPs are executed at the terminal.
Contention control such as waiting in the relay processor if there is another branch transaction so that the terminal processing and online processing in the online processing system consisting of a CPU processing device having a single CPU can be used without being aware of U Is what you do.

Further, when the TCP / IP protocol is used to communicate with the CPU processing unit through the relay processor, the business processing unit can be recognized, so that a TCP frame is created for each business processing unit and the load is distributed to a plurality of CPUs. It is also possible.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing an outline of transaction distribution in an online processing system to which the present invention is applied. This application example is an online processing system that processes a transaction from a terminal connected to a communication line 4 via a CPU processing device 1 having CPUs 11 to 13 and a relay processor 2. When the business process G1 including the transactions T11 to T13 and the business process G2 including the transactions T14 to T16 are transmitted from the terminal 3, the relay processor receives the transaction T11.
→ T12 → T14 → T15 → T13 → T16 may be entered in order. According to the present invention, the relay processor 2 analyzes the transaction and recognizes the transmission source terminal and the business process, so that the CPU to which the process is distributed is determined in the unit of the business process G1 or G2 instead of the transaction unit. Transaction T1
1 to T13 are distributed to the CPU 11, transactions T14 to T16 of the business processing G2 are distributed to the CPU 13, and the load is distributed in business processing units.

FIG. 2 shows an example of a transaction format based on the communication protocol used in the conventional online processing system. According to the present invention, in the relay processor, the transaction format 5 transmitted from the terminal is analyzed, the transmission source terminal is identified from the terminal identification information (D2), the transaction processing is identified from the transaction type (D3), and the inquiry is transmitted by the transaction type (D1). Transaction type such as, response, branch, etc. and start / intermediate /
The final is judged, and the order of transactions is recognized by the transaction serial number (D4).

FIG. 3 is a flow chart of transaction distribution in the relay processor according to the present invention. The relay processor analyzes the transaction having the transaction format as shown in FIG. 2 received from the terminal and distributes the transaction to the CPU according to the flowchart.

FIG. 4 is an example of a typical communication sequence for transmitting an inquiry response transaction from a terminal. From the terminal 3, the inquiry transaction is sent individually or divided and sent in order of the head T21, the intermediate T22, and the last T23. C that received the inquiry transaction
The PU processor 1 separates the response or divides it into the first T
24, intermediate T25, and final T26 are transmitted in this order. Upon receipt of the response, the terminal 3 determines that the inquiry response process has been completed as a completion confirmation transaction T27, and the CPU processing device 1
Send to.

FIG. 5 shows an example of the conventional transaction distribution and the transaction distribution according to the present invention when transmitting the inquiry response transaction shown in FIG. In the conventional online processing system, as shown in (a), the relay processor 2 has the configuration including the CPU processing device 6 having a single CPU in the related art. Therefore, the relay processor 2 processes the transactions T21 to T23 from the terminal as they are. I was passing to.

When the online processing system of (a) is applied to the CPU processing device 1 having a plurality of CPUs as shown in (b), transactions T21 to T23 from the terminal.
When the load is distributed in transaction units by the relay processor, the transaction T21, CPU1
In some cases, the transaction T22 may be assigned to the transaction T2 and the transaction T23 may be assigned to the CPU 13, and the transaction order must be controlled using the shared memory 7.

As in (c), according to the present invention, transaction allocation is performed in consideration of business processing, so that it is not necessary to control transaction order.

FIG. 6 is an example of a table structure in the relay processor when the load of the inquiry response transaction processing from the terminal shown in FIG. 4 is distributed in business processing units.
The relay processor 2 has a CPU processing device side management table 8 and a distribution management table 9 which are created for each business process from the terminal.

The inquiry response transaction process from the terminal will be described with reference to the flowchart of FIG. 3 and the sequence diagram of FIG. When the relay processor receives the inquiry leading transaction T21 (101), since it is the first transaction in the inquiry response sequence (102), the distribution destination in the distribution management table 9 is referred to, and the distribution destination is not determined. All the load information in the processing device side management table 8 is referred to, the CPU 13 with the lightest load is selected, (104) Inquiry is written in the in-process status, and the CPU 13 and Inquiry is written in the hanging state (106) and transmitted to the selected CPU 13 (107). Since the transmitted transaction is not the final sequence (108), it waits for the next receiving process. next,
According to the inquiry response transaction sequence, when the inquiry intermediate transaction T22 is received (101),
Since it is not the first transaction in the inquiry response sequence (102), the distribution destination of the distribution management table 9 is referred to (103), the CPU 13 having the written CPU name is selected (105), and the selected CPU 13 is transmitted. (107). Since the transmitted transaction is not the final sequence (108), it waits for the next reception process. Even when the inquiry final transaction T23 is received, the same processing as the reception processing of the inquiry intermediate transaction is performed. Inquiry response head transaction T2
4. When the intermediate transaction T25 and the final transaction T26 are received from the CPU 13 of the CPU processing device 1, the CPU1 of the CPU processing device side management table 8
Refer to the work-in-progress status of 3 and query processing is in progress.
Send each response transaction to the terminal. When the completion confirmation transaction T27 is received from the terminal side, it is transmitted to the CPU processing unit side according to the flowchart of FIG. 3 and is the last transaction in the inquiry response sequence (108). Therefore, the distribution destination of the distribution management table 9 and the CPU The in-process state of the CPU 13 in the processing device side management table 8 is cleared to the initial state (109). In this way, the transactions are sequentially sent to the same CPU for each business process, so there is no need to wait for the completion of the transaction being processed by another CPU between the CPUs.
Sequence control using a shared memory between PUs is unnecessary.

FIG. 7 is an example of a typical communication sequence when a branch transaction is transmitted from the CPU. The CPU processing device 1 side transmits a blocking request transaction T31 for suppressing transmission from the terminal 3. When accepting the blocking request, the terminal 3 transmits a blocking response transaction T32. CP that received the block response
The U processor singly or divides the branch transaction and sends the branch T33, the intermediate T34, and the final T35 in this order. The terminal 3 which has received the response singly or divides the branch response transaction into a head T36, an intermediate T37,
Transmit in the order of the final T38.

FIG. 8 shows an example of a conventional process and a process according to the present invention when transmitting a branch transaction from the CPU processing device shown in FIG. As shown in (a), a CPU having a single CPU in the conventional online processing system
Because of the configuration of the processing device 6, the transaction T31
However, the relay processor 2 passes the transaction T31 as it is to the terminal.

In the case where the online processing system of (a) is configured of the CPU processing device 1 having a plurality of CPUs as shown in (b), a plurality of transactions T31 may occur simultaneously, but the relay processor 2 However, since the contention control is not performed, it is necessary to deal with the case where the transaction 31 is simultaneously transmitted in the terminal 3.

As shown in (c), in the present invention, when a plurality of transactions T31 occur simultaneously, the relay processor 2 holds the second transaction T31 and controls the competition, so that the terminal 3 performs the conventional processing. Can be used as is.

FIG. 9 is an example of a table structure of the relay processor in the case of branch transaction processing from the CPU processing device shown in FIG. The relay processor 2 has a CPU processing device side management table 8 and a distribution management table 9 which are created for each business process from the terminal.

Branch transaction processing from the CPU processing unit will be described with reference to the sequence diagram of FIG. CP
When the blocking request transaction T31 which is the first sequence of the branch transaction sequence is received from the CPU 13 of the U processing device 1, the in-progress state of the tables of all the CPUs in the CPU processing device side management table 8 is referred to and the other branching state Therefore, the branching state is written in the work-in-progress state of the CPU 13 table of the CPU processing device side management table 8, and the CPU 13 as the CPU name and the branching state in the work-in-progress state are written in the distribution destination of the distribution management table 9 to the terminal To send. Furthermore, C
When the blocking request transaction T31, which is the first sequence of the branch transaction sequence, is also received from the CPU 12 of the PU processing device 1, there is another branching state, so the CPU 12 of the CPU processing device side management table 8
Write the branch transmission wait in the work in progress table of
The blocking request transaction T31 is suspended in the relay processor. When the branch response final transaction T38, which is the end of the branch transaction sequence, is received from the terminal, the distribution destination of the distribution management table 9 and the CPU
The in-process status of the CPU 13 table of the processing device side management table 8 is cleared to the initial state, the branch response final transaction T38 is transmitted to the CPU 13, and the in-process status of the CPU processing device side management table 8 other than the cleared CPU is executed. , There is a branch transmission wait in the CPU 12, so the CPU 1 of the CPU processing device side management table 8
2 is written in the work-in-progress state of the table 2, the CPU 12 which has been waiting for branch transmission is written in the distribution destination of the distribution management table 9, and the branch is written in the work-in-progress state.
By transmitting the blocking request transaction T31 held in the relay processor to the terminal, a plurality of branch transactions for the same terminal that occur simultaneously in a plurality of CPUs are competitively controlled. In this way, the terminal side is guaranteed that the branch transaction is sent from only one CPU, so that the terminal processing used in the conventional online processing system including the CPU processing device having the single CPU is executed. And online processing can be transferred as is.

[0029]

According to the present invention, processing is performed by a plurality of CPUs.
For terminals in multiple branch transactions
Competitive control is possible.

[0030]

[0031]

[0032]

[0033]

[Brief description of drawings]

FIG. 1 is a diagram showing an outline of transaction distribution in an online processing system according to the present invention.

FIG. 2 is an example of a transaction format based on a communication protocol.

FIG. 3 is a flow chart of transaction distribution in the relay processor according to the present invention.

FIG. 4 is a typical communication sequence diagram when transmitting an inquiry response transaction from a terminal.

FIG. 5 shows an example of conventional transaction allocation and a transaction allocation according to the present invention when transmitting a query response transaction from a terminal.

FIG. 6 is a table structure diagram in the relay processor when the load of the inquiry response transaction processing from the terminal according to the present invention is distributed in business processing units.

FIG. 7 is a typical communication sequence diagram when transmitting a branch transaction from a CPU processing device.

FIG. 8 is an example of a conventional process and a process according to the present invention when a branch transaction is transmitted from a CPU processing device.

FIG. 9 is a table structure diagram in the relay processor in the case of branch transaction processing from the CPU processing device according to the present invention.

[Explanation of symbols]

1 ... CPU processing device having a plurality of CPUs 2 ... Relay processor 3 ... Terminal 4 ... communication line 5 ... Transaction format 6 ... CPU processing device having a single CPU 7: Shared memory 8 ... CPU processing device side management table 9: Sorting management table 11-13 ... CPU

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Masahiro Tsunekawa, Inventor Masahiro Tsunekawa 5030 Totsuka-cho, Totsuka-ku, Yokohama, Kanagawa Software Development Division, Hitachi, Ltd. (72) Masahiko Tsuda 5030 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Hitachi, Ltd. Software Development Division (56) Reference JP-A-7-249003 (JP, A) JP-A-6-266597 (JP, A) Patent 2555920 (JP, B1) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 15/16-15/177 G06F 15/00 G06F 12/00 G06F 9/46

Claims (4)

(57) [Claims] 1. A transaction control method in a relay device for relaying a CPU processing device having a plurality of CPUs and a terminal, wherein the relay device is responsive to receipt of a blocking request transaction indicating the beginning of a branch transaction from the CPU.
If the in-process state corresponding to the other CPU is the branching state, the blocking request is suspended, and if not in the branching state , the in-process state corresponding to the CPU that has transmitted the blocking request is set to branching, The relay device receives a branch response final transaction indicating the end of the branch transaction from the terminal.
In response, the transaction control method, characterized in that the work-in-process state corresponds to a CPU that sent the closure request and the initial state. 2. The CP that has transmitted the suspended block request when the in-process state is changed from the branching state to the initial state .
2. The transaction control method according to claim 1, wherein the in-process state of U is branched. 3. The C which has transmitted the pending request for blocking.
The transaction control method according to claim 2, wherein when the in-progress state of the PU is in the branching state, the suspended blocking request is transmitted to the terminal. 4. A relay device for relaying between a CPU processing device having a plurality of CPUs and a terminal, and a work-in-progress state corresponding to another CPU in response to reception of a blocking request transaction indicating the beginning of a branch transaction from the CPU. Is in the branching state, the blocking request is suspended, and if not in the branching state , the in-progress state corresponding to the CPU that has transmitted the blocking request is set to branching, and the terminal ends the branch transaction. the response to receiving the branch response last transaction shown, the relay apparatus, wherein a work-in-process state corresponds to a CPU that sent the closure request and means for the initial state.