JPH03268043A - Disk cache control system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to disk cache control technology, and in particular,
The present invention relates to a technique effective for speeding up data exchange in a storage hierarchy having a cache memory.

[Conventional technology]

For example, in general-purpose computer systems,
Cache memory consisting of semiconductor memory, etc., which operates much faster than magnetic disk devices, is installed between high-level processing units such as central processing units and channel devices that operate at high speed, and magnetic disk drives as external storage devices. It is known to intervene.

When a higher-level processing unit requests a read of data from a magnetic disk device, the higher-level processing unit responds using the data that has been copied in advance to the cache memory whenever possible, or writes data from the higher-level processing unit to the magnetic disk unit. In this case, by performing operations such as reporting the completion of the operation to the upper processing unit at the time the relevant data is written to the cache memory, it is possible to directly access the magnetic disk device that is operating slowly due to waiting for rotation of the magnetic disk, etc. This is intended to alleviate the decrease in the processing speed of the upper-level processing device due to this.

By the way, conventional techniques for managing such cache memory include, for example, as disclosed in Japanese Unexamined Patent Publication No. 59114658, when data is written to the cache memory, regardless of the amount of data, First, a sufficient area corresponding to the capacity of one track in the magnetic disk device is secured, and after writing is completed, the storage area other than the area net used by the writing process is released from the secured area. It has been known.

[Problem to be solved by the invention]

However, with the above-mentioned conventional technology, if the amount of data is less than the storage area secured in advance, unnecessary storage area must be secured and released, and the space between the host processing unit and the magnetic disk device is correspondingly reduced. As redundant time (overhead) increases in data transfer processing via the cache memory of There is a problem in that it causes deterioration.

In addition, when the processing of storing data in the cache memory is concentrated, a shortage of storage area occurs due to wasteful allocation of storage area, and other data stored in the cache memory is frequently purged. As a result of the reduction in the residence time of data in the cache memory, there is a concern that the probability (hit rate) that the data requested by the higher-level processing device exists in the cache memory decreases, and the effectiveness of the cache memory decreases.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to prevent the occurrence of overhead in storing data exchanged between an upper processing unit and a disk device in a cache memory, and to improve data transfer between the upper processing unit and a disk device. The object of the present invention is to provide a disk cache control method that can improve the throughput of the disk cache.

Another object of the present invention is to provide a disk cache control method that can improve the hit rate in the cache memory and make effective use of the cache memory.

The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[Means to solve the problem]

A brief overview of typical inventions disclosed in this application is as follows.

In other words, the disk cache control method according to the present invention uses a cache memory that is interposed between a higher-level processing device and a rotating storage device and temporarily stores data exchanged between the two in the rotating storage device. The cache memory is managed by logically dividing it into slots corresponding to a predetermined physical or logical storage unit, and the storage area of the cache memory is allocated to each slot in units of multiple segments of a predetermined length. A disk cache control method that expands slots in the cache memory or places new slots in the cache memory before writing the data to the cache memory when storing data requested for access from a higher-level processing unit in the cache memory. After calculating the minimum number of segments necessary for the data processing and allocating the minimum number of segments, data is written to the minimum number of segments.

[Effect]

According to the disk cache control method of the present invention described above, when storing data exchanged between the host processing device and the disk device in the cache memory, the minimum number of segments required for storing the data is calculated in advance, Since the storage area for the required minimum number of segments is allocated, for example, regardless of the amount of data, the storage area for the maximum number of segments of the slot is secured, and the excess segments are released after storing the data. As a result, overhead required for securing and releasing unnecessary segments is eliminated, and throughput in data transfer between the host processing device and the disk device can be improved.

In addition, even when data storage processing is concentrated in the cache memory, only the minimum necessary storage space is secured, so the frequency of purging previously stored data will not increase more than necessary in order to secure space. This prevents a decrease in the residence time of data in the cache memory, that is, a decrease in the hit rate, and makes it possible to effectively utilize the cache memory.

〔Example〕

FIG. 1 is a flowchart showing an example of a disk cache control method according to an embodiment of the present invention, and FIG. 2 is a flowchart showing the configuration of an electronic 31w system including a disk cache control device that performs this disk cache control method. FIG. 3 is a block diagram showing an example of the configuration of a disk cache control device.

First, an outline of the configuration of the computer system of this embodiment will be explained with reference to FIGS. 2 and 3.

The central processing unit 10 (CPU) includes a channel 10a and a disk cache control unit 11, which perform control operations for exchanging information with peripheral devices on behalf of the central processing unit 10. A plurality of magnetic disk devices 14 are connected via a magnetic disk connection device 12 .

As shown in FIG. 3, the disk cache control device 11 includes a microprocessor (MPU>20) that performs overall control based on a microprogram stored in a random access memory (RAM) 21, and an upper channel 10a. a channel-to-channel transfer control device 22 that controls data transfer between the
4, and a data transfer control circuit 23 that controls data transfer while synchronizing between the channel and the magnetic disk device.

Further, the disk cache control device 11 is provided with a cache memory 50 composed of a semiconductor memory or the like that can be accessed much faster than the magnetic disk device 14, which involves mechanical operations such as waiting for rotation of a magnetic disk. Among the data stored in the magnetic disk device 14, it holds a copy of data that has a high probability of being accessed from the upper channel 10a side.
It is configured to perform operations such as temporarily holding data to be written to the magnetic disk device 14 from the beginning.

In this case, the cache memory 50 is managed by being divided into predetermined units called slots, and each slot is configured as a unit of segment with a predetermined capacity. The number of segments is set to include the number of segments corresponding to the capacity of the track.

The relationship between the data stored in each slot and the actual storage position of the data in the magnetic disk device 14 is recorded in the search management table 30. When an access request for specific data is generated by specifying an internal record number based on a cylinder number, track number, record number, etc., by referring to the search management table 30 based on this information, It is possible to know whether the data exists in the cache memory 50 (cache hit) or not (cache miss).

Further, each slot in the cache memory 50 is managed by a slot control table 40.

Here, FIG. 4 shows an example of the relationship between the data recording format in the magnetic disk device 14 and the data recording format in the slot in the cache memory 50. It is assumed that the specifications are as follows.

That is, in the magnetic disk device 14, (1) 1
024 cylinders/disk (2) 16 tracks/cylinder (3) 48 bytes/track Also, in the cache memory 50, (4) 204
8) Rack/cache memory (5) 8 bytes/segment (6) 6 segments/track.

Although not particularly shown, the slot control table 40 includes, for example, a slot number, a pointer indicating the address in the cache memory 50 of the beginning of a plurality of segments that make up the slot, the number of segments that make up the slot, and information stored in the slot. Stored data information that records information such as the record lengths of multiple records and the final record length is recorded.

Note that the plurality of segments constituting each slot do not need to be physically contiguous inside the cache memory 50, and by managing them using a well-known list structure etc. The location segment is now used effectively.

An example of the operation of a system having such a configuration and specifications will be described below with reference to the flowchart of FIG. 1 and the like.

First, the magnetic disk device 1 is sent from the central processing unit 10 to the disk cache control device 11 via the channel 10a.
A case will be described in which a read request is issued for data stored in No. 4.

If a read request is issued from channel lOa, for example, in the following CCV (channel command word) chain, the LOC/RD D/RD D/LOG command will read the magnetic field of the record targeted by the read request. A parameter indicating the storage position in the disk device 14 is given, and each RDD command actually transfers each record to be read from the magnetic disk device 14.

In the case of this embodiment, the two records (record count - 2) transferred by the two RDD commands each have 4 bytes, and if the slot containing the record does not exist in the cache memory 50. In other words, a case will be described in which it is necessary to store the record from the magnetic disk device t14 into the cache memory 50 due to a cache miss.

By receiving the LOG command from the channel 10a, the disk cashico control device 11 recognizes the type of data transfer to be executed and the restrictions attached thereto (step 101).

In the case of this embodiment, when the LOG command is received, it is recognized that it is a read request, and further based on the internal search record # etc. corresponding to the storage position information of the record in the magnetic disk unit Wt14. hand,
Searching the search management table 30 (step 10)
2) As mentioned above, since the target record (slot) does not exist in the cache memory 50, the cache memory 50 requires the arrangement of a new slot (segment) in the cache memory 50 and the corresponding information from the magnetic disk device 14. The process moves to slot allocation, which copies the record to a slot (segment).

That is, first, the total transfer data length is calculated from the record count and transfer length accompanying the command.

In this case, record count (2) x transfer length 4 bytes is used to calculate total transfer data length=8 bytes (step 103)'.

Furthermore, from this total transfer data length of 8 bytes, the required allocation is 1 segment (&! transfer data length of 8 bytes/8 bytes/segment = 1 segment) (step 104), and A new slot containing one segment is allocated to the corresponding segment, and the corresponding data (two records) is copied from the magnetic disk device 14 to the one segment (Fig. 5(a)).
.

Note that the transfer of the read-requested data to the upper channel 10a may be performed in parallel with the copying of the data to the cache memory 50, or the data may be transferred to the upper channel 10a side.
After copying the data to 0, the data may be read and transferred.

Here, in the conventional case, as mentioned above, when securing the slot, after data transfer is performed by allocating the maximum number of segments that make up a slot, 6, regardless of the total transfer data length, the remaining 5 segments are Since the operation of releasing the segments is performed, overhead is generated for securing and releasing the five segments.

However, in the case of this embodiment, as described above, prior to data transfer from the magnetic disk device 14 to the cache memory 50, one segment, the minimum number required to configure a slot, is secured in advance, and the corresponding segment is Since the target record is stored in the segment, such overhead does not occur.

Next, the channel 10a is sent to the magnetic disk device 14.
The processing when a data write request is generated from is explained below.

For example, if the write request is a record update process using the following CCW chain, LOC/WRT UP
The D/LOG command specifies various parameters such as the storage location in the magnetic disk device 14 of the record to be written, and the WRT UPD command
Data to be updated and written is transferred.

In the following description, a case will be described in which a slot including a record to be updated exists in the cache memory 50 and the update process is performed as a cache hit.

In other words, in the cache memory 50, two segments are currently used for slots containing records that meet the conditions of the LOG command, and three records each having 5 bytes = 15 bytes of data are stored. Assume that the actual data length in the last record is 3 bytes. It is assumed that the update process is performed on this last record.

First, upon receiving an L0C command (step 101), the disk cache control device 11 searches the search management table 30 based on information such as an internal search record number accompanying the command, and determines whether it is a hit or a miss (step 101). 102).

As described above, since a slot containing the target record exists in the cache memory 50, it is determined that the cache hit is to be made, and the process moves to segment allocation.

In other words, using the pointer to the first segment in the cache memory 50, the number of segments, stored data information, etc. regarding the slot in the slot control table 40, first, calculate the stored data length excluding the record to be updated at the final position. Calculate (step 105).

That is, the stored data length excluding the record to be updated at the final position = (current data length of the corresponding slot, 15 bytes) - (length of the last record to be updated, 3 bytes) = 12 bytes.

Furthermore, the total transfer data length (record length) is calculated from the value of the record counter accompanying the LOG command (1 in this case) and the average transfer record length (5 bytes in this case) (step 106).

Then, by adding bytes to the storage data length 12 excluding the record to be updated obtained in step 105 and adding bytes to the total transfer record length 5 obtained in step 106, the total data length 17 is reduced by bytes after the update process. calculate(
Step 107).

Furthermore, from this total data length 17 bytes and the capacity per segment 18 bytes, calculate the number of segments - 3 required to store the bytes in the total data length 17 (step 108), and further calculate the number of segments. By subtracting the number of segments (2) previously used from (3), the number of segments (1) required due to the increase in the length of the final record due to the update process is calculated (step 109), allocates (adds) a new storage area for the one segment, and then
The target final record is updated using this one segment and the previous two segments (FIG. 5, etc.)).

Here, in the conventional case, first, after allocating four new segments to the number of segments used so far (2) so that the number of segments of the slot to which the update record belongs is maximized, Since only 3 segments are actually used, it is necessary to release the surplus 3 segments, and overhead is generated to allocate and release the surplus 3 segments. However, in the case of this embodiment, As mentioned above, since only the minimum necessary segments are allocated, the overhead unlike the conventional method does not occur.

As explained above, in the disk cache control method of this embodiment, the upper channel 10a (central processing unit 10
) and the magnetic disk device 14 via the cache memory 50 (read/write) is suppressed. 1
The throughput of data transfer between 4 and 4 is definitely improved.

Furthermore, when storing data in the cache memory 50, an unnecessarily large storage area is not allocated, so that the cache memory 50 is
The amount of other data that is evicted from 0 is reduced, that is, the residence time of the entire data in the cache memory 50 becomes longer, and the probability (hit rate) that data requested for access from the channel lOa side exists in the cache memory 50 increases. This will surely increase the cost, and the cache memory 50 can be used more effectively.

Although the invention made by the present inventor has been specifically explained above based on Examples, it goes without saying that the present invention is not limited to the Examples and can be modified in various ways without departing from the gist thereof. Nor.

For example, the storage medium constituting the cache memory is not limited to semiconductor memory, and similar effects can be obtained even when other nonvolatile storage media are used.

〔Effect of the invention〕

Among the inventions disclosed in this application, the effects obtained by typical inventions are briefly described below.

That is, according to the disk cache control method of the present invention, the cache memory that is interposed between the upper processing device and the rotating storage device and temporarily stores data exchanged between the two is controlled by the rotating storage device. Storage device j: is managed by logically dividing it into slots corresponding to a predetermined physical or logical storage unit, and each slot stores the cache memory in units of multiple segments of a predetermined length. A disk cache control method that allocates a storage area, wherein when storing data requested for access from the higher-level processing device in the cache memory, the data is stored in the cache memory prior to writing the data to the cache memory. After calculating the minimum number of the segments necessary for expanding the slot or arranging a new slot in the process, and allocating the necessary minimum number of the segments, performing a write operation of the data to the required minimum number of the segments. This reduces the need to allocate and release unnecessary segments, compared to, for example, securing storage space for the maximum number of segments in a slot regardless of the amount of data, and then releasing the excess segments after storing the data. This eliminates the overhead required for data transfer between the host processing device and the disk device, and improves the throughput of data transfer between the host processing device and the disk device.

In addition, even if the processing of storing data in the cache memory is concentrated, only the minimum necessary storage space is secured, so the frequency of purging previously stored data increases more than necessary in order to secure space. This prevents a reduction in the residence time of data in the cache memory, that is, a reduction in the hit rate, which is caused by this, and makes it possible to effectively utilize the cache memory.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing an example of a disk cache control method according to an embodiment of the present invention, and FIG. FIG. 3 is a block diagram showing an example of the configuration of a computer system; FIG. 3 is a block diagram showing an example of the configuration of a disk cache control device; FIG. 4 is a block diagram showing an example of the configuration of a disk cache control device; FIG. A conceptual diagram showing an example of the correspondence between data recording formats in a magnetic disk device and a cache memory in a computer system, Figures 5 (a) and 5) are conceptual diagrams showing an example of data storage states in slot allocation and segment allocation. It is a diagram. DESCRIPTION OF SYMBOLS 10...Central processing unit (CPU), 10a...Channel (upper processing unit), 11...Disk cache control device, 12...Magnetic disk device, 14...Magnetic disk device (rotating type) 20... Microprocessor (MPU), 21... Random access memory (RAM), 22... Channel transfer control device, 23... Data transfer control circuit, 24... Disk disk Transfer control device, 30...Search management table, 4
0...Slot control table 50...Cache memory, 101-109...Processing steps showing an example of disk cache control method.

Claims (1)

[Claims] 1. A cache memory that is interposed between the upper processing unit and the rotating storage device and temporarily stores data exchanged between the two is a physical or logical predetermined storage in the rotating storage device. A disk cache control method that logically divides and manages slots corresponding to units, and allocates storage areas of the cache memory in units of multiple segments of a predetermined length to each slot. , when storing data requested for access from the higher-level processing device in the cache memory,
Prior to writing the data to the cache memory, calculate the minimum number of segments necessary for expanding the slot in the cache memory or arranging a new slot, and after allocating the required minimum number of segments. , a control method for a disk cache, characterized in that a write operation is performed to a necessary minimum number of said segments.