JPH05181721A - Method for reusing and retrieving shared buffer - Google Patents

Abstract

PURPOSE:To eliminate the necessity of access time to a magnetic disk device at the time of reusing it and to attain rapid access by determining a shared buffer which is not in an updating state as a shared buffer for reuse. CONSTITUTION:This system is provided with a CPU 1 including a shared buffer retrieving means 10 and an asynchronous input procedure 11, a main storage device 2 including a shared buffer storing part 12 and a magnetic disk device 13 as an external storage device. Shared buffers are connected to a queue in an access time sequence from the oldest buffer up to the newest one and the existence of updating state is successively retrieved and judged from the queue. If the updating status exists at the time of purging a certain shared buffer from the device 2, the buffer is asynchronously written in the device 13, and at the time of accessing the device 13, the retrieval and judgement of the succeeding queue can be continued without waiting the completion of the writing. Since the contents with that of the device 13, the unupdated shared buffer can be determined as a shared buffer for reuse without accessing the device 13.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to shared buffer management of shared files in a file system, and more particularly to a shared buffer reuse search method effective for speeding up the shared buffer reuse.

[0002]

2. Description of the Related Art Management of a shared buffer in a shared file is a technique for reducing access to a magnetic disk device as an external storage device and speeding up file access. The shared buffer is a buffer in which the contents of files accessed by a plurality of programs are recorded in the main storage device. Since the access from the program only requires the memory transfer time in the main storage device, the access speed is 10 times or more as high as the access from the magnetic disk device. It is possible to increase the speed by placing all the contents of the magnetic disk device in the main storage device as a shared buffer. However, the number of shared buffers is limited to a fixed value when the system is constructed because the memory of the main storage device is limited. Therefore, the shared buffer manages a part of the contents in the magnetic disk device. The shared buffer manages the contents of the disk in blocks. A block is a unit for inputting / outputting with a magnetic disk. Normally, the block has a fixed length.

An LRU (Least Recently Used) method is used to efficiently manage a limited number of shared buffers. The LRU method is a method of arranging shared buffers in the order of block access, and expelling the oldest accessed shared buffer from the main storage device. When the buffer that has been accessed once is accessed again, it is rearranged last. When the shared buffer is expelled, if there is a change in the block, it takes time to access the magnetic disk drive for writing, but if there is no change, the shared buffer can be reused immediately. In the following description, eviction from the main memory and using it for another block will be referred to as buffer reuse. Whether or not the block has changed shall have a state indicating that the block has been updated when the shared buffer was written.

[0004]

However, the above-mentioned shared buffer management method cannot eliminate the access time for writing to the magnetic disk device when the shared buffer is reused. Therefore, the advantage of accessing the shared buffer is lost. That is, there is a drawback that the access time to the magnetic disk device is the same as the file access time.

It is an object of the present invention to eliminate the access time of the magnetic disk device when reusing the shared buffer and to speed up the access to the shared buffer.

[0006]

In order to solve the above-mentioned problems, a method for reusing a shared buffer according to the present invention sequentially searches the shared buffer queue to determine whether it is in an updated state or not, and updates the state. The shared buffer of is asynchronously written to the external storage device, and the shared buffer that is not in the updated state is determined as the reused shared buffer.

In the above method, when the shared buffer is in the updated state until the end of the queue, the head shared buffer of the queue may be determined as the reused shared buffer.

[0008]

The shared buffer is connected to the queue in the order of oldness of access, and the presence or absence of the update state is sequentially searched and determined from the queue. If there is an update when flushing the shared buffer from the main storage device, it can be asynchronously written to the external storage device, and the search judgment for the next queue can be continued without waiting for completion when accessing the external storage device. . If it is not in the updated state, the contents of the shared buffer and the contents of the external storage device match, so the external storage device is not accessed and the shared buffer that is not in the updated state is determined as the reused shared buffer.

When the shared buffer is in the updated state up to the end of the queue, the completion of the asynchronous writing of the head shared buffer of the queue is waited for, and this is determined as the shared buffer for reuse to make effective use of the CPU.

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a flow chart showing an embodiment of a reuse procedure according to the present invention. FIG. 2 is a functional block diagram for reusing a shared buffer. In the CPU 1, a shared buffer search means 10 and an asynchronous input / output procedure 11 are provided, and a shared buffer storage unit 12 is provided in the main storage device 2. A magnetic disk device 13 is provided as an external storage device. FIG. 3 is a diagram showing a state in which the shared buffers are arranged in the order of access (queue). As shown in FIGS. 3A and 3B, the shared buffers are queued in chronological order of access in the order of oldness. The queue points to the shared buffer one by one from the beginning, and the last NULL 14 is at the end.

First, as shown in FIG. 3 (a), the queue has a1, a2, and a3 that have an update status and b1 and b that have no update.
The case where 2 and 2 are lined up will be described. When reusing, the shared buffer a1 is selected from the head of the queue (FIG. 1, step P1). It is checked whether it is the end of the queue (step P2), and the process goes to no. The update status of a1 is checked (step P3), and since it is the update status, the process proceeds to yes. Here, the asynchronous input / output procedure 11 of FIG. 2 is used to asynchronously write from the shared buffer storage means 12 to the magnetic disk device 13 (step P4). Since asynchronous writing does not wait for completion when the magnetic disk device 13 is accessed, the procedure execution of the shared buffer search means 10 can be continued. The next shared buffer a2 is selected from the queue (step P5). Return to step P2, a2
Repeats the same processing as a1 since it is not the last in the queue, and selects the next shared buffer b1 in step P5.
Furthermore, the process returns to step P2, and since it is not the end of the queue, it is checked in step P3 if it is in the updated state. Since b1 is not updated, the process proceeds to no. Since the contents of b1 have not been updated, the contents of the shared buffer and the contents of the magnetic disk device 13 match. Therefore, the magnetic disk device 13
You can discard the contents of the shared buffer without accessing. Therefore, b1 is determined as the reuse shared buffer (step P8).

Next, as shown in FIG. 3B, a case where there is no update until the end of the queue will be described. In FIG. 3B, c1, c2, c3, c4 having update status in the queue.
c5 are lined up. Shared buffer c1 from the beginning of the queue
To select the shared buffers in sequence as c2, c3, c4, c5
The procedure for selecting is as described above. When the next shared buffer in step P5 is selected in the shared buffer c5,
It becomes NULL 14 at the end of the queue. At step P2, since the last NULL 14 in the queue, the process proceeds to yes. This is the case when the shared buffer that is not updated cannot be found. At this time,
The oldest shared buffer c1 at the head is reused. Using the asynchronous input / output procedure 11, the completion of the asynchronous writing of the shared buffer c1 at the head is completed by the magnetic disk device 13
Waits for notification from (step P6). When the writing is completed, the top shared buffer c1 is selected (step P7). In step P8, c1 is determined as the reuse buffer.

[0013]

As described above, according to the present invention, when the shared buffer is reused, the access time of the magnetic disk device can be shortened and high speed access can be achieved. Further, the asynchronous input / output allows the CPU to be effectively used because the oldest buffer is written.

[Brief description of drawings]

FIG. 1 is a flowchart showing an embodiment of a reuse procedure according to the present invention.

FIG. 2 is a functional block diagram for reusing a shared buffer.

FIG. 3 is a diagram showing a state of an access order arrangement (queue) of a shared buffer.

[Explanation of symbols]

 10 shared buffer search means 11 asynchronous input / output procedure 12 shared buffer storage unit 13 magnetic disk device (external storage device)

Claims (2)

[Claims] 1. A shared buffer in which a shared buffer in an updated state is asynchronously written to an external storage device, and a shared buffer that is not in an updated state is reused. A method for reusing a shared buffer for reusing a shared buffer. 2. A shared buffer in which the shared buffer in the updated state is asynchronously written to an external storage device, and the shared buffer in the non-updated state is reused. In the shared buffer reuse search method determined in step 1, when the shared buffer is in an updated state up to the end of the queue, the shared buffer for reuse is characterized in that the first shared buffer in the queue is determined as the shared buffer for reuse. retrieval method.