JPH0459650B2 - - Google Patents

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides a method for multiple processors to independently transfer data to an auxiliary storage device such as a common hard disk device via a common disk transfer buffer. Regarding a disk transfer buffer management method in a data processing system capable of In addition, the system is configured to start up without initializing the disk transfer buffer.

[Conventional technology]

Generally, in a data processing system in which a common hard disk device is accessed independently from multiple processors via a common disk transfer buffer, a disk transfer buffer is provided between the main memory of the processor and the hard disk device, and this is also blocked. Tag information is easily stored for each block for easy management, and if there is data to be accessed on the disk transfer buffer, the data on the disk transfer buffer can be transferred without actually transferring the data to the hard disk. A disk transfer buffer management method is known that aims to improve system performance by reducing accesses to hard disk devices through processor access.

For example, when starting up the system,
The disk transfer buffer and tag information are all empty and there is no data on the disk transfer buffer that can be accessed, and even if the system is running, the disk transfer buffer can still be accessed. If there is no data,
Search for empty tag information, read data from the hard disk device into the corresponding disk transfer buffer block, and write the file address on the hard disk in the tag information, then transfer the data in the data read block to the processor. .

Furthermore, when there is no free tag information, a so-called swap-out is performed in which the contents of a block with a disk transfer buffer are written to the hard disk device to create a free block, and then the tag information is made free and data is transferred from the hard disk device to the free block. Load.

By the way, when multiple processors independently transfer disks to a common hard disk device via a common disk transfer buffer, multiple processors may simultaneously access the same data on the disk transfer buffer, or one processor may When the data to be accessed is newly read from the hard disk device to a block with a disk transfer buffer and the tag information is manipulated, that is, the file address is rewritten, etc., if another processor accesses the same data, the access The target tag information does not exist, and the same data is independently read into another block, resulting in multiple pieces of data corresponding to one block on the hard disk device being stored on the disk transfer buffer. usage efficiency decreases.

To prevent this, a buffer management block is provided so that only one processor is granted the right to use the disk transfer buffer, and the processor that makes the disk transfer request occupies the buffer management block and performs the disk transfer. A disk transfer buffer management method has been proposed in which the occupancy information is released after the processor is used, and if the processor cannot be occupied because another processor is in use, the system waits until the processor is occupied.

[Problem that the invention seeks to solve]

However, in the conventional disk transfer buffer management method described above, even if data on the disk transfer buffer is rewritten, the corresponding data on the hard disk device is not updated immediately.
If the system goes down due to something other than a normal system startup, there is a problem that data on the hard disk device will be lost, and in a system where multiple processors are running, one processor may occupy the management block. If the processor goes down, the disk transfer buffer remains occupied by the down processor, and other operating processors are unable to perform disk transfers.

Therefore, the present invention allows other processors to release the disk transfer buffer occupied by the down processor to continue disk transfer.
Moreover, it is an object of the present invention to provide a disk transfer buffer management system that allows continuous use of the disk transfer buffer without data loss even when a downed processor is started up.

[Means for solving problems]

In order to achieve the above object, the present invention provides a system in which when a plurality of processors independently use a common auxiliary storage device via a common disk transfer buffer, one processor uses the common disk transfer buffer. In a disk transfer buffer management method in which a buffer management block is occupied, when a processor occupies a buffer management block, the number of the occupied processor is stored in the buffer management information block, and if the processor goes down, Then, the other normal processors are notified of this, and based on the down processor number, the other normal processors determine whether or not the processor occupying the buffer management block is a down processor, and the determination result is When the processor is a down processor, release the buffer management block and initialize the disk transfer buffer when the contents of the externally inputtable buffer storage register are in the saved state when starting up the processor. I try to start up the processor without any problems.

[Effect]

In this invention, when a processor goes down while occupying a disk transfer buffer, the processor occupying the disk transfer buffer transfers the memory to the down processor based on the processor number stored when another normal processor occupied the disk transfer buffer. If the result of the determination is that the processor is down, the buffer management block is released, allowing another normal processor to continue disk transfer. In addition, by setting the buffer storage register that can be input externally to the saved state when starting up a processor that has gone down, it is possible to maintain the contents of the disk transfer buffer in the state before it went down, allowing one or more processors to Even if the hard disk drive goes down, data loss in the hard disk device can be prevented without deteriorating system performance.

〔Example〕

Embodiments of the present invention will be described below based on the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention.

In the figure, P ₀ , P ₁ to P _N are processors connected to each other via a system bus 1, and 2 is a hard disk controller connected between the system bus 1 and a hard disk device 3 as an auxiliary storage device.
4 is a buffer storage register connected to the system bus 1 and set by an external switch 5; 6 is a common memory connected to the system bus 1;
In this common memory 6, a buffer management block 7, tag information 8, and a disk transfer buffer 9 are set.

As shown in FIG. 2, the buffer management block 7 stores occupancy information 11 indicating whether the buffer management block 7 is occupied or not, and the processor number 12 that occupies it. Further, the tag information 8 includes, as shown in FIG. 3, block validity information 13 indicating whether or not a block in the disk transfer buffer 9 is free, and data on the hard disk device 3 shown in FIG. 5 in M+1 blocks. The file address 14 is composed of the block number when the file is divided into two. Furthermore, as shown in FIG. 4, the disk transfer buffer 9 stores the block B ₀
It is configured to be able to store a number of block data corresponding to the number of processors N+1 in ~B _M.

Thus, for example, when a disk transfer request occurs to processor P ₀ , processor P ₀ transfers memory 6
Read the occupancy information 11 of the buffer management block 7 of
, and also sets the occupied processor number 12 to occupy the disk transfer buffer 9.

Next, the tag information stored in the tag information area 8 is searched to determine whether data to be accessed on the disk exists on the disk transfer buffer 9. At this time, if the block valid information 13 is "1", the block on the corresponding disk transfer buffer is valid, and the file address 14 on the hard disk device 3 corresponding to the block is read and the file address to be accessed is read out. If the block exists, data is transferred to the corresponding block on the disk transfer buffer 9.

When the target file address does not exist, empty tag information 8 whose block valid information 13 is "0" is searched, or when there is no empty tag information 8, the disk transfer buffer corresponding to one piece of tag information is searched. 9 to the hard disk device 3 indicated by the file address 14.
Swap out to the upper block, empty the tag information 8 (set the block valid information to "0"), import the data to be accessed from the hard disk device 3 to the disk transfer buffer 9, and read the tag information 8. Enter the file address 14 in the field, and set the block valid information 13 to “1”.
After that, transfer the data. When this data transfer is completed, the processor _P0 transfers the occupancy information 11 of the buffer management block 7 and the occupancy processor number 1.
2 and release the right to use the disk transfer buffer 9.

The above describes the processing when processor P ₀ completes data transfer in a normal state. However, if processor P ₀ goes down while occupying the buffer management block 7, the system will notify that processor P ₀ has gone down. The other processors P ₁ -P _N are notified by a processor down signal on bus 1.

Therefore, the processor that receives the down notification will
The process shown in FIG. 6 is executed. That is, in step it is determined whether or not the management block 7 is occupied, and if it is, the flow advances to step and it is determined whether the occupied processor number and the down processor number match. At this time, if the occupied processor number and the down processor number match, the process moves to step where the occupied information 11 and the occupied processor number 12 are deleted, and then the process ends.

On the other hand, if the determination result of the step is that the management block is not occupied, or if the determination result of the step is that the occupied processor number and the down processor number do not match, the processing is directly terminated.

As a result, the state of occupation of the disk transfer buffer by the down processor _P0 is released, allowing other normal processors to occupy the disk transfer buffer, allowing the normal processor to continue performing disk transfers.

Also, a certain processor P _i (i=0, 1, 2...
N), the processing shown in FIG. 7 is executed in its processor P _i . That is, in this step, it is determined whether another processor is already in operation. At this time, if another processor is in operation, the process is immediately terminated.
This allows a downed processor to be incorporated into a running system.

On the other hand, if there are no other processors running,
Proceed to step 7, buffer management block 7.
The buffer management block 7 is forcibly occupied by setting the occupancy information 11 to "1" and writing in the occupancy processor number 12. As a result, even if the entire system goes down while a certain processor is occupying the buffer management block 7, or in other words, even if the buffer management block 7 cannot be released from its occupation, it is possible to continue using the disk transfer buffer 9, and the next step to move to. In this step, it is determined whether the buffer save register 4 is set to "1" to save the contents of the disk transfer buffer. At this time, buffer storage register 4 is “0”
If it has been reset, the process moves to step, where the disk transfer buffer and tag information are initialized to become empty, and then the process moves to step.

In this step, the occupancy information and processor number are deleted before the process ends.

On the other hand, if the judgment result of the step is that the buffer storage register 4 is set to "1", it is determined that the disk transfer buffer 9 is to be used continuously, and the process moves directly to the step without proceeding to the step. After deleting the exclusive information 11 and the exclusive processor number 12, the process ends.

Here, by setting the contents of the buffer save register 4 to "1" or "0" by the external switch 5, it is possible to select whether or not to save the data on the disk transfer buffer 9 when the processor is restarted. By setting the buffer storage register 4 to "1", it is possible to save the data on the disk transfer buffer 9 and prevent data loss.

In the above embodiment, a case has been described in which the hard disk device 3 is used as the auxiliary storage device, but it goes without saying that a storage device such as a magnetic disk device or an optical disk device can be used instead.

〔Effect of the invention〕

As explained above, according to the present invention, by setting the occupied processor number together with the occupation information in the buffer management block, even if one processor goes down while occupying the buffer management block, other processors can still operate the buffer management block. This allows other processors to continue using the disk transfer buffer.In addition, a buffer storage register that can be input externally is provided, which allows the disk transfer buffer to be released from the When restarting, you can choose whether to save the contents before the downtime or initialize it. By choosing to save the contents before the downtime, you can continue to use the disk transfer buffer. This provides the effect that data loss in the disk transfer buffer in a data processing system having a plurality of processors can be reliably prevented.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram showing the contents of a management block,
Fig. 3 is an explanatory diagram showing the contents of tag information, Fig. 4 is an explanatory diagram showing the contents of the disk transfer buffer, and Fig. 5 is an explanatory diagram showing the contents of the disk transfer buffer.
FIG. 6 is an explanatory diagram showing the blocks of the hard disk device, FIG. 6 is a flowchart showing the processing procedure when accessing the disk transfer buffer from the processor, and FIG. 7 is a flowchart showing the processing procedure when starting up the processor. In the figure, P ₀ to P _N are processors, 1 is a system bus, 2 is a hard disk controller, 3 is a hard disk device, 4 is a buffer storage register, 5
1 is an external switch, 6 is a memory, 7 is a buffer management block, 8 is tag information, 9 is a disk transfer buffer, 11 is occupancy information, 12 is an occupancy processor number, 13 is block valid information, and 14 is a file address.

Claims (1)

[Claims] 1. When multiple processors independently use a common auxiliary storage device via a common disk transfer buffer, one processor occupies the buffer management block while using the common disk transfer buffer. In this disk transfer buffer management method, when a processor occupies a buffer management block, the number of the occupied processor is stored in the buffer management information block, and when the processor goes down, this is notified to other normal processors. Then, it is determined whether the processor occupying the buffer management block is a down processor based on the processor number of the down processor in the other normal processor, and when the determination result is that the processor is down, the buffer management block is A disk drive system characterized in that when a management block is released and the processor is started up, the contents of a buffer storage register that can be input externally are in a saved state, and the processor is started up without initializing the disk transfer buffer. Transfer buffer management method.