JPS61267862A - File processing system - Google Patents

Abstract

PURPOSE:To improve the processing efficiency and the reliability of a file processing system by adding a block control table to a file device applying an exclusive control system and performing the exclusive control of a block level automatically when a request for input/output execution is given to a virtual volume. CONSTITUTION:A network system comprises data processors 1-1-1-n, a virtual disk type file device 2 including a fixed disk device and a bus 3 connected in common to these processors 1-1-1-n and the device 2 respectively. A block control table is provided to the device 2 to control a state where each task uses exclusively the space of each virtual volume for each block forming a virtual volume within the device 2. Then the exclusive control is carried out automatically to the block level as necessary when a request for input/output execution is given to the virtual volume from each task.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a file processing system including at least one file device shared by a plurality of data processing devices.

[Conventional technology]

In a file processing system in which at least one file device is shared by multiple data processing devices via a network, appropriate exclusive control is applied to requests to update data stored in the file device by tasks on the data processing device side. If this is not done, update operations required by the task cannot be performed correctly, as will be described in detail later.

The virtual disk type file device used as the file device described above has a function of emulating, on the file device side, individual disk devices owned by individual data processing devices that share the file device. By connecting such a virtual disk type file device to a network, the data processing device side can access the virtual disk type file device as if it were a friend disk device connected at hand. An example of a conventional virtual disk type file device having such a function is disclosed in Nikkei Electronics Magazine, February 27, 1984, pp. 179-198.

[Problem that the invention seeks to solve]

However, as will be described in detail later, in the conventional system, task execution is serialized on a virtual volume basis, which has a major drawback in that it takes a long time to execute the task as a whole.

[Means for solving problems]

The system of the present invention includes: a plurality of data processing devices each having at least one task; at least one file device shared by the plurality of data processing devices and having a plurality of storage areas; sidecar state storage means storing allocation information indicating an allocation state to a task; lock information storage means storing lock information for a plurality of storage blocks constituting each of the plurality of storage areas; When a first task updates the contents of the block in the storage area indicating that the block is not allocated, this block until the read and write operations of this first task to this block are completed. Registering and retaining the lock information for this block in order to prohibit access from at least one other second task to the block, and removing this lock information from the lock information storage means after these operations are completed. It consists of a processing means.

〔Example〕

In order to facilitate understanding of the present invention, a conventional system will first be described.

Referring to Figure 2, a general network system is
n (positive integer) data processing devices 1-1 to l-n, a virtual disk type file device 2 including a fixed disk device, and a bus to which these devices 1-1 to 1-n and 2 are commonly connected. It consists of 3. As the file device 2,
Virtual server VS100/2 manufactured by Compact Microelectronics and Minato Electronics
0°VS100/40. VS100/80t or vs1
00/160 can be used. For example, VS100/16
0 is LSI-11/23 manufactured by DEC as a CPU
- PLU8, a main memory unit with a storage capacity of 1 MB, a Winchester disk unit with a storage capacity of 160 MB, and two floppy disk units each with a storage capacity of 0.5 MB. The operating system used is EC's R8X-11M, which is capable of real-time multitasking. Now, as shown in FIG. 3, at timing tl to t6, the data processing device 1
-Task person on 1 and task B on data processing device 1-2
Storage record R of the disk in the Toyoshi file device 2
The update operation is performed in parallel for 9. In this case, the read operation for the same record by task B at time t2 is also performed in advance of the write operation for the same record R by task A at time t4. This is because, at time t2, the update operation for all records by task A has not been completed. If we look at time t4, which is the end point of the update operation of the record by task B, it can be seen that the update operation of the record by the task person has become invalid.

For these reasons, when data in the same record is updated in parallel by multiple tasks, the exclusive control function for data is used to serialize the update operations of multiple tasks on the same data so that they do not overlap with each other. It is necessary. That is, the read operation of record R by task B may be delayed until the end time t4 of the write operation to record R by task AK.

In the conventional exclusive control method for the file device 2, the file device 2 receives a MOUNT command indicating a virtual volume allocation request via the bus 3 from the data processing device of the user who does not want to access the file device 2. do. This command includes two pieces of information regarding the virtual volume allocation conditions at the time of allocation. One of them is to allocate the same virtual volume to other tasks at the same time and use this virtual policy.

This is information that specifies whether you want the task to be shared (shared allocation) with other tasks or to be exclusively allocated (exclusive allocation), and this specification information is called a shared mode designation. The other is the read operation (after allocation to the virtual volume).
This information specifies whether to perform only reference allocation (reference allocation) or whether not only read operation but also write operation (update allocation) is to be performed. This specification information is called processing mode designation. Regarding the shared mode specification, in addition to the above-mentioned shared allocation and exclusive allocation, there is a second shared mode in which the same virtual volume can be allocated at the same time only to other tasks that request reference allocation. In some cases, it is possible to designate "assignment" (referred to as "assignment"), but for the sake of simplicity, we will limit the case here to the case where only shared assignment and exclusive assignment are designated as described above. The file device 2 not only has the function of receiving the above-mentioned mount commands and input/output execution commands, but also has the function of managing the state of allocation of each virtual volume to individual tasks. FIG. 4 shows an example of a management table (referred to as a volume allocation control table (VACT)) stored in the main storage device 22 used for this purpose. In this figure, virtual volumes 1, 2, 4, and 6 are not assigned to any task, and virtual volumes 3 and 5 are not assigned to any task.
is assigned to the task. For example, virtual volume 5 is simultaneously allocated to tasks A and B, respectively, with shared allocation/reference allocation. CPU in file device 2
Upon receiving a MOUNT command indicating a virtual volume allocation request from a task on the data processing device, 21 examines the shared mode designation information included in the command. CPU21
checks VACT in MM22, and if it detects that the virtual volume specified by this shared mode specification information is not allocated to any other task at that time (that is, unallocated), it will issue this allocation request. Acceptance is accepted, the allocation result is reflected in VACTK, and the virtual volume allocation operation is completed. However, if the requested virtual volume has already been assigned to another task, the CPU 21
If the allocation request is found to be consistent with the allocation state of the virtual volume (if it is shared with another task and this request is also shared allocation), the request will be accepted. However, if they are mutually contradictory (if the request is shared with other tasks and the request is exclusive assignment),
Or, if the task is exclusively assigned to another task and the request is for shared assignment or exclusive assignment), this request will not be accepted.

Needless to say, if the request is approved, this result will be reflected in VCAT. In this way, exclusive assignment of virtual volumes to tasks becomes possible.

Furthermore, the CPU 21 in the file device 2 issues a request command (Read command or Write command) to execute an input/output operation by a task.
Each time a te command is received, processing is performed according to the processing flow shown in FIG. Here, check whether the virtual volume has been allocated correctly in advance by the task.
Furthermore, it is reliably checked whether a task attempts to perform a write operation even though the virtual volume is reference-allocated.

Finally, when the task no longer needs to allocate a virtual volume and receives a DEMOUNT command from the task to cancel the virtual volume allocation, the CPU 21
Modify the CT to unassign the virtual volume by that task. As described above, serialization of task volume update operations using the conventional exclusive control method is realized.

That is, when multiple tasks update the same code as shown in FIG.

Each task specifies exclusive allocation/update allocation as an allocation condition when requesting virtual volume allocation using the MOUNT command, so that each task is sequentially allocated one record at a time. As a result, update operations for the same record are serialized.

In such conventional methods, during update processing, the virtual volume to be updated can be assigned to only one task at a time. Therefore, for example, if we define this number of multiple tasks as m
(a positive integer), it will take a long time, m times the time required for an update operation by one task, for all tasks to perform an update operation on the same virtual volume.

One embodiment of the present invention has a configuration similar to that shown in FIG. Therefore, the task on the data processing device is the shared file device 2.
When using , the task first issues a MOUNT command for a virtual volume allocation request to the file device as in the conventional case.

The method of specifying virtual volume allocation conditions in this embodiment is the same as the conventional method. Furthermore, the file device 2 in this embodiment also has a VACT in the main storage device 22 similar to the conventional one, and the CPU 21 processes the MOUNT command by the task in exactly the same manner as the conventional one. That is, the virtual volume allocation process is performed in exactly the same way as in the conventional system.

Thereafter, the task that successfully allocated the virtual volume requests the virtual volume to perform an input/output operation.

At this time, if the shared mode designation at the time of virtual volume allocation by the task is exclusive allocation, the CPU 21 performs the processing according to the processing flow shown in FIG. 5 in exactly the same manner as before. If the result is normal, input/output operations are executed. However, if the shared mode specification at the time of virtual volume allocation by a task is shared allocation, there is a possibility that there are update tasks that are simultaneously allocated to the same virtual volume, so the CPU 21 performs the following process. This method performs block-level automatic exclusive control.

There are various ways to implement block-level exclusive control, but the simplest one is based on the principle that each virtual volume allows exclusive use of at most one block at a time by one task. In other words, in this embodiment, a lock control table (LCT) as shown in FIG. When executing this task, end vrtc lock information of the LCT containing the data block to be read is registered, and this lock information is removed from the LCT after this task executes a write operation on the same data block.In FIG. None of the blocks in virtual volumes 1.2.3.4 and 6 are locked (access prohibited).However, in virtual volume 5, the third block B3 is locked by task A, and the fourth block B4 is locked by task A. Locked by B.

Next, details of the process executed by the CPU 21 when the CPU 21 receives an input/output operation execution request command will be explained with reference to FIG. First, in step 61, when a task receives a command requesting execution of an input/output operation to a virtual volume (step 61), the volume specified by this command is
VA is assigned to this task.
A CT scan is performed (step 62). R
It is determined whether it is an (ead) operation or a write operation (step 64). If it is a read operation, first check the LCT, and if there is a block in the target volume that is locked (lock information is registered) by that task, it is unlocked (lock information is removed). (step 65), attempts to lock the block to be read (step 66)
. This attempt will succeed only if the block is already locked by another task. If not successful,
This task is made to wait until the block is unlocked by another task and can be locked. When the lock is successful, the CPU 21 reads the contents of the block from the corresponding disk and transmits the data to the task on the data processing device that issued the read command (step 67).

On the other hand, if the execution request command from the task is a data write command, it is found that the corresponding volume has been updated and allocated (step 68).
, confirms with the LCT that the block is locked by the task (step 69), and writes data from the task to the block (step 70). After this, the block is unlocked (step 71).

Finally, upon receiving a DEMOUNT command for virtual volume deallocation from a task, the CPU 21 in the file device 2 uses the LCT to check whether there are any remaining blocks locked by that task, and if so, to unlock. Subsequently, by deleting all information related to this task from VACT, the assignment of the virtual volume to this task is released.

In this way, in this embodiment, even though the same virtual volume is allocated to multiple tasks at the same time, block-level update operations are automatically serialized during actual input/output operations. Therefore, the overall execution time of multiple tasks does not increase.

In the above description, a simple embodiment has been described based on the premise that each virtual volume allows exclusive use of at most one block at a time by one task.

However, in the above embodiment, even if it is extended so that each task can lock multiple consecutive blocks to only one location on the virtual volume, the basic method of realizing exclusive control remains the same. Add . Furthermore,
It is also possible to extend this so that each task can lock multiple blocks or multiple locations in the virtual volume. In this case, locking is not required each time it is needed, and unlocking can be performed all at once when virtual volumes are deallocated.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, the space of each virtual volume is allocated to a file device that uses the conventional exclusive control method in units of blocks that constitute each virtual volume in the file device. A means (block control table) is added to manage the state of exclusive use by individual tasks, and when an individual task requests execution of input/output to a virtual volume, it is automatically executed according to the processing flow shown in Figure 6. By performing block-level exclusive control, multiple tasks can perform parallel update operations on the same data without increasing the overall task execution time and without running the risk of losing updates. The effect of improving processing efficiency and reliability is significant -0.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is a block diagram showing details of a file device, and Fig. 3 is an update when multiple tasks update the same data in parallel according to the conventional technology. FIG. 4 is a diagram illustrating how a loss of operation can occur, and is used to manage the state of assignment of each virtual volume to a task under the conventional exclusive control method and the exclusive control method according to the present invention. FIG. 5 is a diagram showing an example of the configuration of a control table, FIG. 5 is a diagram showing an example of a processing flow when an input/output request command received by a file device is processed under the conventional exclusive control method, and FIG. 6 is a diagram showing an example of the processing flow according to the present invention. Figure 7 shows an example of the processing flow when a file device receives and processes a nine-person output request command under the exclusive control method according to the present invention. FIG. 3 is a diagram illustrating a configuration example of a control table used to manage lock states by each task in a block. In the figure, 1-1 to 1-n... data processing device, 2... file device. Agent Patent Attorney Uchihara (Shinmata-3 Figure 5 Kaya z TM

Claims (4)

[Claims] (1) a plurality of data processing devices each having at least one task; at least one file device shared by the plurality of data processing devices and having a plurality of storage areas; allocation state storage means storing allocation information indicating an allocation state; lock information storage means storing lock information for a plurality of storage blocks constituting each of the plurality of storage areas; When a first task updates the contents of the block in the storage area indicating that the first task has not updated the contents of the block in the storage area, at least and processing means for registering and retaining the lock information for this block in order to prohibit access from other second tasks, and for removing this lock information from the lock information storage means after these operations are completed. A file processing system characterized by: (2) The processing means, in response to an allocation request issued by the first task, checks the allocation state of the storage area specified by this allocation request with reference to the allocation state storage means, and Claim (1) characterized in that, in order to allocate this storage area to a task, a first process is executed to determine whether or not the allocation information is stored in the allocation state storage means. File processing system described. (3) The processing means receives a read request that is issued from the first task and includes first designation information that designates one of the storage areas and second designation information that designates at least one of the blocks; by referring to the allocation state storage means in response to fourth designation information including third designation information designating one of the storage areas and fourth designation information designating at least one of the blocks; Said first
or a second process for detecting whether the storage area specified by the third specification information is exclusively allocated to the first task; When it is detected that the block specified by the second specification information is not specified by the lock information storage means in response to the read request, the block specified by the second specification information is
a third task that detects whether lock information belonging to the task and at least one other second task is registered;
a fourth process of prohibiting a read operation for the first task in response to the detection of the lock information of the second task in the third process; When unregistered lock information belonging to the first and second tasks is detected, lock information for the first task is registered in the lock information storage means, and the file device specified by the second designation information is registered. a fifth process of reading out at least a part of the block for the first task; and in response to the write request when it is detected in the second process that the block is not exclusively allocated. a sixth process of writing data from the first task into at least a part of the block specified by the fourth specification information; and a sixth process of writing data from the first task to at least a part of the block specified by the fourth specification information, and a process for writing data to the corresponding block after the write operation in the sixth process is completed. Claim (1), characterized in that a seventh process of deleting the lock information is executed.
File processing system as described in Section. (4) The processing means, in response to an allocation release request issued from the first task, processes the first task in the allocation state storage means corresponding to the storage area specified by the allocation release request. Claim 8, characterized in that an eighth process is executed to remove the allocation information and also remove the lock information in the lock information storage means of the first task corresponding to the block belonging to this storage area. (2)
File processing system as described in Section.