JPH0682342B2 - Disk cache control method - Google Patents

Description

The present invention relates to a disk cache control technique, and more particularly,
The present invention relates to a technique capable of replacing data in a cache memory so as to match different access forms of a plurality of application programs.

[Conventional technology]

For example, in an information processing system such as an electronic computer system, as one technique for improving the effective performance of data transfer between a magnetic disk device used as an external storage device and a host main storage device, The cache technology is known.

That is, in the data transfer path between the magnetic disk device and the main storage device, a cache memory such as a semiconductor memory that can be accessed at a much higher speed than that of the magnetic disk device is installed and stored in the magnetic disk device. Copy some data to this cache memory.

Then, when a read command for transferring data from the magnetic disk device to the main memory device or a write command for transferring data from the main memory device to the magnetic disk device is issued, respectively, from the cache memory to the main memory device or to the main memory device. When data is transferred from the device to the cache memory, the completion of the read or write command is reported to a higher-level central processing unit, etc., so that a plurality of concentric circles are provided on the magnetic disk that is the storage medium of the magnetic disk device. Do not include in the read / write command the time required for the seek operation to position the magnetic head on one of the tracks, the rotation waiting time until the recording position of the target data in the track arrives just below the magnetic head, etc. In order to speed up the data input / output processing between the main storage device and the magnetic disk device. That.

Generally, the storage capacity of the cache memory as described above is
It is orders of magnitude smaller than the total storage capacity of the magnetic disk devices under it. Therefore, it is necessary to effectively copy the data having the highest probability of being accessed by the next read / write instruction from the magnetic disk device to the cache memory. As an example of such a technique, so-called LRU (Lea
st Recently Used) method is widely used. This LP
As an example of U, if the accessed data is in the cache memory, leave it in the cache memory as it is,
If there is no such data, there is such a method that the data accessed at the oldest time among the data in the cache memory and the data currently accessed are exchanged.

However, no matter what cache memory data replacement method is used, another application program that accesses another data (file) that is completely different from the data (file) that was being accessed up to a certain point has started operating. In that case, of course, the target data cannot be found in the cache memory.

Also, when configuring the cache memory with non-volatile memory such as battery-backed semiconductor memory for power failure countermeasures, etc., when the power of the disk cache subsystem including it is turned on, data is not in the cache memory. As in the case described above, a state occurs in which the target data does not exist in the cache memory at all.

As a means for solving this, there is a technique disclosed in Japanese Patent Laid-Open No. 58-105360.

That is, a non-volatile memory is provided to save the contents of the extent memory that stores the information that associates the data stored in the cache memory with the data in the magnetic disk device, and after the power is turned on, the first cache memory becomes full. The contents of the extent memory at that time are saved in the non-volatile memory, and based on the contents of the extent memory saved in the non-volatile memory before the access to the magnetic disk device is started when the power is turned on again. The data is copied from the magnetic disk device onto the cache memory to restore the state of the cache memory.

[Problems to be Solved by the Invention]

In an electronic computer system having a magnetic disk device as an external storage device, there are relatively many operating modes in which different types of application programs as described above operate in tandem.

For example, a system that provides online services during the day and batch processing at night is one of them.
In this case, the access to the magnetic disk device including the files for online processing is concentrated during the day, and the access to the magnetic disk devices including the files for batch processing is concentrated during the night. In general, online processing files and batch processing files are often stored in different magnetic disk devices.

Therefore, immediately after shifting from batch processing to online processing, and immediately after shifting from online processing to batch processing, the target data does not exist in the cache memory,
Until a state (miss state) that requires direct access to the magnetic disk device with a long access time occurs, the data in the cache memory is replaced, and it becomes a steady state,
The state where the hit rate (probability that the target data exists in the cache memory) remains low continues, and during this period, there is a problem that the provided cache memory cannot function effectively.

Such a problem occurs not only when the batch / online business is switched but also when the various application programs are switched. In such a case, the technique disclosed in Japanese Patent Laid-Open No. 58-105360 can be used. It cannot be effective.

Further, in recent years, the capacity of the cache memory has been increasing due to the cost reduction of the semiconductor memory and the like, and the state of the low hit rate until the cache memory reaches the optimum steady state tends to be long. The decrease in hit rate when switching application programs
It becomes an increasingly important issue from the viewpoint of improving the performance of the entire information processing system.

Therefore, it is an object of the present invention to provide a disk cache control method capable of preventing a reduction in hit rate when switching a plurality of application programs.

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

[Means for Solving the Problems]

The following is a brief description of the outline of the typical inventions among the inventions disclosed in the present application.

That is, the disk cache control method according to the present invention includes a high-order processing device, a magnetic disk device that stores data exchanged with the high-order processing device, and a magnetic disk device and the high-order processing device. It consists of an intervening cache memory that holds a copy of the data stored in the magnetic disk device. Whenever possible, the data stored in the cache memory is responded to in response to an access request from the host processor to the magnetic disk device. In the information processing system adapted to respond by using the first information indicating the storage position in the magnetic disk device of the data stored in the cache memory during execution or at the end of execution of an arbitrary program executed in the host processing device. The storage location information, which is an information group, is converted by the channel command word from the host processor. A second process consisting of an identification number for identifying a program for each of a plurality of first time points is recorded and stored in a non-volatile memory, and at a second time point different from the first time point, upper processing is performed. The channel command word from the device selects and reads the first information group recorded and stored in the nonvolatile memory at the first time point based on the second information and reads the magnetic field based on the first information group. Data is copied from the disk device to the cache memory.

[Action]

According to the above-described disk cache control method of the present invention, for example, a first instruction such as during execution or termination of an individual application program is triggered by a command from a host processing device or an external hardware switch setting.
At the point of time, the first information group indicating the storage position of the data existing in the cache memory in the magnetic disk device is recorded and saved in the non-volatile memory together with the identification number corresponding to the application program. At the time of starting or at a second time point during operation after starting, the first information group is read from the non-volatile memory based on the identification number corresponding to the application program, and the magnetic field is read using the first information group. Data can be copied from the disk device to the cache memory and the state of the cache memory can be restored to the optimum state for the application program. rate It can be reliably prevented from being lowered.

[Embodiment 1] An example of a disk cache control method according to an embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an essential part of an information processing system in which a disk cache control method according to an embodiment of the present invention is implemented, and FIG. 2 is a block diagram showing an example of its overall configuration. Further, FIG. 3 is a diagram showing an example of a storage format of information.

First, the outline of the configuration of the information processing apparatus in this embodiment will be described.

The central processing unit 1 forming the center of the system and performing desired arithmetic operations and control operations stores programs and data to be processed for controlling the operations of the central processing unit 1 to perform desired operations. The main storage device 2 and the channel device 3 are connected.

The channel unit 3 receives from the central processing unit 1 a channel command word CCW issued when a read / write command is issued to an external storage unit or the like, and instead of the central processing unit 1, the main storage unit 2 and the external storage unit. It is designed to control the exchange of data and commands with and the like.

A plurality of magnetic disk devices 7 having an order of magnitude larger storage capacity than the cache memory 5 are connected to the channel device 3 via a disk control device 4 having a cache memory 5 composed of a semiconductor memory or the like. .

The disk control device 4 includes at least two devices between the upper channel device 3, the cache memory 5, and the magnetic disk device 7.
It has a function to asynchronously transfer data between the two. Then, at any time, a part of the data of the magnetic disk device 7 is copied to the cache memory 5, and for the read / write command issued from the channel device 3, the data stored in the cache memory 5 is copied as much as possible. By responding with the use, the frequency of direct access to the magnetic disk device 7 which involves mechanical operations such as seek operation and rotation waiting is reduced, and data transfer between the magnetic disk device 7 and the channel device 3 is reduced. It is designed to improve performance in operation.

The cache memory 5 includes a data area 51 for storing a copy of the data stored in the magnetic disk device 7, and a directory 52 in which the storage position information of the data stored in the data area 51 in the magnetic disk device 7 is recorded.
It consists of and.

Although various units can be considered as a unit for copying data from the magnetic disk device 7 to the cache memory 5, in the case of the present embodiment, a case where the track in the magnetic disk device 7 is used as a unit will be described as an example.

That is, in the magnetic disk device 7, usually, a plurality of magnetic disks fixed coaxially to the rotation axis are used as a storage medium, and information is recorded concentrically on each recording surface of each magnetic disk. A plurality of tracks, which are areas, are provided, and the target track is selected by the radial displacement (seek operation) of the magnetic head provided for each recording surface. Further, in order to reduce the seek operation and improve the efficiency of access, in general, a plurality of magnetic disks in which a group of tracks equidistant from the rotation center are connected is managed by the concept of a cylinder. As a result, in the case of the present embodiment, the directory 52 has a device number indicating which of the plurality of magnetic disk devices 7 is used to identify one target track in the plurality of magnetic disk devices 7. The DEVNO, the cylinder number CYLNO indicating which cylinder of the device belongs, the head number HDNO corresponding to the track, and the storage address in the cache memory of the data corresponding to the storage position are stored in association with each other. ing.

In this case, the disk control device 4 is connected to a non-volatile memory 6 in which at least a part of the contents of the directory 52 in the cache memory 5 is saved at any time, and inside the non-volatile memory 6, a plurality of tracks are stored. A number table 61 is provided.

The individual track number table 61 is stored in the cache memory 5
Device number DEVNO and cylinder number C corresponding to the capacity of
Multiple storage location information 61 consisting of YLNO and head number HDNO
3 is recorded, and further, an identification number 612 and a valid flag 611 are recorded in each track number table 61.

Thereby, as will be described later, the contents of the directory 52 of the cache memory 5 are saved in the non-volatile memory 6 at a plurality of arbitrary times, and later the cache memory 5 is read based on the contents read from the non-volatile memory 6 at any time. It is possible to restore the state of (1) to the state of the saved point at any time.

The operation of this embodiment will be described below with reference to the flow charts of FIGS.

For example, basic software such as an operating system (not shown) resident in the main storage device 2 loads an application program or the like for performing a normal task into the main storage device 2 and activates it, so that the central processing unit 1 executes the application program. It is started (step 80).

At this time, it is checked whether or not there is restoration information of the cache memory 5 corresponding to the application program in the non-volatile memory 6 (step 81).

When the application program is first started,
Since the necessary information is not yet stored in the non-volatile memory 6, the execution of the application program is immediately started.

Then, in response to an input / output request that occurs at any time during the execution process of this application program, the central processing unit 1
The data stored in the magnetic disk device 7 is accessed through the channel device 3 and the disk control device 4 (step 82).

At this time, the disk control device 4 copies the data from the magnetic disk device 7 to the cache memory 5 by an algorithm that improves the hit rate in the cache memory 5, and copies the data to the cache memory 5 as much as possible. The operation of responding to the access request from the upper channel device 3 side is performed using the data, and while such an operation is repeated, the internal state of the cache memory 5 is adapted to the access form of the application program. The hit rate is good.

Then, when the application program ends (first time) (step 83), the disk controller 4 reads the magnetic disk of the data existing in the data area 51 of the cache memory 5 from the directory 52 of the cache memory 5 at that time. Device number DEVNO, cylinder number CYLNO, head number, which shows the storage location in device 7.
The HDNO or the like is recorded in the track number table 61 of the nonvolatile memory 6 as the storage position information 613 (step 8
Four).

FIG. 5 is a flowchart showing a more detailed example of the process of step 84.

That is, first, from the plurality of track number tables 61 provided in the non-volatile memory 6, the valid flag 611 is turned off, that is, the writable one is selected (step 841).

Next, from the directory 52, information such as a device number DEVNO, a cylinder number CYLNO, and a head number HDNO indicating the storage position in the magnetic disk device 7 of the data currently existing in the data area 51 of the cache memory 5 is stored. The position information 613 is written in the track number table 61 of the nonvolatile memory 6 (step 842).

After that, in order to prevent destruction by the same writing operation, the valid flag 611 of the track number table 61 is turned on (step 843), and writing is prohibited.
Further, the identification number 612 corresponding to the application program executed in the immediately preceding step 82 is written (step 844).

In this way, when the execution of one application program is completed, it is checked whether or not there is an execution command for another application program (step 85). If there is an execution command, the process returns to the beginning of step 80 and the application program is started. .

Such an operation is performed for each of a plurality of application programs, and the storage location information corresponding to each application program is stored in the nonvolatile memory 6.
A plurality of track number tables 61 in which 613 is recorded exist.

After that, when the same application program is started again (second time point) (step 80), it is determined whether or not there is restoration information in the nonvolatile memory 6 (step 8).
1) Since the restoration information of the cache memory 5 exists in the non-volatile memory 6 by the processing of the previous step 84, the magnetic disk device is based on the information of the track number table 61 recorded in the non-volatile memory 6. Processing for copying and restoring predetermined data from 7 to the cache memory 5 is started (step 86).

The process of step 86 is performed by the upper central processing unit 1.
The process is executed independently of the application program executed in the step 1. and asynchronously, and the execution of the application program in the central processing unit 11 or the like is not kept waiting during the process of step 86.

FIG. 6 shows an example of the process of step 86 in more detail.

That is, first, the identification numbers of the plurality of track number tables 61 recorded in the non-volatile memory 6 are checked, and the track number table 61 corresponding to the application program is selected (step 861).

Next, the storage location information 613 recorded in the track number table 61 is used to restore the directory 52 of the cache memory 5 (step 862).

In addition, the device number DEVN of the restored directory 52
Using O, cylinder number CYLNO, head number HDNO, etc., the data of the magnetic disk device 7 is copied to the data area 51 of the cache memory 5 (step 863) and repeated until all the data is copied (step 864).

After that, the valid lag of the track number table 61 is turned off so that the track number table 61 can be written in the subsequent step 84 (step 86).
Five).

Thus, when the application program is started, the internal state of the cache memory 5 becomes a state suitable for the access form to the magnetic disk device 7 by the started application program, and the subsequent step 82
From the beginning of the execution of the application program in (1), it is possible to access the data of the group of magnetic disk devices 7 with a high hit rate of the cache memory 5.

As a result, for example, when switching execution of a plurality of types of application programs having different access modes to the magnetic disk device 7, the cache memory 5 as in the conventional case is switched.
Until it reaches a stable state, the hit rate of the cache memory 5 is surely prevented from being greatly reduced, and the frequency of inefficient direct access to the magnetic disk device 7 does not increase, The data transfer amount per unit time between the main storage device 2 and the magnetic disk device 7 can be stably maintained at a good value.

In addition, the effect becomes greater especially in the case of a large capacity cache memory 5 which takes a long time to reach a good hit rate.

[Second Embodiment] Next, a disk cache control method according to another embodiment of the present invention will be described.

In the following description, the magnetic disk device 7 will be used as an example.
The case of copying the data from the cache memory 5 to the cache memory 5 in units of tracks will be described.

In the case of the second embodiment, as shown in FIG. 7, the directory 52 of the cache memory 5 has an in-cache slot management table 521 and a cache hit ratio table 522.

The in-cache slot management table 521 improves the information indicating the storage position in the magnetic disk device 7 of the data existing in track units in the data area 51 of the cache memory 5, the storage address in the cache memory 5, and the hit rate. Various information for realizing the algorithm for is stored.

In the cache hit rate table 522, a relative frequency (hit rate) in which a plurality of track units of data are accessed from the higher channel device 3 inside the cache memory 5 is recorded.

Then, when the predetermined application program ends and the information in the directory 52 is saved in the non-volatile memory 6, the processing in step 84 is performed in step 84 in FIG.
As shown in 2a, by referring to the cache hit rate table 522, information such as device number DEVNO, cylinder number CYLNO, head number HDNO corresponding to the relevant track is stored by the capacity of the cache memory 5 in descending order of hit rate. The storage position information 613 is written and recorded in the track number table 61 of the nonvolatile memory 6.

After that, as in the case of the first embodiment, when the same application program is re-executed, the magnetic disk device 7 to the cache memory 5 are stored based on the storage position information 613 recorded in the corresponding track number table 61.
Data is copied to.

As a result, in the case of the second embodiment, an access mode in which simply restoring the state of the cache memory 5 at the end of an application program at the time of re-execution does not necessarily contribute to the improvement of the hit rate. In this case as well, a high hit rate of the cache memory 5 can be realized from the beginning of execution of the application program.

[Third Embodiment] In the case of the present embodiment, the directory 52 of the cache memory 5 is saved to the non-volatile memory 6, and the state of the cache memory 5 is restored based on the storage location information 613 saved in the non-volatile memory 6. The channel command word CCW issued from the operating system or the application program at any time is used as the trigger.

That is, the channel command word CCW for issuing the operation of registering the information in the directory 52 of the cache memory 5 in the non-volatile memory 6 is issued at the time when the application program ends or at any time during execution, and A channel command word CCW for issuing an operation to restore the state of the cache memory 5 based on the information saved in the non-volatile memory 6 is issued at the start of execution of an application program or at any point during execution. Programming.

Then, a channel command word for registering the information in the directory 52 of the cache memory 5 in the nonvolatile memory 6
At the time when the CCW is issued, the disk controller 4 uses the method shown in the flow chart shown in FIG. 5 of the first embodiment or the flow chart shown in FIG. 8 of the second embodiment at that time. The state of the cache memory 5 is saved in the track number table 61 of the non-volatile memory 6.

At this time, the identification number 612 is received at the same time and is written in the corresponding track number table 61.

Further, when the same application program is executed again later, the channel command word CCW for restoring the cache memory 5 is restored from the upper channel device 3 to the application program at an initial stage of execution or at any time during execution. The disk control device 4 which issued the corresponding identification number 612 and received it, based on the storage position information 613 recorded in the track number table 61 of the received identification number 612, receives a cache memory from the magnetic disk device 7. 5 is copied in track units, for example, and the operation of restoring the state of the cache memory 5 to the state suitable for the access form of the application program is performed.

An external storage subsystem including the disk control device 4, the cache memory 5, the magnetic disk device 7, etc. is provided by giving a trigger for saving and restoring the state of the cache memory 5 at any time from a program as in the third embodiment. It is possible to realize more detailed and efficient operation of.

Although the invention made by the present inventor has been specifically described based on the embodiments, the present invention is not limited to the embodiments and various modifications can be made without departing from the scope of the invention. Nor.

For example, in the description of the above embodiments, the case where the cache memory and the non-volatile memory are installed at the level of the disk control device has been described, but the present invention is not limited to this, and they are installed in the lower magnetic disk device or the upper channel device. However, it goes without saying that the effects of the present invention are not impaired.

Further, the unit of data copied from the magnetic disk device to the cache memory is not limited to a track unit, but a sector unit, which is a concept of equally dividing a track in the circumferential direction, and a cylinder unit composed of a plurality of tracks,
Alternatively, it may be an arbitrary data set unit or the like.

〔The invention's effect〕

The effects obtained by the typical ones of the inventions disclosed in the present application will be briefly described as follows.

That is, according to the disk cache control method of the present invention, a high-order processing device, a magnetic disk device in which data sent and received between the high-order processing device is stored, the magnetic disk device, and the high-order processing device Intervenes between
A cache memory that holds a copy of the data stored in the magnetic disk device, and is stored in the cache memory as much as possible in response to an access request from the host processor to the magnetic disk device. In the information processing system for responding using the data, the magnetic disk device of the data stored in the cache memory during execution or at the end of execution of an arbitrary program executed in the upper processing device. The storage position information, which is the first information group indicating the storage position in, is the second information consisting of the identification number for identifying the program for each of the plurality of first time points by the channel command word from the host processor. Is stored in a non-volatile memory in a state where
At a second time point different from the second time point by the channel command word from the host processor.
Based on the information, the first information group recorded and stored in the non-volatile memory at the first time point is selected and read out, and the magnetic disk device causes the cache memory to read out based on the first information group. Since the operation of copying the above-mentioned data to the above is performed, for example, at the first point of time during execution or termination of each application program triggered by a command from the host processor or setting of a hardware switch from the outside. In, the first information group indicating the storage position of the data existing in the cache memory in the magnetic disk device is recorded and saved in the non-volatile memory together with the identification number corresponding to the application program, and the arbitrary application program is restarted later. At the second point in time of operation or after startup, The first information group is read from the non-volatile memory based on the identification number corresponding to the program, the data is copied from the magnetic disk device to the cache memory by using the first information group, and the state of the cache memory is determined by the application. It becomes possible to restore the optimum state for the program, and it is possible to surely prevent the hit rate of the cache memory from being lowered when the plurality of application programs are switched.

[Brief description of drawings]

FIG. 1 is a block diagram showing an essential part of an information processing system in which a disk cache control method according to an embodiment of the present invention is implemented, and FIG. 2 is a block diagram showing an example of its overall configuration. Is a diagram showing an example of a storage format of information, FIG. 4 is a flow chart showing an example of operation of a disk cache control method according to an embodiment of the present invention, and FIG. 5 is a flow chart showing a part thereof in more detail, FIG. 6 is a flow chart for explaining a part thereof in more detail, FIG. 7 is an explanatory view showing an example of the structure of a directory in the disk cache control method according to the second embodiment of the present invention, and FIG. It is a flowchart which shows an example. 1 ... Central processing unit (upper processing unit), 2 ... Main storage unit, 3 ... Channel unit, 4 ... Disk control unit, 5
... cache memory, 51 ... data area, 52 ... directory, 521 ... cache slot management table, 522 ... cache hit rate table (counting means), 6 ...
… Nonvolatile memory, 61 …… Track number table, 611
...... Valid flag, 612 ... Identification number (second information), 613
...... Storage location information (first information group), CYLNO ... Cylinder number, DEVNO ... Device number, HDNO ... Head number, 7 ... Magnetic disk device, 80,81,82,83,84,85, 86 ...
... steps showing an example of a disk cache control method,
841,842,842a, 843,844 ... An example of a series of steps constituting the process of step 84, 861,862,863,864,865 ... An example of a series of steps constituting the process of step 86, CCW ...
… Channel command word.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Mitsuo Nagasawa 2880, Kozu, Odawara-shi, Kanagawa Hitachi Ltd. Odawara Plant (72) Inventor Morihiko Yotsuya 2880, Kozu, Odawara, Kanagawa Hitachi Odawara Plant (56) Reference JP 62-130440 (JP, A) JP 1-131940 (JP, A)

Claims (1)

[Claims] 1. A high-order processing device, and a magnetic disk device for storing data exchanged with the high-order processing device,
An access between the magnetic disk device and the host processing device, the cache memory holding a copy of the data stored in the magnetic disk device, and accessing the magnetic disk device from the host processing device In an information processing system that responds to a request by using the data held in the cache memory as much as possible, during execution or at the end of execution of an arbitrary program executed in the upper processing device Storage position information, which is a first information group indicating a storage position of the data stored in the cache memory in the magnetic disk device, is stored at each of a plurality of first time points according to a channel command word from the upper processing device. With the second information consisting of the identification number for identifying the program for each The data is recorded and stored in a volatile memory, and at a second time point different from the first time point, the channel command word from the higher-level processing device is used to generate the channel information at the first time point based on the second information. A disk cache characterized by selecting and reading the first information group recorded and stored in a non-volatile memory, and copying the data from the magnetic disk device to the cache memory based on the first information group. Control method.