JP2875325B2 - Method of controlling storage device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a storage device control method for distributing a load to be executed among storage devices.

[Conventional technology]

Japanese Patent Application Laid-Open No. 60-114947 discloses a disk cache (hereinafter, referred to as a disk cache).
Sometimes simply called cache. ) Discloses a technique relating to a double writing function of a disk device in a control device having the following. Here, the two disk devices for writing the same data are referred to as dual-write disks.

Japanese Patent Application Laid-Open No. 60-114947 describes a method of controlling a disk device with double (two) write data using a disk cache as follows. The control device is
An input / output request from the CPU is processed for one of the two disk devices. When receiving a read request (input request), the control device executes the received request as it is. When a write request (output request) is received from the CPU, the following processing is executed. That is, data is written to one of the dual writing disks, and this data is also written to the disk cache. Then, the control device later writes the data written in the disk cache to the other disk device using the vacant time.
This writing process is called a write after process. As described above, the same data is written to the double writing disk device.

Japanese Patent Publication No. 61-28128 discloses a technique relating to load distribution of a dual-write disk device as a method of controlling a duplicate file. Japanese Patent Publication No. 61-28128 aims to speed up the processing by selecting an empty disk device when processing an input / output request. However, JP-B 61-281
No. 28 does not use a method of double-writing data by write-after control using a disk cache as disclosed in Japanese Patent Application Laid-Open No. 60-114947.

Nishigaki et al .: Effect analysis of disk cache in sequential access input processing, Transactions of Information Processing Society of Japan, Vol.25, No.2, p.
The following technology is disclosed in the article on p.313-320 (1984). This is a technique in which a read-ahead process for a sequential process in a control device having a disk cache, that is, data that is not requested from the CPU, is staged in a cache. This stage processing is based on the CP
Responding to an input / output request from U is independently executed by the control device.

[Problems to be solved by the invention]

Japanese Patent Application Laid-Open No. 60-114947 does not pay attention to the advantage that there are a plurality of disk devices that can be processed by the control device in a double-write disk. That is, CP
The disk device to be selected in response to the input / output request received from U is limited to a specific one.

On the other hand, as disclosed in Japanese Patent Publication No. 61-28128, a method of selecting an empty disk device in response to an input / output request is excellent in performance. However, this method can be implemented by a write-after control using a cache.
If applied to the case where the overwriting function is realized, the reliability is deteriorated. This is because it is highly probable that there is post-write data received from CPUs corresponding to all disk devices. Therefore, if the power down of the cache and the failure of one of the disk devices overlap, the write data received from the CPU will be lost.

Further, in the case of a control device having a cache, the control device executes input / output processing between the cache and the disk device independently of an input / output request received from the CPU.
(This is Nishigaki et al .: Effect analysis of disk cache in sequential access input processing, Transactions of Information Processing Society of Japan, Vol.2
5, No. 2, pp. 313-320 (1984). Therefore, it should be noted that there are a plurality of disk devices that can be selected as targets to which the processing is to be assigned, even for input / output processing executed independently of the CPU by the control apparatus. It is.

The present invention relates to a control method and a control device for writing the same data to a disk device group including one or more disk devices by write-after processing using a cache.

An object of the present invention is to provide a control device and a control method for distributing the load of input processing to be executed between the disk devices in a disk device group and improving the parallelism of execution of input / output processing. .

[Means for solving the problem]

In order to make the description of the method and the device for solving the problem easy to understand, the input / output processing executed by the control device with the disk device is classified.

Input / output processing executed by the control device with the disk device can be classified into the following four types.

(1) Processing for a write request received from the CPU, which requires access to the disk device.

(2) Processing for a read request received from the CPU, which requires access to the disk device.

(3) Stage processing for transferring data from the disk device to the cache, independent of input / output requests (read requests / write requests) from the CPU.

(4) Write-after processing executed with the disk device.

Of the above input / output processing, the write-after processing is not a target of load distribution. Hereinafter, the reason will be described. The write-after process is a process executed when executing a write request received from the CPU for all disk devices in a disk device group other than the disk device to which data has been written. Therefore, there is no flexibility in selecting a disk device on which the write-after processing is to be performed.
Therefore, of the above processes (1) to (4),
The processes (1) to (3) are targets of load distribution.

In this specification, two methods are shown as a method of executing load distribution. For convenience, these methods are
1, referred to as load distribution method 2. Hereinafter, each method will be described.

Load distribution method 1... When selecting a disk device to execute the input / output processing of (2) or (3) above, the control device selects an arbitrary empty disk device in the disk device group. When selecting a disk device in response to the write request received from the CPU which needs to access the disk device in the above (1), a specific disk device in the disk device group is selected.

Load distribution method 2... When selecting a disk device in response to the above-described processing category (1), that is, a write request received from a CPU that needs to access the disk device, Select a specific disk device. When selecting a disk device to be subjected to the input / output processing shown in the processing classification (2) or (3), a disk device other than the specific disk device is preferentially selected.

[Action]

The operation corresponding to each of the load distribution method 1 and the load distribution method 2 will be described.

(1) In the case of the load distribution method 1, the control device sends a read request for accessing a disk device to a certain disk device group to the CPU.
, The following processing is executed. In response to the read request, any one free disk device in the disk device group is selected. If there is no empty disk device, the control device gives the read request. When a write request that needs to access a disk device is received, one specific disk device in the disk device group is selected in response to the write request. However, if the particular disk device is not empty, the control device waits for the write request.

Also, when the control device attempts to execute a stage process independent of an input / output request from the CPU, the following process is performed. That is, one free disk device in the disk device group corresponding to this stage processing is selected. If there is no empty disk device, the control device does not execute the stage processing.

The reliability and performance characteristics of the load distribution method 1 are described below.

The load distribution method 1 is slightly less effective than the method disclosed in JP-A-60-114947.
Excellent performance can be obtained as compared to the method disclosed in 61-28128.

Hereinafter, the reason will be described. The load distribution method 1 imposes a limit on the degree of freedom in selecting a disk device for a write request. Therefore, the effect of dispersion is inferior to the method disclosed in JP-A-60-114947. However, an empty disk device is selected in response to a read request. Normally, the ratio of input / output requests to the disk device is considerably higher in read requests than in write requests. Accordingly, the load distribution method 1 does not cause much performance degradation as compared with the method disclosed in JP-A-60-114947. On the other hand, compared with the method of JP-B-61-28128 in which all input / output requests are concentrated on one disk device, the load distribution method 1 can obtain excellent performance.

The reliability of the load distribution method 1 is higher than that of the method disclosed in Japanese Patent Application Laid-Open No. 60-114947, and is substantially the same as that of the method of Japanese Patent Publication No. 61-28128.

Hereinafter, the reason will be described. Load balancing method 1, or
In Japanese Patent Publication No. 61-28128, there is no data to be subjected to write-after processing for a disk device that concentrates write requests.
Therefore, even if the power supply of the cache goes down, the write data received from the CPU does not disappear unless a failure occurs in the disk device that concentrates the write requests.

As described above, according to the load distribution method 1, it is possible to perform load distribution that achieves high performance / high reliability of the disk device group in a well-balanced manner.

(2) In the case of the load distribution method 2 When the control device receives a write request from the CPU which needs to access a disk device of the disk device group, the control device specifies the disk device group in response to the write request. Is selected. However, if this particular disk device is not empty, the control device waits for the write request. When a read request for accessing a disk device of a certain disk device group is received from the CPU, the following process is executed. First, one of the disk devices other than the above-mentioned specific disk device is selected as an empty disk device. When there is no empty disk device, it is checked whether or not the specific disk device is empty. If so, the controller selects this particular disk device for the received read request. If not, the control device waits for the read request.

Also, when the control device attempts to execute a stage process independent of an input / output request from the CPU, the following process is performed. First, an arbitrary one of the disk devices other than the above-mentioned specific disk device is selected. If there is no empty disk device, it is checked whether the specific disk device is empty. If so, the controller selects this particular disk device for the stage processing that was to be performed. If not, the control device does not execute the stage processing.

The reasons for using the above-described dispersion method are as follows. For example, assume that a process corresponding to a read request is assigned to a specific disk device that concentrates a process corresponding to a write request from the CPU. In this case, if a write request is received before the processing corresponding to the read request is completed,
It will not be possible to enter the processing corresponding to the write request. Therefore, a CPU that does not need to select a specific disk device
For processes other than the process corresponding to the write request from, a higher priority is given to assigning a disk device other than the specific disk device to increase the effect of dispersion.

〔Example〕

Hereinafter, two types of embodiments of the present invention will be described. First, contents common to both embodiments will be described.

FIG. 2 shows the configuration of a computer system to which the present invention is applied. The computer system includes a processing device 210 including a CPU 200, a main memory 201, and a channel 202, a control device 203,
It comprises one or more n disk devices 204.
It will be apparent from the following description that the present invention can be applied to a case where a plurality of processing devices 210 are connected to the control device 203.

The n disk devices 204 are grouped into m disk device groups 211 each including one or more disk devices 204. The number of disk devices 204 belonging to each disk device group 211 may be different. Each disk device 204 belongs to one disk device group 211. Each disk device 2
The method for determining which disk device group 211 the 04 belongs to is not directly related to the present invention, and thus the description is omitted.

The control device 203 includes one or more directors 205, a cache (memory) 206, a control information memory 207, and a directory 208. Each director 205 is a channel
202 and the disk device 204, between the channel 202 and the cache 206, and between the cache 206 and the disk device.
Transfer data to / from 204. In the cache 206, data frequently accessed among the data stored in the disk device 204 is staged. Directory 20
8 stores management information of the cache 206. The stage processing is executed by the director 205. Specific examples of the stage data include data to be accessed by the CPU 20 and data whose storage position of the disk device 204 is close to the data.

The control device 203 according to the present invention has a function of writing the same data to the disk devices 204 belonging to a certain disk device group 211, that is, a so-called multiple writing function. Therefore, it may be considered that the processing device 210 issues an input / output request to each disk device group 211.

In the present invention, when viewed from the control device 203, the input / output requests received from the processing device 204 can be classified as follows.

(1) A request for data transfer between the cache 206 and the processing device 210, which is an input / output processing pattern that does not access the disk device group 211. This is, for example,
This is a processing pattern executed when data corresponding to a read request received from the processing device 210 is staged in the cache 206.

(2) This is an input / output processing pattern that requires access to the disk device 204 in the disk device group 211.

(3) Further, when the control device 203 has the cache 206, the control device 203 executes the following input / output processing independently of the input / output request received from the processing device 210.

This is an input / output processing pattern between the cache 206 and the disk devices 204 in the disk device group 211, and is a data transfer pattern that does not involve the processing device 210.

The present invention relates to a load distribution method between disk devices 204 belonging to the same disk device group 211. Therefore, the processing pattern that does not need to access the disk device group 211 shown in (1) is not directly related to the present invention. Of the input / output processing patterns executed by the control device 203, the input / output processing patterns shown in (2) and (3) are the objects of the present invention. In addition,
Disk device 204 not assigned to the input / output processing shown in (2) or (3) (processing is not being executed)
Is called an empty disk device.

The outline of each of the two embodiments will be described below. First, the outline of the first embodiment will be described.

FIG. 1 is a diagram for explaining the operation of the control device 203 in the first embodiment.

In the configuration of FIG. 1, one mask disk A, B, and C exists in each disk device group A, B, and C. The difference between the master disk and other disk devices will be described later. What is a master disk?
This is a specific disk device determined in advance in each disk device group.

In FIG. 1, a disk device group A, a disk device group B and a disk device group C are connected to the control device 203.

The disk device group A is composed of master disks A, disks A1,..., Disks Ai. Similarly, the disk device group B is composed of master disks B,..., Disk Bj, and the disk device group C is composed of master disks C,.

The input / output requests that the control device 203 receives from the processing device 210 and needs to access the disk device group 211 will be described separately for write requests and read requests. In FIG. 1, reference numeral 110 denotes a data flow associated with a write request, and reference numeral 113 denotes a data flow associated with a read request. Hereinafter, a data transfer pattern for each request will be described.

The control device 203 has received a write request 110 that needs to access the disk device group A. The control device 203 selects the master disk A from the disk device group A. That is, the master disk A is a disk device that concentrates write requests that need to access the disk device group A.

The control device 203 writes data accompanying the write request received from the processing device 210 to the master disk A,
Write it to cache 206 as well. The control device 203 executes the same data write processing on the disk devices belonging to the same disk device group, for example, the master disk A, the disk devices A1,..., The disk device Ai. The write data 111 written in the cache 206 is used to execute the above-described write processing. The details will be described later.

When a write request for accessing the disk device group A is received, the reason for selecting the master disk A is as follows. If a write request that needs to access the disk device group A is always assigned to the master disk A, the master disk A
, The entire write data 112 received from the processing device 210 is written. For this reason, for example, a disk Ai other than the master disk A fails and the cache 20
When the power of 6 is turned off, complete data is held in the master disk A.

However, since the degree of freedom in selecting a disk device is limited to the assignment of the write request received from the processing device 210, the performance is reduced. Specifically, the control device 203
However, when a write request 110 that needs to access the disk device group A is received, even if a disk A1 other than the master disk A is vacant, if the master disk A is not vacant, the process is immediately started. You can't.

FIG. 1 also shows a case where the control device 203 receives a read request that needs to access the disk device processing group C. At this time, the control device 203 selects one disk device (disk C1 in FIG. 1) from any empty disk devices in the disk device group C.

The control device 203 sends the requested data from the disk C1 to the processing device 210. At this time, the data requested by the processing device 210 (or the data requested by the processing device 210 and the data near the data C1 on the disk C1) is set as the stage data 114, and the cache is cached. You may stage at 206. This is indicated by a broken line.

The input / output processing between the disk device group and the cache 206 executed by the control device 203 independently of the input / output request received from the processing device 210 includes the following processes shown in FIG. Write-after processing for writing write data to the disk device is a stage process independent of input / output requests from the processing device 210. An example of a stage process independent of the processing device 210 corresponds to a case where the control device 203 executes a prefetch process for a sequential process. Less than,
Each case will be described.

The write-after process is a process of writing the write data 111 in the cache 206 to the unwritten disk device Ai. In the disk device group A, it is not necessary to execute the write-after process on the master disk A to which the write data 111 received from the processing device 210 is directly written. Therefore, the disks A1,..., Disks Ai, except for the master disk A, are sequentially subjected to the write-after processing. The order of execution may be random.

When executing a stage process on a certain disk device group B independent of the processing device 210, the control device 203 selects any one of the disk device groups B that is vacant in the disk device group B. . In FIG. 1, the control device 203 has staged the stage data 116 on the cache 206 from the disk Bj.

The above is the operation of the control device 203 in the present embodiment. The features are as follows.

For a write request that needs to access the disk device group received from the processing device 210, a master disk is selected to ensure reliability. On the other hand, when reading data from a disk device group, an empty disk device is selected. As described above, high reliability and high performance are realized in a well-balanced manner.

Hereinafter, an example of the parallel operation will be described with reference to FIGS.

FIG. 3 shows a form in which the control device 203 executes the following pattern of input / output processing in parallel.

The input / output process of the first pattern is an input / output process that needs to access the disk device group A received from the processing device 210.

The input / output processing of the second pattern is an input / output processing executed by the control device 203 with the cache 206 independently of the processing device 210.

For example, as shown in FIG. 3, the control device 203 executes a stage process independent of the input / output process from the processing device 210 with the disk A1, and a write-after process with the disk A2. And At this time, the control device 203
However, it is assumed that an input / output request that requires access to the disk device group A from the processing device 210 is received. FIG. 3 shows an example in which a read request that requires access to the disk device group A is received. In this case, the control device 203
Can select the empty disk device Ai from the disk device group A to execute the received input / output request. However, if the input / output request received from the processing device 210 is a write request that requires access to the disk device group A, the same processing cannot be executed unless the master disk A is free.

Further, in FIG. 3, each of the write after process and the stage process independent of the processing device 210 are executed one by one. However, if there is an empty disk device, the control device 203 can execute two or more multiplex write-after processes and a stage process independent of the processing device 210.

4 and 5 show the contents related to the parallel operation of the human output processing received from the processing device 210.

FIG. 4 shows a case where a plurality of processing devices 210 are connected to the control device 203, specifically, the processing devices 210 and 210a are connected.

It is assumed that the control device 203 is executing a process of a read request that needs to access the disk device group A received from the processing device 210 with the disk A1. At this time, it is assumed that the control device 203 has received a read request that requires access to the disk device group A from the processing device 210a. The control device 203 selects an optional disk device in the disk device group A, that is, the disk Ai, and starts processing the received read request. Of course, even when a write request that needs to access the disk device group A is received from the processing device 210a, if there is an empty disk device, the processing can be immediately started. However,
Due to contention of the master disk A, a parallel operation of write requests that need to access the disk device group A cannot be executed.

Further, when three or more processing devices are connected to the control device 203, if there is an empty disk device, an I / O request that needs to access three or more disk device groups A is required. Can be executed in parallel.

FIG. 5 shows a case where one processing device 210 is connected to the control device 203. It is assumed that the processing device 210 has a data function for issuing a new input / output request to the disk device group A before the processing for the input / output request to the disk device group A is completed.

In FIG. 5, it is assumed that the control device 203 is executing a read request process which needs to access the disk device group A received from the processing device 210 with the disk A1. In this state, the control device 203
, Further receives a read request that needs to access the disk device group A. Control device 20
3 selects an arbitrary disk device in the disk device group A which is in an empty state, that is, a disk Ai,
The processing of the received read request is started. Of course, even if a write request that needs to access the disk device group A is received from the processing device 210, if there is an empty disk device, the processing can be immediately started. However, due to the competition of Master Disk A,
A parallel operation between write requests that need to access the disk device group A cannot be executed.

Further, when the processing device 210 issues an I / O request for three or more multiplexes to the same disk device group A, if there is only an empty disk device, the control device 203 executes the issued I / O request. Requests can be executed in parallel.

The above is the outline of the first embodiment. Next, an outline of the second embodiment will be described.

 An outline of the second embodiment is shown in FIG.

The operation of the control device 203 of the second embodiment differs from that of the first embodiment in the following points. For a read request that needs to access the disk device group, a disk device other than the master disk is preferentially selected. Similarly, for the stage processing independent of the processing device 210, a disk device other than the master disk is preferentially selected. The reason why the above-described method is adopted is that if a disk device other than the master disk is selected, the speed can be further increased. This is because the probability that the master disk is free when a write request that needs to access the disk device group is received can be increased.

 Hereinafter, details of the embodiment will be described.

 First, the first embodiment will be described in detail.

The configuration of the computer system shown in FIG. 2 can be applied to the first embodiment as it is. Hereinafter, each of them will be described in detail.

FIG. 8 shows the configuration of the disk device 204. Disk 801
Is a medium for recording data.
Reference numeral 204 denotes a plurality of rotating bodies. Read / write head 802
Is a device for reading and writing data on the disk 801, and exists for the disk 308. Control device interface 803
Is an interface with the control device 203.

A circular recording unit accessible by the read / write head 802 during one rotation of the disk 801 is called a track 800. A plurality of tracks 800 exist on the disk 801.

FIG. 9 shows the configuration of the track 800. Truck 800
, A track head 902 and a track tail 903 are determined based on a certain position. Also, on Track 800,
One or more records 900 exist. The record 900 is a minimum input / output processing unit between the processing device 210 and the control device 203. The location of record 900 on track 800 is cell 901
Is expressed in units of fixed-length bytes. (Record 90
0 always starts storing from the beginning of cell 901 and cell 901
Is not started from the middle of the process. ) The number of the cell 901 is assigned in ascending order by 1 with the head of the track 800 being 0.

FIG. 10 shows the structure of the cache 206. Cache 2
06 is composed of a segment 1000. In the present embodiment, one segment is assigned to one track 800, and data of the entire track 800 is stored in the segment 1000. However, in the present invention, the allocation unit of the segment 1000 does not need to be limited to the track 800, and is effective as a small unit, for example, the record 900 which is a read / write unit of the processing device 210 and the control device 203. .

FIG. 11 shows the configuration of the directory 208. The directory 208 includes segment management information 1100, track votes 1101,
And an empty segment head pointer 1102. The segment management information 1100 exists in units of 1000 segments. One track slip 1101 and one free segment pointer 1102 exist in the control device 203.

FIG. 12 shows information required in this embodiment in the segment management information 1100. Hereinafter, each parameter and its contents are shown.

Empty segment pointer 1200: a pointer to segment management information 1100 corresponding to another segment 1000 not allocated to the track 800.

Cache track number 1201... Represents the number of the disk device group 211 and the number of the track 800 stored in the segment 400 corresponding to the segment management information 500.

Record bit map 1202 ... Segment management information
Track 800 stored in segment 400 corresponding to 500
Represents the starting position of the record 900 above. Here, it is assumed that each bit exists corresponding to the number of the cell 901. For example, if the n-th bit in the record bit map 1202 is on, it means that the storage of the record 301 has been started from the n-th cell 901 corresponding to the segment management information 1100. If the n-th bit is off, it means that there is no record 301 starting from the n-th cell 901 in the record 301 whose storage has been started. FIG. 13 shows the storage format on the track 800 in the segment 1000 in this embodiment. In the segment 1000, records 900 are stored in order from the track head 901 on the track 800. Therefore, if the number of the cell 901 to be stored on the track of the record 900 is known, the storage start position in the segment 1000 of the record 301 can be known.

Update record bitmap 1203 ... stored in the segment 400 corresponding to the segment management information 500,
In addition, this is a bit map of the record 900 requiring the write-after processing 112. The record 900 requiring the write process 112 is hereinafter referred to as a write after record. Each bit exists in correspondence with the number of the cell 901 as in the case of the record bit map 1202. Specifically, if the n-th bit in the update record bit map 1203 is ON, the record 301 starting to be stored from the n-th cell 901 corresponding to the segment management information 1100 is called a write-after record. Will be. Update Record Bitmap 1203
Exist for one disk device 204. Specifically, which disk device 204 the update bit map 1203 corresponds to will be described in the description of the configuration of the control memory 207. Update Record Bitmap 1203
Are prepared by the number of disk devices 204 that can be defined in one disk device group 211. However, the number of update record bitmaps 1203 used depends on the number of disk devices that constitute the disk device group.
This is 204 units.

Stored flag 1204: Indicates whether the record 900 on the allocated track 800 is stored in the segment 1000 corresponding to the segment management information.

Busy flag 1205: Indicates that the input / output processing corresponding to the track 800 assigned to the segment management information 1100 is being executed.

A segment pointer 1206 is a pointer to the segment 1000 corresponding to the segment management information.

FIG. 14 shows the configuration of the track slip 1101 and the free segment head pointer 1102.

Track vote 1101 is assigned to all disk device groups 21
Regarding one track 800, it indicates whether the segment 1000 is allocated to the track 800 or not. If it is assigned, it indicates a pointer to the segment management information 1200 corresponding to the segment 1000 assigned to the track 800. In the track vote 1201, the track 8 on the same disk device group 211
Information on 00 is collectively stored in the order of the track 800 number.

The segment management information 1100 corresponding to the segment 1000 to which the track 800 is not allocated is, in order from the free segment head pointer 1102, the free segment pointer 1200,
Be combined. A set of segment management information 1100 that is combined is called an empty segment queue 1400.

FIG. 15 shows the configuration of the control information memory 207. In the control information memory 207, disk device group information 1500 which is control information corresponding to each disk device group 211 is stored.
Is included. The number of disk device group information 1500 is
The number of disk device groups 211 that can be defined in one control device 203 is prepared.

FIG. 16 shows the configuration of a disk device group device 1500.

The number of disk devices 1600 ... the relevant disk device group 211
Is the number of disk devices 204 currently defined within.

Disk packaging information 1601... Information corresponding to each disk device 204 constituting the disk device group information 1500. As the disk device information 1600, only the disk devices 204 that can be defined in one disk device group 211 are prepared. However, valid information is stored from the first disk device information 1601 to the number of disk device information 1601 defined in the number 1600 of disk devices. Here, the first disk device information 1601 is information corresponding to the master disk.

Also, the n-th update record bitmap 1203 in the segment management information 1200 contains the n-th disk device information 16.
This is a disk device 204 corresponding to 01.

Processing device I / O wait bit 1602: A bit indicating that an input / output request received from the processing device 210 to the disk device group 211 is in a wait state. The number of bits can be expressed as follows: number of processing device I / O waiting bits 1602 = number of processing devices 210 connectable to control device 203 (here, I units) x
According to the number of input / output processing requests that can be processed in parallel by one processing device 210 with respect to one disk device group 210 (here, J requests), each processing device 210 controls the control device 203. When issuing an input / output request to the control device 210, the processing device 210 notifies the control device 203 of the following two points. First, the first point is that the number of the processing device 210 from the first to the I-th is the processing device 210 that issues the input / output request. The second point is that the I / O request is assigned to the specified disk device group.
The number of the input / output request from No. 1 to No. J to the 211 is notified.

 FIG. 17 shows the structure of the disk device information 1601.

Disk device number 1700: The number of the disk device 204 corresponding to the disk device information 1700.

Processing device I / O executing bit 1701... Indicates by one bit that the disk device 204 corresponding to the disk device information 1700 is executing an input / output request from the processing device 210.

Write-after executing bit 1702... Indicates by one bit that the disk device 204 corresponding to the disk device information 1700 is executing the write-after processing 112.

Independent stage execution bit 1703... Indicates by one bit that the disk device 204 corresponding to the corresponding disk device information 1700 is performing the stage processing 115 independent of the processing device 210.

Only one piece of information that is turned on at the same time among the processing device I / O executing bit 1701, the write data executing bit 1702, and the independent stage executing bit 1703 is at the same time. Also, the processing device I /
The disk device 204 in which all of the O-executing bit 1701, the write-fat-executing bit 1702, and the independent-stage-executing bit 1703 are off is the disk device 204 in an empty state.

Segment management information pointer 1704... Indicates a pointer to the segment management embodiment 1100 assigned to the track 800 accessed by the input / output processing being executed by the disk device 204 corresponding to the disk device information 1700.

There is a problem if the information in the control information memory 207 is lost due to a power failure or the like.
Is desirably nonvolatile.

The input / output processing to be executed by the control device 203 is actually
Each director 205 in the control device 203 executes in parallel.

FIG. 18 shows each procedure used when each director 205 executes this embodiment. Hereinafter, each function will be described.

Input / output request receiving unit 1800: Processes input / output requests received from the processing device 210.

Write-after processing schedule section 1801 ... Schedules the write-after processing.

Independent stage processing schedule section 1802 ... Processing device 210
Schedule the stage processing independent of.

Disk device transfer section 1803: A portion for executing read / write transfer with the disk device 204.

FIG. 19 is a processing flowchart of the input / output request receiving unit 1800. The execution timing of the input / output request receiving unit 1800 is when a new input / output request is received from the processing device 210. Hereinafter, the processing flow will be described.

In step 1900, it is checked whether the received I / O request is an I / O request that needs to access the disk device 204. The specific input / output request that needs to access the disk device 204 is not directly related to the present invention, and thus the details are omitted. If the received input / output request does not require access to the disk device 204, the process jumps to step 1915.

The processing after step 1901 is a processing in the case where the received input / output request is an input / output request that needs to access the disk device 204.

In step 1901, it is checked whether the track 300 to be accessed by the input / output request is allocating the segment 1000. If so, the process jumps to step 1903.

In step 1902, the segment management information 1100 is allocated to the track 300 to be accessed by the input / output request, and is linked to the corresponding area of the track slip 1101. Further, the stored flag 1205 of the allocated segment management information 1100 is turned off, and the busy flag is turned on. At this time, the segment management information 1100 to be allocated is the segment management information 1100 that is free from the free segment head queue 1102.
Select Empty segment management information 1100
If there is no segment management information, the segment management information currently assigned is selected by a known method. The specific selection method is not related to the present invention, and thus the description is omitted. Thereafter, jump to step 1905.

In step 1903, it is checked whether the busy flag 1205 of the segment management information 1100 being allocated to the track 300 to be accessed by the input / output request is on. If on, tracks targeted for access by other I / O processes
300 is currently in use. Therefore, the received input / output request is jumped to step 1914 because it cannot be immediately executed.

If the busy flag 1205 is off, in step 1904,
Turn on the busy flag 1205.

Next, in step 1905, it is checked whether the input / output request is a read request or a write request. In the present invention, a write request that needs to access the disk device group 211 causes the master disk to be accessed. Therefore, if the input / output request is a read request, the flow branches to step 1908.

In the case of a write request, it is checked in step 1906 whether the master disk is in an empty state. This check is for the disk device group 211
Device information 1601 corresponding to the master disk in the
The following information in the first disk device information 1601 in the disk device group information 1500 is checked. That is, the processing device I / O execution bit 1701, the write after execution bit 1702, and the independent stage execution bit 1703 are executed.
Whether all bits of are off (is empty)
Check. If the master disk is free, the master disk is selected as an access target in step 1907. Specifically, the bit 1701 of the processing device I / O in the disk device information 1601 corresponding to the master disk is turned on. When the above processing is completed, the process jumps to step 1910, and enters the same processing as in the first embodiment.

If the master disk is not empty, the process jumps to step 1913 to wait for the input / output request.

In step 1908, the disk device 204 is assigned to a read request that requires access to the disk device 204. In this embodiment, an arbitrary free disk device is assigned to a read request that requires access to the disk device 204. Therefore, it is checked whether there is an empty disk device 204 in the disk device group 211 to be input / output. The specific processing contents are as follows. That is, the disk device information 1601 in which all the bits of the processing device I / O execution bit 1701, the write after execution bit 1702, and the independent stage execution bit 1703 are off is found. If not found, there is no empty disk device 204, and the process jumps to step 1913 because the disk device 204 cannot enter the input / output processing.

If found, in step 1909, the disk device 204 corresponding to the disk device number 1700 in the disk device information 1601 is selected as an access target. In particular,
The processing device I / O execution bit 1701 in the found disk device information 1601 is turned on.

Further, in step 1910, a pointer to the segment management information 1100 assigned to the track 800 to be accessed by the input / output request is set in the segment management information pointer 1704.

At step 1911, a positioning processing request is issued to the disk device 204 selected at step 1909.

In step 1912, a process for once disconnecting the connection with the processing device 210 is executed with the processing device 210 until the positioning process of the disk device 204 is completed. Thereafter, the processing of the input / output request receiving unit 1800 is terminated.

If the empty disk device 204 has been lost, step 191
The following process is executed in 3 and later.

In step 1913, the corresponding segment management information 1100
Turn off the busy flag 1205.

Thereafter, in step 1914, it is checked whether the stored flag 1204 is on. If it is on, jump to step 1916. If it is off, the segment management information is registered in the empty segment queue 1400 in step 1915 to indicate that no data is contained in the segment 1000 corresponding to the segment management information 1100.

Further, at step 1916, the processing device 210 is informed that the processing for the input / output request has not been executed for another input / output processing.
The processing request I / O waiting bit 1602 in 1500 is set. Specifically, the position of the processing request I / O wait bit 1602 to be set is determined from the following two points, and the bit is set.

First, the first point is the processing device that issued the input / output request.
That is, the number of the processing device 210 from the first to the I-th.

Next, the second point is how many I / O requests from No. 1 to J to the specified disk device group 210.

When the segment management information 1100 of the track 800 to be accessed by the input / output request is used by another input / output process, it is not necessary to operate the information in the segment management information 1100 in particular. Therefore, step 19
You will jump to 16.

Finally, step 1917 reports to processor 210 that the process for the I / O request will be in a wait state because it will not be executed for another I / O process. Thereafter, the processing of the input / output request receiving unit 1800 is terminated.

In step 1918, a process that needs to be executed is executed for an input / output request that does not need to access the disk device 204. The specific processing content is not directly related to the present invention, and thus the description is omitted.

FIG. 20 is a processing flow of the write-after processing schedule section 1801. Light after treatment schedule part 1801
Is executed using the time when the director 25 is available.

In step 2000, a disk device group 210 to be a write-after object is determined. Since this determination method has no direct relation to the present invention, the description is omitted.

In step 2001, an empty disk device to be an input / output target is found in the determined disk device group other than the master disk. The specific processing contents are as follows. That is, other than the master disk, the disk device information 1601 in which all the bits of the processing device I / O executing bit 1701, the write after executing bit 1702, and the independent stage executing bit 1703 are off is found. If not found, the write-after processing cannot be performed,
The processing of the write-after processing schedule section 1801 is terminated.

If found, in step 2002, the write-after-execution bit 1702 in the disk device information 1601 found in step 2001 is turned on.

In step 2003, the track 80 found in step 2001 and capable of executing the write-after
Check if there is 0. The specific check information is obtained from the track slip 1101 and the selected disk device 204.
The segment management information 1100 having the ON bit is searched for in the update record bit map 1203 in correspondence with. Further, since it is necessary that another processing request is not being used for the segment management information, the fact that the in-use flag 1205 in the segment management information 1100 is off is also a condition under which the write-after processing can be executed. If found, step
Jump in 2005. If there is no track 800 in which the write-after process can be performed, in step 2004, the write-under-execution bit 1702 is turned off. Thereafter, the processing of the write-after processing schedule section 1801 is terminated.
Select 800. When there are a plurality of tracks 800 capable of executing the write-after processing, which track 800 to select is not related to the present invention, and the description is omitted.

In step 2006, the track selected in step 2005
The busy flag 1205 in the segment management information 1100 corresponding to 800 is turned on.

In step 2007, the segment management information pointer 1704
The track 800 to be accessed by the input / output request
A pointer to the segment management information 1100 assigned to is set.

In step 2008, a positioning processing request is issued to the disk device 204 selected in step 2001. Thereafter, the processing of the rat-after-processing schedule section 1801 is terminated.

FIG. 21 is a processing flow of the independent stage processing schedule section 1802. Independent stage processing schedule section 1802
Also, it executes using the time when the director 25 is available.

In step 2100, a disk device group 210 to execute the stage processing 115 independent of the processing device 210 is determined. This determination method is also not directly related to the present invention, similarly to step 2000, and thus the description is omitted.

In step 2101, it is checked whether or not there is a track 800 in the disk device group 211 found in step 2100 to execute a stage process independent of the processing device 210. Since this check method is not directly related to the present invention, the description is omitted. If there is no track 800 to be executed, the processing of the independent stage processing schedule section 1802 ends.

In step 2102, a track 800 to be subjected to stage processing independent of the processing device 210 is selected. Processing unit 2
Track 800 that can execute stage processing independent of 10
In the case where there are a plurality of tracks, which track 800 is selected is not relevant to the present invention. Therefore, the description is omitted.

In step 2103, select and track in step 2004.
The segment management information 1100 is allocated to 800. (The track 800 to be subjected to the stage processing 115 independent of the processing device 210 is the track 800 not staged in the cache 206.) This allocation method is as shown in step 1902. Further, the stored flag 1204 in the allocated segment management information 1100 is turned off, and the busy flag 1205 is turned on.

In step 2104, an empty disk device to be input / output is found in the disk device group determined in step 2100. Since the specific processing contents are the same as those in step 1908, the description is omitted. If not found, the stage processing 115 independent of the processing device 210 cannot be executed. Therefore, in step 2105, the allocated segment management information 1100 is returned to the empty segment queue 1400. Thereafter, the processing of the independent stage processing schedule section 1802 is terminated.

If found, in step 2106, the independent stage execution bit 1702 in the disk device information 1601 found in step 2103 is turned on.

In step 2107, the segment management information pointer 1704
The track 800 to be accessed by the input / output request
A pointer to the segment management information 1100 assigned to is set.

At step 2108, a positioning process request is issued to the disk device 204 selected at step 2001. Thereafter, the processing of the independent stage processing schedule section 1802 is terminated.

FIG. 22 is a processing flowchart of the disk device transfer unit 1803. The trigger for the execution of the disk device transfer unit 1803 is when the director 205 receives the positioning completion report of the disk device 204.

At step 2200, the segment management information pointer in the disk device information 1601 corresponding to the disk device 204 is set.
The segment management information 1100 pointed to by 1704 is selected as a processing target. Hereinafter, when simply describing the segment management information 1100, it refers to the segment management information 1100 selected in step 2200. When the information in the segment management information 1100 is simply indicated, it indicates the information in the segment management embodiment 1100 selected in step 2200.

In step 2201, the processing device I / O execution bit 1 in the disk device information 1601 corresponding to the disk device 204 is set.
Check if 701 is on. If it is off, it indicates that the input / output processing being executed is not the processing for the input / output request received from the processing device 210.
Jump to 12.

If the processing device I / O execution bit 1701 is turned on, it indicates that the input / output process being executed is a process for the input / output request received from the processing device 210. Accordingly, in step 2202, the completion of the positioning process is reported to the processing device 210, and the connection state is established again.

In step 2203, it is determined whether the input / output request received from the processing device is a read request or a write request. In the case of a read request, the process jumps to step 2209.

In the case of a write request, in step 2204, the data received from the processing device 204 is written to the disk device 204 and the segment 1000 corresponding to the segment management information 1100 selected as a processing target. At this time, it is necessary to recognize the number of the cell 901 of the record 800 on the disk device 204 to which the data has actually been written, and perform the following processing. First, the segment
The data to be written in 1000 is also written in the position corresponding to the recognized cell 901. Further, the following processing is performed on the update record bit map 1203 corresponding to all the disk devices 204 other than the master disk in the segment management information 1100 selected as the processing target. That is, in the update record bitmap 1203, the cell recognized above
Turn on the bit corresponding to the number 901. Further, the completion of the input / output processing is reported to the post-processing device 210.

In step 2205, it is checked whether the stored flag 1204 in the segment management information 1100 to be processed is on. If it is on, the record 900 on the track 300 to be processed is staged in the segment 1000, and thus jumps to step 2215.

When the stored flag 1204 is off, the record 900 on the track 300 to be processed is not staged in the segment 1000. Therefore, the steps
The following processing is performed after 2206. In step 2206, the bit position of the record bit map 1202 corresponding to the number of the cell 901 recognized in step 2204 is turned on.

Next, in step 2207, processing is performed to stage the remaining sludge 900 of the track 800 being processed into the segment 1000. Also in this case, it is necessary to recognize the number of the cell 901 of the record 900 to be recorded and perform the following processing. First, records to be staged in segment 1000
900 is also written in the position corresponding to the recognized cell 901.
Further, the bit position of the record bit map 1202 corresponding to the recognized number of the cell 901 is turned on. Thereafter, in step 2208, the stored flag 1204 is turned on, and the process jumps to step 2215.

In steps 2209 and below, processing of the read request received from the processing device 210 is performed.

In step 2209, it is checked whether the stored flag 1204 in the segment management information 1100 to be processed is on.
If stored flag 1204 is on, record 900 is already
Are stored in the segment 1000. Therefore,
Copy record 900 requested in step 2210 to disk device 2
04 is sent to the processing device 210 from above. Thereafter, the completion of the input / output processing is reported to the processing device 210. Next, step 22
Jump to 15.

If the stored flag 1204 is off, the record 900 on the track 300 being processed is
Not staged within 1000. Therefore, the steps
The following processing is performed after 2211. First, in step 2210, the requested record 900 is sent from the disk device 204 to the processing device 210, and is staged in the segment 1000 as well. Also in this case, it is necessary to recognize the number of the cell 901 of the record 900 to be staged and perform the following processing.
First, the record 900 to be staged in the segment 1000 is also written at a position corresponding to the recognized cell 901. Further, the recognized cell 901 in the record bit map 1202 in the segment management information 1100 selected as the processing target is processed.
Turn on the bit corresponding to the number. Thereafter, the completion of the input / output request is reported to the processing device 210. next,
Remaining records 900 on the track 300 to be processed
To stage 2207.

In step 2212, the write-after-execution bit in the disk device information 1601 corresponding to the disk device 204 is set.
Check if 1702 is on. If it is off, the process jumps to step 2214 to execute the stage processing 115 independent of the processing device 210.

In step 2213, all the write-after records are recognized by the update record bitmap 1203 in the segment management information 1100 to be processed. In addition, all recognized light after records are stored on the disk device 204.
Write on top. Thereafter, the update record bit map 1203 corresponding to the disk device 204 is all cleared to zero. Next, jump to step 2215.

In step 2214, a stage process independent of the processing device 210 is executed. Specifically, all records 900 on the track 800 to be processed are staged in the segment 1000. Also in this case, it is necessary to recognize the number of the cell 901 of the record 900 to be staged and perform the following processing. First, records to be staged in segment 1000
900 is also written in the position corresponding to the recognized cell 901.
Further, the following processing is performed on the record bit map 1202 in the segment management information 1100 to be processed. That is, the bit position of the record bit map 1202 corresponding to the number of the recognized cell 901 is turned on. In addition, the stored flag 1204 is turned on.

At step 2215 and below, the following processing is performed as termination processing.

First, in step 2215, the in-use flag 1205 in the segment management information 1100 is turned off as a processing target. Next, in step 2216, all bits of the processing device I / O executing bit 1701, write after executing bit 1702, and independent stage executing bit 1703 in the disk device information 1601 corresponding to the disk device 204 are turned off. To

Finally, in step 2217, the following processing is performed to release the waiting state of the input / output request for which the processing request I / O waiting bit 1602 corresponding to the disk device group 211 to be processed is turned on. That is, the waiting state of all the input / output requests determined by the processing devices 210 from No. 1 to I and the input / output numbers from No. 1 to J are released from the bit being turned on. . Specifically, the respective processing devices are again notified to issue those input / output requests. Thereafter, the disk device transfer unit 18
The process of 02 ends.

Next, a second embodiment will be described. The differences between the second embodiment and the first embodiment are as follows.

A read request that needs to access the disk device group 211, a stage process independent of the processing device 210,
The point is that disk devices 204 other than the master disk are preferentially selected.

In the first embodiment, the respective data structures shown in FIGS. 8 to 17 can be used as they are in the second embodiment.

The configuration of the module necessary to execute the first embodiment in the director 205 shown in FIG. 18 can be applied to the third embodiment as it is. However, regarding the processing flow of each module, the input / output request receiving unit 180
0, Independent stage processing scheduler 1802 is slightly different from the first embodiment. However, the processing flow of the first embodiment can be applied as it is to the processing flow of another module.

FIG. 23 is a processing flow of the input / output request receiving unit 1800 in the third embodiment. I / O request receiving unit 1800
Is executed in the same manner as in the first embodiment.

Hereinafter, differences between the processing flow of the first embodiment shown in FIG. 19 and the processing flow of FIG. 23 will be described. Note that, in the processing flow of FIG. 23, steps that are the same as those in the processing flow of FIG. 19 have the same step numbers. The difference between FIG. 23 and the processing flow of FIG. 19 is that step 1908 in FIG.
This is the point where 00 is entered.

At step 2300, it is preferentially selected whether a disk device 204 other than the master disk is empty. This is because the second embodiment preferentially selects the disk device 204 other than the read request and the master disk which needs to access the disk device group 211. Specifically, the following information in the disk device information 1601 of the disk device 204 other than the master disk is checked. That is, it is checked whether all bits of the processing device I / O executing bit 1701, the write after executing bit 1702, and the independent stage executing bit 1703 are off.

If there is an empty disk device 204, that disk device 2
Step 1909 to select 04 as the access target
Then, processing similar to the processing flow in FIG. 19 is started. If there is no vacant disk device 204, the process jumps to step 1906 to check the vacancy of the master disk and enters the same processing as the processing flow of FIG.

Except for the following points, the processing flow in FIG. 23 and the nineteenth processing flow are exactly the same, and a description thereof will be omitted.

FIG. 7 is a processing flow of the independent stage processing scheduler 1802 in the second embodiment. The execution timing of the independent stage processing schedule unit 1802 is the same as that in the first embodiment.

Hereinafter, differences between the processing flow in the first embodiment shown in FIG. 21 and the processing flow in FIG. 7 will be described. In the processing flow of FIG. 7, the steps having the same processing contents as those of the processing flow of FIG. 21 have the same step numbers.

The processing flow of FIG. 7 differs from the processing flow of the first embodiment shown in FIG. 21 in the following points.

First, when the empty disk device 204 is found after step 2102, the second embodiment differs in that an empty disk device 204 other than the master disk is found in step 2400. This is because the second embodiment is different from the processing device 210 in that the stage processing is independent of the master disk 1.
This is because a disk device 204 other than 20 is preferentially selected. The specific processing content is the same as that in step 2300, and thus the description is omitted.

If there is an empty disk device 204, that disk device 2
In order to select 04 as a disk to be accessed, the process jumps to step 2104, and the same processing as in the first embodiment is started.

When the disk device 204 other than the master disk is in an empty state, it is checked in step 2401 whether the master disk is in an empty state. The above processing is performed in step 19
Since the process is the same as the process in step 06, the description is omitted. If the master disk is free, in step 2402, the master disk is selected as an access target. Specifically, the processing device I / O in the disk device information 1601 corresponding to the master disk
O Turn on running bit 1701. After this, step 2107
Then, processing similar to that of the first embodiment is started.

If the master disk is not empty, stage processing independent of the processing device 210 cannot be executed. Therefore,
The process jumps to step 2105 to enter the same processing as in the first embodiment. Except for the points described above, the processing flow in FIG.
Since the processing flow is the same, the description is omitted.

〔The invention's effect〕

According to the present invention, a high-performance / high-reliability control device having a cache having a function of writing the same data to a disk device group including one or more disk devices can be realized in a well-balanced manner. This is because, according to the present invention, the input / output processing can be distributed among the disk devices within a range that does not impair the reliability, and the parallelism of the input / output processing executable by the control device can be improved.

[Brief description of the drawings]

FIG. 1 shows a control device 203 according to a first embodiment of the present invention.
This is the basic operation of. FIG. 2 shows the configuration of a computer system to which the present invention is applied. FIG. 3 shows an input / output request received from the processing device 210,
The input / output request from the processing device 210 indicates a parallel operation of the input / output processing independently executed by the control device 203. FIG. 4 shows a parallel operation of a plurality of input / output requests received from a plurality of processing devices 210. FIG. 5 shows a parallel operation of a plurality of input / output requests received from one processing device 210. FIG. 6 shows a control device 203 according to the second embodiment of the present invention.
This is the basic operation of. FIG. 7 shows a processing flowchart of the independent stage schedule processing unit 1802 in the second embodiment. FIG. 8 shows the configuration of the disk device 24. FIG. 9 shows the configuration of the track 800. FIG. 10 shows the structure of the cache 206. FIG. 11 shows information required in the present invention in the directory 208. FIG. 12 shows information necessary for the present invention in the segment management information 1100. FIG. 13 shows the configuration of the track slip 1101 and the free segment queue start pointer 1102. FIG. 14 shows the storage format of the record 901 on the track 800 in the segment 1000. FIG. 15 shows information stored in the control memory 207. FIG. 16 shows the structure of the disk device group information 1500. FIG. 17 shows the structure of the disk device information 1601. FIG. 18 shows the modules in the director 205 related to the present invention. FIG. 19 shows a processing flowchart of the input / output request receiving unit 1800. FIG. 20 is a processing flowchart of the write after schedule processing unit 1801. FIG. 21 shows a processing flowchart of the independent stage schedule processing section 1802. FIG. 22 shows a processing flowchart of the disk device read / write processing unit 1803. FIG. 23 shows an input / output request receiving unit according to the second embodiment.
FIG. 1 shows a processing flow diagram of 1800. FIG. EXPLANATION OF SYMBOLS 203: control device, 210: processing device, 211: disk device group, 110: write request, 113: read request, 112: write after process, 115: stage process
120 …… Master disk.

Continued on the front page (72) Inventor Yoshihiro Azumi 2880 Kozu, Kokuzu, Odawara City, Kanagawa Prefecture Inside the Odawara Plant, Hitachi, Ltd. Inventor Hiroyuki Kitajima 1099 Ozenji Temple, Aso-ku, Kawasaki-shi, Kanagawa Prefecture Hitachi, Ltd. Systems Development Laboratory Co., Ltd. (56) References JP-A-58-219656 (JP, A) JP-A-59-118113 (JP, U) JP 61-28128 (JP, B2) (58) Fields investigated (Int. Cl. ⁶ , DB name) G06F 3/06-3/08

Claims (2)

(57) [Claims] 1. A method for controlling a storage device in a system including at least one storage device group including two or more storage devices and a control device connected to each storage device of the storage device group. In response to a read request for the storage device group received from a processing device, the control device responds to a read request for the storage device group other than a specific storage device in the storage device group to which no input / output processing is assigned. And assigning a process corresponding to the received read request to the selected one of the storage devices. Input / output processing is assigned to a storage device other than the specific storage device in the storage device group. If there is no storage device that has not been read, a specific storage device in the storage device group is assigned in response to the received read request. A method for controlling a storage device, comprising: selecting and assigning a process corresponding to the received read request to the selected specific storage device. 2. An at least one storage device comprising two or more storage devices.
A method for controlling a storage device in a system including one storage device group and a control device connected to each storage device of the storage device group and having a cache memory, wherein the control device receives an input / output request from a processing device. In a stage process of reading unprocessed data from a storage device in the storage device group to the cache memory, an input / output process is assigned to a storage device other than a specific storage device in the storage device group. Select any storage device that does not exist, assign the stage processing to the selected storage device, and store the storage device other than the specific storage device in the storage device group to which the input / output process is not assigned. When there is no device, a specific storage device in the storage device group is corresponded to the stage processing. A storage device, and allocating the stage processing to the selected specific storage device.