JP2555227B2 - Message processing based data processing 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 message communication-based system that realizes asynchronous message transmission in a system that realizes mutual requests between global services by message communication and uses a non-volatile memory to prevent message loss. The present invention relates to a data processing device.

[0002]

2. Description of the Related Art FIG. 8 is an explanatory diagram of a conventional technique. A computer system using a message communication means is used as a communication method between programs operating in different address spaces. FIG. 8A shows a message communication environment in such a message communication-based data processing device. In FIG. 8A, 80 is an address space in which the caller program operates, 81 is an address space in which the callee program operates, 11,
21 is a port existing in each address space, 82 to 84
Represents a thread that is the execution unit of the program.

The ports 11 and 21 are windows for programs to exchange messages with other spaces.
The threads 82 to 84 are execution units of a program to which the execution right of the CPU is given, and the program execution environment such as hardware registers and stacks can be switched in that unit.

In the message communication-based data processing device as shown in FIG. 8A, for example, when a thread 82 operating in the space 80 transmits a message M2, a call destination program operates in the address space 81. A thread 83 that receives at port 21 and belongs to that port,
The called program is run on any free thread of 84.

A service program that runs in a unique space within the system, confines system resources in the space, and controls the resources of the entire system is called a global service (G
S). The GS is, for example, a service program that performs exclusive control of files or a service program that controls printing to a printer. Here, the active GS that actually performs the service during operation and the standby GS that is switched to the active when a failure occurs are regarded as one GS. When the active GS goes down, the standby GS takes over the processing.

When message communication is used as a communication means between a program operating in a GS and a program operating in another GS, the calling GS is called again for the processing in the called GS. If such a process is performed, both GSs may be in a message waiting state and both GSs may stop the service function as GSs.

FIG. 8B shows an example of the deadlock state. That is, the calling GS1 space 9
0 receives the message M1 and issues a request by the message M2 to the called GS2 space 91 in the processing. In the GS2 space 91, the thread 93 further issues a request by the message M3 to the GS1 space 90 for processing the message M2.

In the GS1 space 90, the thread 92 is stopped because it waits for a reply message of the message M2 to the GS2 space 91, and if there is no empty thread, the receiving process of the message M3 is kept waiting. As a result, both GS
Both 1 and GS2 become immobile and deadlock occurs.

[0009]

In order to avoid the deadlock described above, it is necessary to eliminate the threads waiting for the mutual message reception by, for example, not waiting for the reply message. . However, if the processes are not synchronized, for example, when the active system GS goes down due to a failure and the service is switched to the standby system GS, the processing request by the message becomes halfway, and the message may be missed. There is a new problem that may occur.

The present invention solves the above problems and provides a means for avoiding a deadlock due to waiting for a message to be received and for preventing a message from being missed even when the caller GS is down. It is an object.

[0011]

FIG. 1 is a block diagram showing the principle of the present invention. In FIG. 1, reference numeral 1 is a computer including a CPU and main memory, and 10 is a GS in which an active GS operates.
1 working space, 10 'is a GS1 standby space in which the standby GS operates, 11 is a port serving as a window for messages to be received,
Reference numeral 12 is a thread belonging to a port, 13 is a transmission-dedicated thread dedicated to processing message transmission, 14 is a non-volatile memory writing section that operates when a message is transmitted, 15 is a non-volatile memory reading section that operates when switching to a standby system, 1
6 is a non-volatile memory in which stored information is not lost even if GS goes down, 17 is message transmission information, 20 is a GS2 working space in which the working GS operates, 2
0'is a GS2 standby space in which the standby GS operates, 21 is G
A port of the S2 active space, 22 represents a thread belonging to the port.

The GS1 active space 10 and the GS2 active space 20 are spaces in which the service programs operate, and perform processing requests and notification of processing results by message communication, respectively.

In the present invention, GS1 working space 10 to GS
2 When sending a message to the active space 20 to request processing, a thread 12 belonging to the port in the calling GS1 active space 10 and a send-only thread 13 for sending the message are separately provided, and the thread 12 belonging to the port is provided. Request from is done asynchronously.

The transmission-only thread 13 comprises a non-volatile memory writing section 14 and a non-volatile memory reading section 15. When a message transmission request is issued from the thread 12 belonging to the port, the transmission-only thread 13 writes the message transmission information 17 to the non-volatile memory 16 and stores it by the non-volatile memory writing unit 14, and stores the message in
2 Send to the working space 20.

The non-volatile memory reading unit 15 does not operate in a normal state, but operates in the transmission-only thread 13 in the GS1 standby space 10 'when a failure occurs. GS1 working space 10 to GS1 waiting space 1
Non-volatile memory reading unit 1 when switched to 0 '
5 reads the message transmission information 17 stored in the non-volatile memory 16 and reads the GS2 working space 20 or G
Send to S2 standby space 20 '.

[0016]

In the present invention, the calling GS1 working space 10
A request from the callee to the GS2 working space 20 of the callee is treated as an asynchronous process. Therefore, the calling GS1
In the active space 10, a dedicated transmission-dedicated thread 13 for transmitting a message to the call destination is created, and the program operating in the thread 12 belonging to the port executes processing without waiting for a reply to the message to the call destination GS. Let it complete. That is, the processing of the thread 12 belonging to the port of the call source GS is completed regardless of the completion of the request to the call destination GS.

As a result, a deadlock due to waiting for a message to be received between the calling GS and the called GS is avoided.

With the above processing alone, when the caller GS goes down, the message transmission processing that was being processed by the transmission dedicated thread 13 may be omitted. In order to deal with this, the transmission-only thread 13 saves the content of the message in the non-volatile memory 16 before sending the message, and if the calling source GS goes down, the non-volatile memory reading unit in the standby GS. The message transmission information 17 is read by operating 15 and message transmission processing to the callee GS is performed again based on this information. As a result, it is possible to avoid the message transmission failure to the call destination GS.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A system according to an embodiment of the present invention is a hardware configuration having a general-purpose computer with a single processor or multiprocessor equipped with an ordinary main memory and a non-volatile memory whose contents are guaranteed even when down. It has become.

The software necessary for implementing the present invention is composed of a program group as shown in FIG. 2, for example. The message communication control unit 30 is a program that controls message communication that operates in the supervisor. The non-volatile memory control unit 31 is a supervisor program that provides a writing function and a reading function for the non-volatile memory 16 shown in FIG. Since the processing functions of these supervisors are well known, detailed description thereof is omitted here.

GS1 working space 10 and GS1 standby space 1
0 ', GS2 working space 20 and GS2 standby space 20'
Then, the same program works respectively. Asynchronous message transmission unit 32, 32 ', non-volatile memory writing unit 1
4, the non-volatile memory reading unit 15 is a program that operates on the transmission-only thread 13 of FIG. 1, and the message receiving units 33 and 33 ′ are programs that operate on the thread 22 belonging to the port.

The asynchronous message transmitter 32 transmits a message to the GS2 working space 20 via the message communication controller 30. In the GS2 active space 20, the message receiving unit 33 receives a message from the port serving as a message receiving window.

Before sending the message, the asynchronous message sending unit 32 writes the contents of the message to the non-volatile memory via the non-volatile memory control unit 31 by the non-volatile memory writing unit 14. When the GS1 active space 10 is taken over to the GS1 standby space 10 'due to the down, the nonvolatile memory reading unit 15 reads the content of the message in the transmission state by the nonvolatile memory control unit 31 and transmits the asynchronous message. It is retransmitted by the unit 32 '.

FIG. 3 is a diagram showing a state in which the message of the called GS is desynchronized according to an embodiment of the present invention, FIG.
3 is a flowchart of steady processing in the state shown in FIG. The message communication processing between GS1 and GS2 will be described below with reference to FIGS. ~ In the text and ~ in FIGS. 3 and 4 correspond to each other.

The GS1 working space 10 receives a request for a message M1 from a user. This message M1 activates the thread 12 belonging to the port.
When a processing request for the GS2 is required from the thread 12 belonging to this port, the transmission-only thread 13 is activated to send a message to the GS2 working space 20.

The transmission-dedicated thread 13 shows the contents of the message M2 necessary for requesting the GS2 again when the GS1 goes down during the message transmission from the GS1 to the GS2. It is stored in the non-volatile memory 16 via 31.

After that, the asynchronous message transmission unit 32 shown in FIG.
The request for 2 is sent from GS1 as message M2.
In the GS2 working space 20, the thread 22 belonging to the port is activated, and in this process, the request for the GS1 is transmitted as the message M3. In the GS1 active space 10, it is considered that the message transmission to the GS2 has been completed by the processing, and the processing of the thread 12 belonging to the port is completed. Therefore, it is possible to wait for the thread 12 belonging to the port to be activated by the message M3. Absent.

FIG. 5 is an explanatory diagram for avoiding message transmission omission according to one embodiment of the present invention, and FIG. 6 is a flowchart of processing by the standby space of the GS1. Hereinafter, the message communication processing when the GS1 is down due to a failure will be described with reference to'- 'shown in FIGS.

In FIG. 5, the processes from to are the same as the processes described in FIG. By processing, when the message transmission information 17 is written in the non-volatile memory 16,
Assume that the GS1 working space 10 is down. GS
The processing from "to" is performed by one standby space 10 '.

The'GS1 standby space 10 'is notified of the resend of the message M1 requested by the user. '
Upon receipt of the message M1, the thread 12 ′ belonging to the port is activated, and the thread 12 ′ belonging to the port
Activates a send-only thread 13 'for sending a message to GS2.

The'transmission-dedicated thread 13 'reads the content of the message M2 to be transmitted, which is left in the non-volatile memory 16 by the GS1 active space 10. 'After that, the asynchronous message transmission unit 32' shown in FIG. 2 operating on the transmission dedicated thread 13 'transmits the message M2 having the content restored in the processing' to the GS2. The process is the same as the steady process shown in FIGS.

FIG. 7 is a more detailed processing flowchart of the transmission dedicated thread 13. Below, shown in Figure 7 (a)
It will be explained according to (i). (a) At the time of IPL, the necessary buffers and other work areas are initialized. (b) The WAIT processing waits for a processing request from the port thread. (c) If the end of processing is notified, the necessary post-processing is performed and the processing ends.

(D) If it is not the end notification, it is judged whether it is the first processing request after the active down in the standby space. If so, the process moves to the process (h) to retransmit the message that has not been processed. (e) In the case of a normal message transmission request from the thread 12 belonging to the port, the content of the message is written to the non-volatile memory 16.

(F) Send a processing request message to GS2. (g) After that, if it is confirmed that the GS 2 has completed processing or received the message, the information stored in the non-volatile memory 16 is erased. Then, the processing returns to the processing (b) and waits for the next processing request.

(H) If the processing request to the transmission dedicated thread 13 is the first request when the standby space is switched to the working space, the content of the message stored in the non-volatile memory 16 is read. . (i) If there is no message stored in the non-volatile memory 16, the process returns to the process (b) and waits for a process request. If there is a message, move to process (f) and send the message.

[0036]

As described above, according to the present invention,
By making the request between the calling global service and the calling global service asynchronous, it is possible to avoid deadlock due to waiting for message reception during that time. Even if the caller's global service goes down, the message contents saved in the non-volatile memory are used by the standby global service to send a message to the callee's global service and send the message. Leakage can be avoided.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a diagram showing a software configuration example according to an embodiment of the present invention.

FIG. 3 is a diagram showing a state in which a message of a callee GS is desynchronized according to an embodiment of the present invention.

FIG. 4 is a flowchart of steady processing according to an embodiment of the present invention.

FIG. 5 is an explanatory diagram of avoiding message transmission failure according to an embodiment of the present invention.

FIG. 6 is a flowchart of processing by the standby space of the GS1 according to an embodiment of the present invention.

FIG. 7 is a processing flowchart of a transmission-only thread according to an embodiment of the present invention.

FIG. 8 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

 1 Computer 10 GS1 Working Space 10 'GS1 Standby Space 11 Ports 12 Threads belonging to 12 Ports 13 Threads for Sending 14 Nonvolatile Memory Writing Section 15 Nonvolatile Memory Reading Section 16 Nonvolatile Memory 17 Message Transmission Information 20 GS2 Working Space 20' GS2 Standby space 21 ports Threads belonging to 22 ports

Claims (1)

(57) [Claims] 1. In a data processing device that realizes a mutual request between service programs controlling resources of the entire system by message communication, a service program operating space is provided with a window for exchanging messages with other spaces. A thread (12) belonging to a port and a send-only thread (13) for sending a message to another space at the request of the thread belonging to the port A means (14) for writing to the non-volatile memory (16) to take over to the standby space at the time of a failure is provided, and when the space for service is switched to the standby space due to the failure, the transmission written to the non-volatile memory A message characterized by being configured to read and resend information about the message Communication based data processing apparatus.