JPH01286048A - Duplex computer system - Google Patents

Abstract

PURPOSE:To make unnecessary the copying of memory data at every break point, and to reduce the load of a CPU by constituting the title system so that the live computer system side may refer to the all memory spaces of a broken- down computer system. CONSTITUTION:Computer systems SY1 and SY2 respectively have CPUs 11 and 21 and memories 12 and 22, and share memory spaces 41 and 42 of a single virtual memory system. Memory copying means 31 and 32 copy information on the real memory of the broken-down computer system through a cross memory access means 6 on the real memory of the live computer system. In virtual memory space access means 51 and 52, for the information to exist on the virtual memory, whereas, not to exist on the real memory, the live system accesses the virtual memory space used by the broken-down system based on the information on the real memory of the broken-down system.

Description

[Detailed description of the invention] [1llt Required 1 Regarding a redundant computer system configured by duplicating a system consisting of a CPU, memory, etc., if one computer system goes down without causing a drop in CPU performance, The purpose of this system is to allow one computer to smoothly take over a running zip, and it consists of two computer systems each having a CPU and memory, and each sharing a single virtual memory space. , -1
A cross-memory access means is installed between the two computer systems to mutually access the memory of the other system from the CPU of the other system. a memory copying means for copying stored information to the large memory of a living computer system via the cross-memory access means; and a system in which the living system has gone down based on information on the real memory of the system that has gone down. A virtual memory space access means for accessing information in the virtual memory space that was used by the user and does not exist in the real memory is configured in the installation group.

[Industrial Field of Application] The present invention relates to a redundant computer system configured by duplicating systems consisting of CPUs, memories, etc. More specifically, if one system goes down, the other system takes over and handles the job ( This invention relates to a redundant computer system that enables continuous takeover of processes (processes).

[Prior Art] Conventionally, one method for increasing system reliability has been to duplicate needle bar machines and memories.

FIG. 6 is a block diagram illustrating an example of a conventional duplex needle n machine system of this type. In the figure, SYl, SY
Reference numeral 2 denotes a computer system having substantially the same configuration, and is connected to a plurality of terminals (eg, workstations) TMl, . . . , TMn via a common interface (eg, LAN). Each computer system SY1. In SY2,
11.21 is CPU, 12.22 is memory, 13.23
is a workstation controller that controls multiple workstations via an interface, 14.24 is a disk, and 15.25 is a disk υJIIl device that controls disk 14.24, respectively, and each of them is a disk in the other system. It is configured so that cross-calls can be made to.

Memory within each system 12.22. Workstation controller 13.23. The disk controller 15.25 is connected to the bus BS1. It is interconnected to CPI Jll, 21 via BS2.

Reference numeral 30 denotes a memory transfer means provided between the memories 12 and 22 in each system, so that data can be transferred between the two memories.

One computer system, for example, runs on SYl and I
In the job currently running, data necessary for restoring the job is copied to the memory 22 in the other computer system SY2 via the memory transfer means 30 at each breakpoint that is a breakpoint in the process.

Now, if an abnormality occurs in one computer system SY1 for some reason and the system has to stop, the other computer system SY2 will use the breakpoint information copied to the memory 22. The job that was previously running on the crusher system SY1 is restored, and the computer system SY2 takes over the operation.

[Problems to be Solved by the Invention] In the conventional redundant computer system with such a configuration, the memory transfer means 30 needs to transfer restoration information between memories for each breakpoint, which reduces the processing load on the CPU. There is 1mff1 that increases the performance and reduces the performance. If the restoration information is to be transferred without degrading performance as much as possible, the amount of transferred information can be reduced, but in order to do so, it is necessary to configure the application programmer to intentionally narrow down the information. arise. Furthermore, Ilo control information (workstation screen information, etc.) is not directly related to job operations and must be copied one by one.

The present invention has been made in view of these points, and consists of C.
The purpose of this invention is to provide a redundant computer system that can smoothly take over jobs that were currently being executed to the other 41 computer system when one computer system goes down, without degrading the performance of the PU. .

[Means for Solving the Problems] FIG. 1 is a block diagram of the principle of the present invention. In the figure,
SYl and SY2 are CPU 11゜21 and memory 1, respectively.
2.22, they share a single virtual memory storage space 41.42.

6 are two computer systems SY1.6. This is cross memory access means provided in SY2 for mutually accessing the memory of the other system from one CPU. Provided within each computer system SY1 and SY2, 31.
32 is a memory copying means 51 for copying information on the real memory of a down computer system to the real memory of a living computer system via the cross memory access means 6;
゜52 is a virtual system that uses information that exists in virtual memory but does not exist in real memory to access the virtual memory space used by the down system based on information in the real memory of the down system. It is a storage space access means.

[Operation] When one system goes down, the cross memory access means 6 copies the information in the memory of the down system to the other system, and the information that is in the virtual memory but does not exist in the memory is copied to the other system. The virtual storage space access means of the other system accesses the virtual storage space used by the system that has gone down based on the information in the memory of the system that has gone down.

This allows the other system to take over the virtual space of one system that is down.

[Example] Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 2 is a block diagram of an embodiment of the present invention. Components that are the same as those in FIG. 1 are not shown with the same symbol. Virtual storage spaces 41 and 42 are provided for each computer system SY1. S.Y.
It is provided in the second disk device 14.24. or,
One mail for each system [Covey means 31.32 is C
The virtual storage space access means 51 and 52 are provided in the disk controllers 15 and 25, respectively. Incidentally, the memory copy means 31 and 32 are connected to each CPU 1.
1.21 can be configured by a MOVE command or the like for copying information between the memory 12 and the memory 22.

In addition, although the virtual storage space access means 51 and 52 are described as being provided in the disk controller [15, 25], in reality, the address within the virtual storage space is specified by the page management function of O8. This is a function that accesses the file in which virtual memory information is stored in the disk device 14.24 corresponding to this address.In terms of hardware, the disk device [14.24] can be accessed from the CPU 11.21. It is sufficient if a bus (disk control device) is provided.

FIG. 3 is a conceptual diagram of a virtual storage space 41.42 provided in each disk device 14.24.

These virtual storage spaces 41 and 42 are each divided into four areas. Here, kernel 1 (kernel 2) is an area in which a part of O8 owned by computer system SY1 (computer system 5Y2) is stored. The "A" area is a job space that the computer system SY1 can normally access, and in the normal (normal) operating state, the computer system SY1 performs operations while accessing only the kernel 1 and "A" areas. . The B" flj area is a job space where the computer system SY2 normally accesses, and is projected from the computer system SY1 to the same address as the computer system SY2. Each area jl A I+ and B
” can be accessed from the other system's space in backup mode when one of the computer systems goes down, and is protected in other modes. 'C' and 'D II The areas are projection areas of the opponent's space of Kernel 2 and Kernel 1, respectively, and can be accessed from the opponent's space in backup mode.In other modes, they are protected.

FIG. 4 is a configuration block diagram of an address conversion circuit provided in the CPU 11 (or CPU 21) for converting a logical address into a real address.

71 is an address determining means for inputting a logical address from the CPU and determining which area of the virtual storage space 41 (42) is to be accessed by the logical address; 72 is a means for determining when the computer system SY1 (SY2) is down; When the backup mode is entered, the backup mode registers G1 and G2 that indicate the state input the signal from the backup mode register 72 and the signal from the address determination means 71, and request access to the memory of the other system. Gates 73 and 74 create signals and signals, and 73 and 74 are tables for inputting logical addresses and creating real addresses to the memory of one's own system and the memory of the other system. Although not shown in the memory 12.22, the CPU 1
1 or CPLI 21, and means is provided for determining which access is valid.

FIG. 5 shows two computer systems SY1. 2 is a configuration block diagram of cross memory access means 6 provided between SY2. FIG. Driver/receiver DR inserted into the address bus AB1 connected from CPtJll of one computer system SY1 to the memory 22 of the other computer system SY2
1. Driver/receiver DR2, C inserted into the address bus AB2 connected to the memory 12 from the CPU 21 of the other computer system SY2 to the one computer system SY1.
Data transmission/reception buffer yBF1゜8F2 and drivers/receivers DR3, DR4 inserted into the data bus 881 connected from the PU 11 to the memory 22; data transmission/reception buffer BF3゜8F4 and drivers/receivers inserted into the data bus 8B2 connected from the CPU 21 to the memory 12. OR
5, DR6.

The operation of the device configured as described above will be explained as follows.

Now, there are two 311 aircraft systems SY1. When both SY2 are operating normally, one computer system (for example, 5Y1) is the main system, and the other t7'PXi system (for example, 5Y2) is the slave system, forming a normal mode.

In this normal mode, each computer SY1. SY2 can access only his own virtual space, and access to the other party's virtual space is protected. That is, in FIG. 4, the address determining means 71 determines which area of the virtual storage space 41 is accessed by the logical address from the CPU 11. Based on this determination result, if the logical address is to the kernel 1.0A'' area, an access request to the memory 12 is unconditionally MI EN
Send out. As a result, the memory 12 is accessed (by the CPU 11) given the real address converted by the table 73.

On the other hand, the logical address is 11 B according to the judgment result.
If it is within the area ", C", the handling will differ depending on the contents of the backup mode register 72. That is, in the normal mode, the memory 22 is
The access request signal to CP is suppressed, and the protect signal indicating that it is protected by gate G2 is suppressed.
1 notified to LJll. . ,＜Tsukuup,−
Do, if there is, me.

The access request signal M2EN to the computer system SY2 becomes valid, and the real address converted by the table 74 is given, so that the virtual storage space of the other computer system SY2 can be accessed by the CPU 11 of the one computer system SY1. .

Therefore, in this state, the CPU of the computer system that is not down accesses the memory of the other system through the cross-memory access means 6, and the computer system that is down accesses the memory of the other system, and stores the status of the job that was being executed until then in the memory. It becomes possible to expand (copy) the job onto the memory of the other computer system using the copying means and to continue taking over the job. In other words, the computer system that does not go down in this state, for example, the CPU 21 of SY2, is MOV.
By issuing the E command, it is possible to access the memory 12 of the down computer system SY1, read out the information in the memory 12, and store it in the own memory 22. Here, if the information necessary to restart the job of one of the down computer systems does not exist in the real memory, the other computer system that has taken over will update the information on the real memory of the down computer system. Based on this information, the disk containing the virtual memory previously used by the downed system can be accessed. This allows the other computer system to take over the virtual space of one computer system that has gone down.

In addition, each computer system SY1. In SY2, the disk devices 114.24 that store the virtual space are separately provided for exclusive use, but by realizing the cross-call function in each disk control device 15.25, it can be done without any problem. Can access virtual spaces of other systems.

In this embodiment, it is assumed that in each computer system, each memory 12, 22 is made more reliable by providing, for example, an error correction circuit or alternate memory, so that it is difficult to go down. Therefore, memory 12
．． If 22 goes down, relief will be provided by other means.

[Effects of the Invention] As explained in detail above, the present invention has a configuration in which all memory spaces (virtual spaces) of a down computer system can be referenced from the living computer system side.
According to the present invention, by referencing memory only when one of the computer systems goes down, there is no need to copy memory data at each break point as in conventional systems, and the load on the CPU can be reduced. , performance can be improved. In addition, by referring to the information of the O8 core such as Te3, it is possible to extract the downed job and restart the remaining jobs.Also, by taking over the I10 control information, the system can be controlled under the control of the downed system. I
Ill lit'' with a system that lives as it is.
You can do it.

[Brief explanation of the drawing]

Figure 1 is a block diagram of the principle of the present invention, Figure 2 is a block diagram of the configuration of an embodiment of the present invention, Figure 3 is a conceptual diagram of the configuration of virtual storage space, and Figure 4 is an address provided within the CPU. FIG. 5 is a block diagram of the configuration of the conversion circuit, FIG. 5 is a block diagram of the cross memory access means in FIG. 1, and FIG. 6 is a block diagram of the configuration of an example of a conventional system. 1 and 2, SYl and SY2 are computer systems, 6 is cross memory access means, 11.21 is cpul, 12.22 is memory, 31.32 is memory copy means, 41.42 is virtual space, 51 .52 is a virtual space access means. Patent applicant: Fujitsu Limited
Company Representative Patent Attorney Fuji Ijima
Figure 1 of the Atsuka block copper color of the present invention

Claims (1)

[Claims] Two computer systems (SY1, SY
2), between the two computer systems (SY1, SY2),
A cross-memory access means (6) is provided for mutually accessing the memory of the other system from one CPU, and the cross-memory access means (6) transmits information on the real memory of the down computer system to the two computer systems.
) to the real memory of a living computer system (31, 32), and a memory copy means (31, 32) that copies information on the real memory of the living system to the real memory of the down system. 1. A redundant computer system comprising virtual storage space access means (51, 52) for accessing information in a virtual storage space that does not exist in real storage.