JPS60144863A - Rough-coupled multi-processor system - Google Patents

Abstract

PURPOSE:To restart and continue the processing which is disconnected owing to the breakdown of a certain processor system with the back-up of another processor system, by providing two control information memories exclusive for storage of control information. CONSTITUTION:An input telegram is received by a line controller 301 and stored to a main memory 201 via a channel device 111. A processor 101 checks the telegram and reads the desired file data out of a disk device 310 to produce an updated data from the telegram data and the file data. This updated data is first written to a control information memory 211 in the form of the journal information then to a control information memory 212. Therefore a processor 102 checks the contents of memories 211 and 212 in case the processor 101 is brokendown. Thus the processing can be restarted and continued by making use of the contents of an area corresponding to the processor 101.

Description

[Detailed description of the invention] [Technical field to which the invention pertains] The present invention relates to information processing systems, and particularly to information processing systems.

The present invention relates to a loosely coupled multiprocessor system with high reliability.

[Prior art] Conventionally, this type of high reliability system has one example.

"Written by Tsuyoshi Umeuri: "No'n7Stop System
The principle of "Busy Scomi x 2'f V Yon'81.
Vol-18°No2. pp, 132-138J
As seen in , a system in which multiple processors are equipped with multiple operating systems (hereinafter referred to as OS) and these are coupled together, a so-called loosely coupled multiprocessor system, is configured, and phase 1-r is configured. The system is operated while backing up control information. In this system, even if one processor system goes down,
Other processors that were backing up this system were able to take over the job and continue operating the system, achieving a high level of reliability.

However, in the case of this system, each of the plurality of processors independently owns the main memory, so the memory address spaces are each independent and different. Therefore.

When control information is banked up between multiple processor systems, the processor on the backup side should intervene and back it up: Is it necessary to receive the lr control information and then save it correctly in its own address space? . However, each processor has to perform extra processing to back up the li'l fill information, resulting in a drop in performance.

In addition, as a solution to the above-mentioned shortcomings, a common method (iii) is proposed to solve the above-mentioned drawbacks. Memory (- means each block [1 setsa] corresponds to the rhyme 11
When Aoyuki is injected into the system and one of the processors goes into system town, the other processors read the control information from the shared memory and, based on the read control information, perform the actions of the downed processor. A processing continuation method has been proposed that takes over the processing that was previously in progress. However, in this method, there is only one shared memory, and it is dedicated to backup, so if the processor system goes down while the control information is being written to the processor or shared memory, the contents of the shared memory will be disturbed. Otherwise, other processors will no longer be able to use the control information. Also, although the correct control information is stored in the main memory of the downed processor, Iya's Brocenose cannot access the above device, so Processing of the towned Flosser cannot be continued.

[Purpose of the invention]

The object of the present invention is to solve the above-mentioned conventional drawbacks.

The object of the present invention is to provide a highly reliable loosely coupled multiprocessor system.

[Structure of the invention]

According to the present invention, in a loosely coupled multiprocessor system in which a plurality of processors each have a main memory, control information for two processors other than the main memory is recorded in the plurality of processors. After updating the control information of one of the two control information storage devices to the latest state, each processor updates the control information of one of the two control information storage devices to the latest state. the other '+IJ
, Control Shield Report 1.0 Device Control gl! I information is also updated with the same content, and if a certain processor system goes down due to a failure and the processing is postponed, E L < is stored in one of the two control information recorders/1G devices. System 3 [11] Based on the information, other processor systems execute the suspended process i! +: Lith it ($1'L) is changed to the cow 11' and -j-ruj'tI combined multi-brosenoreno system is deleted.

[Example of this invention]

Next, regarding the present invention, 1. Explain in detail.

Referring to FIG. 1, an embodiment of a loosely coupled multiprocessor system according to the present invention is shown. In this embodiment, two processors 101, 102, . Main storage device 201, 202° control information storage device 211, 212.
Channel devices 411, 112. Line control device 301, 3
02° It is composed of disk control devices 311, 312, tape control devices 621, 522, and one disk device 610 and one tape device 6320 each.

Referring to FIG. 2, there is shown a processing flow of complex text processing to provide a detailed explanation of the present invention.

The operation of the present invention will be explained below with reference to FIGS. 1 and 2.

The human text received via line 401 is as follows.

Time bJ: It' is received by the control device 301 (e), and the 4 line control information is divided i'jl[, and only the necessary message data is stored in the main storage device 201 via the channel device 111 (the second [Fig. 10 ).The processor 101 checks this telegram data (FIG. 2, 11), determines the corresponding file from its contents, and instructs the Chao Lu device 111 to read the file in order not to read it. Channel device 11
1 reads necessary file data from the disk device 310 via the disk control device 311 according to the instructions, and stores it in the main storage device 201 (FIG. 2, 12). Based on this, update data for updating file data is created (FIG. 2, 13).

Next, in preparation for the unlikely event of file destruction, the processor 101 first stores the updated data in one control information storage device 211 as journal information (this is called journal ring 8).

Next, the same information is written into the other control information storage device 212 to match the contents of both the control information storage devices 211 and 212 (FIG. 2, 14). next.

The processor 101 instructs the channel device 111 to journal, and the channel device 111 writes the updated data as journal information to the tape device 620 via the tape control device 321 according to the instruction (this is called journaling B). Figure 2 15).

Furthermore, processor 101 instructs channel device 111 to write a file in order to re-input the updated data into the file. In accordance with this instruction, the channel device 111 sends the update data to the disk device 5 via the disk controller 511.
10 (FIG. 2, 16).

Next, in response to completing the processing of the message, the processor 101 sends one control information record ↑, 117. Device 2
Update the contents of 11, y, and delete the journal information, and 1, then the other control information record? The contents of the No. 5 device 212 are updated and the journal information is similarly deleted, and the contents of both control information stored in the control information storage devices 211 and 212 are made to match (second [D17). Here, the control information is 9 normal.

It includes resource management information, task status information, address correspondence information, etc. The processor 101 is.

When the update of the control information is completed, an output message is created next (Fig. 2 18), and this created output message is sent to the line control device 3.
01 to line 401 (FIG. 2, 19).

The above description of the operation describes the operation when the system is operating normally.

The following describes the operation when a failure occurs in the processor 101 or the main storage device 201 during execution of the above processing and the processor 101 system goes down.

When the processor 101 system goes down, the processor 102 is notified of this. When the processor 102 receives this notification, it first checks the contents of the control information storage device 211, and if the contents of the control information storage device 211 are disturbed, it writes the control information in the primary device. 2
Check the contents of 12. Here, as is clear from the above explanation of the operation, the processor 101 writes the updated data to the control information storage devices gQ211 and 212 simultaneously (without interfering with each other, but serially for 9 hours). , even if the processor 101 goes down during the update of the control information, the control information will be updated to the 0 devices 211, 212.
is stored normally in at least one of the files. Thus, processor 102 examines the two corresponding areas of control information storage devices 211 and 212.

The state in which the message was interrupted is determined by using the correctly saved contents.

If the journaling A14 has not yet been completed, the message must be processed 341B from the beginning, so a retransmission of the message is requested. Furthermore, if the control information update 17 has already been completed, the process continues and output message creation 18 and output message transmission 19 are performed.

The processor 102 looks at the control information and executes journal link Δ1.
4 has been completed, but when it is determined that the control information update 17 has not yet been completed, the contents of the tape device 320 and disk device 310 are further read and compared with the journal information stored in layering A14. It is determined whether journaling B15 or file update 16 is completed or not.

In this way, the processor 102 can clearly grasp the interrupted state of the message processing, so that it is possible to resume and continue unfinished processing based on the journal information stored in journaling Δ14. can.

In the above description, it is assumed that the processor 101 system is down, but it is clear that even if the processor 102 system is down, the processor 101 can resume and continue processing in exactly the same way.

Also, did I explain that there are two processor systems: the Florenoza 101 system and the Processor 102 system?

It is clear that even if there are more than one system, the processing can be continued in the same way if the two control information registers and three devices are respectively shared by the processors of all systems.

〔Effect of the invention〕

As explained below, in the present invention, by providing two 111 control information recording devices for storing control information, processing interrupted in a down system can be resumed in another system. The effect is that it can be continued.

Margin below

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of a loosely coupled multiprocessor system according to the present invention. FIG. 2 is a flowchart showing the operation of the embodiment of the present invention. 101, 102... processor, 111, 112...
Channel device, 201, 202 ・Main description, ・U device, 211.212 ・Control information storage device, 301.3
02 Line control device, 310/disk device, 311
, 312... Disk control device, 320... Tape device, 321, 322... Tape control device, 401, 40
2-Line. Agent (7127) bu1-"Ihik1j2%' (Yosuke -1-27"゛'+go

Claims (1)

[Scope of Claims] 1. In a loosely coupled multiprocessor system having a plurality of processors and a plurality of main storage devices held separately by the plurality of processors, Each includes two control information storage devices that are shared, and if a certain processor system goes down and processing is interrupted, the control information correctly stored in one of the two control information storage devices can be saved. Based on. A loosely coupled multiprocessor system, characterized in that processor systems other than the certain processor system resume and continue the interrupted processing. Remaining days below