JP2004199711A - Semiconductor disk device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To restore into a semiconductor disk device the data saved in the disk device prior to changing the configuration of a semiconductor disk. <P>SOLUTION: When an instruction is given to restore the data saved in the disk device 18 into the semiconductor disk after the configuration of the semiconductor disk 14 is changed, a disk adapter 19 uses the respective configuration information tables 21, 22 of the semiconductor disk and the disk device to create an address conversion table 23 for converting a disk address into the address of the semiconductor disk for each logic drive. During the restoration, the disk adapter 19 uses the address conversion table to restore the data saved in each logic drive of the disk drive into the corresponding logic drive of the semiconductor disk. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

  The present invention relates to a semiconductor disk device, and more particularly, to a high speed access from a host device without mechanical operation by writing all user data stored in a magnetic disk device (DASD: Direct Access Storage Device) to a semiconductor memory. And a semiconductor disk device.

  The semiconductor disk device replaces the recording medium with a semiconductor memory from a magnetic disk while maintaining the behavior (command code, data transfer method, etc.) of the magnetic disk device. For this reason, the interface between the host device (CPU) and the semiconductor disk controller is exactly the same as the interface between the CPU and the magnetic disk controller. This semiconductor disk device has the advantage that it can be accessed instantly because the movement of the head is not necessary unlike the magnetic disk device, and has the advantage that the software resources between the CPU and the magnetic disk controller can be used as they are.

FIG. 57 is a configuration diagram of the semiconductor disk device. 1a is a CPU, 2 is a semiconductor disk device (SSD: Shared Storage Device), 3 is a semiconductor disk controller, 4 is a semiconductor disk, and a plurality of semiconductor memory modules (MS: Main Storage) 4a, 4b, 4c. 4n and a memory interface adapter (ESP: Extended Storage Adapter) 4s for controlling writing / reading of data to / from the semiconductor memory module. Reference numeral 5 denotes a maintenance panel or a personal computer.
In the semiconductor disk control device 3, reference numeral 3a denotes a channel adapter CA having a single or a plurality of interfaces (upper interface) with the upper device (CPU) 1a. Is provided. 3b is provided with an exclusive control table (not shown). When none of the upper interfaces uses a predetermined semiconductor memory module, the other upper interface permits the use of the semiconductor memory module. Is a resource manager RM that executes exclusive control that does not permit use. Actually, the semiconductor memory module is divided into a plurality of logical drives, and the resource manager RM performs exclusive control for each logical drive. Reference numeral 3c denotes a service adapter (SA) for performing IML (initial microprogram loading) processing, state monitoring processing, and recovery processing in the event of a failure in each unit. 3d, 3e and 3f are control storage units (Control Storage) for storing various control tables and programs.

First Problem In a semiconductor disk device, a failure of a semiconductor memory module is fatal. Conventionally, when a failure occurs in a semiconductor memory module, the data is evacuated, then the power is turned off, and the failed semiconductor memory module is replaced with another semiconductor memory module. After the replacement, the power is turned on, the semiconductor disk device is started, and data is restored. However, in this method, a device for saving data is separately required, and the semiconductor disk device cannot be used when the power is turned off and when the data is saved / restored, so that there is a problem that a request for a non-stop device cannot be answered. A method of replacing a failed semiconductor memory module without stopping the semiconductor disk device has also been proposed (Japanese Unexamined Patent Publication No. 3-268020, name: Nondisruptive maintenance method for semiconductor disk). However, the proposed non-stop maintenance method for semiconductor disks has a problem that a large-scale maintenance device for enabling non-stop maintenance is required separately from the semiconductor disk device.

Second problem By the way, when the power of the semiconductor disk device is turned off, its stored contents are lost. For this reason, a backup disk device may be connected to the semiconductor disk device. Each semiconductor memory module constituting a semiconductor disk is divided into a plurality of logical drives, and a higher-level device specifies a logical drive by a Start I / O command, and accesses a predetermined position of the logical drive if the logical drive is usable. It is supposed to. There is a 1: 1 relationship between the configuration of the logical drive of such a semiconductor disk and the configuration of the logical drive in the backup disk device. However, if the size or the position of the logical drive of the semiconductor disk is changed, the configuration of the logical drive between the semiconductor disk and the backup disk device does not correspond to 1: 1. Therefore, even if data is backed up to the backup disk device before the configuration of the semiconductor disk is changed, if the configuration of the logical drive of the semiconductor disk device is changed, the data backed up to the backup disk device is copied. There is a problem that the user data before the change is invalidated and cannot be used.

Third Problem Since the semiconductor disk device emulates a magnetic disk device, it has a control information portion called a directory (DIRECTORY) for each track in addition to the user data portion of the actual device, and the directory uses the control information portion. It manages the address on the memory and the sector information of the record (user data) in the track field currently being emulated.
When accessing the user data area of the specified track field, the channel adapter acquires the control information of the target track field by incorporating the directory of the specified track into the channel adapter, and stores the user data according to the information. to access. This directory includes (1) the record number of the last record among the records written in the track field, (2) the sector directory, and (3) the record directory. The sector directory is a table in which the number of the record which can be read first in the set sector processing is written. The record directory is a table in which a relative address (offset address) from the track head of each record is written. Yes, it is used to directly access the record field positioned by record number. By using these information, the orientation (virtual head position) can be moved directly during the set sector processing or search ID processing, and the access can be performed at high speed.

If the target directory cannot be read due to a two-bit error in the memory or the like, all user data in the track field controlled by the directory becomes inaccessible, and effective user data is effectively lost. . For this reason, a data loss avoidance mechanism is required as a data loss avoidance mechanism that enables access to user data even when a directory falls into an unreadable state.
In a conventional semiconductor disk device, when an attempt is made to access a track field specified by the firmware of the channel adapter, first, a directory in which control information of the track field is written is read, and then normal processing is started. Therefore, if the directory cannot be read, even if an attempt is made to access the track field, access is impossible because there is no recovery means at the time of directory collapse. Therefore, in order to be able to use the unreadable track again, it is necessary to initialize the minimum directory (for example, one cylinder unit), initialize the corrupted directory, and validate the directory. However, since the initialization process causes the user data in the area to be initialized to be lost, the data in the area to be initialized must be saved in advance. Further, the data in the track field which has become inaccessible due to the collapse of the directory can be read out only by a memory dump instruction prepared as one of the channel commands. For this reason, it is impossible for anyone other than those who know the track format inside the semiconductor disk device to extract and reproduce the necessary part of the track field as user data, and it is impossible to completely restore the original data. It was almost impossible.

Fourth Problem In a semiconductor disk device, a data storage / storage medium is a semiconductor memory chip. Therefore, the storage cost per bit is higher than that of the magnetic disk device. Also, the storage capacity per semiconductor disk device is reduced. In order to solve the problem concerning the capacity, a method of restoring the compressed data when writing and reading the data by compressing the data has been proposed. The problem with such a compression scheme is that when rewriting user data, the size after compression may differ from the size before compression. If the data size after compression is small, it is necessary to release an extra area for effective use of memory. Conversely, if the data size after compression is large, it is necessary to secure a new area and rewrite. There is. Such memory management is complicated. Conventionally, there has been a problem that the memory area cannot be released and allocated in a simple manner while effectively using the memory.

  From the above, it is an object of the present invention to be able to restore data saved in a backup disk device before a configuration change to a semiconductor disk device after a configuration change even if the configuration of the semiconductor disk is changed, and to use the user data before the change. To provide a semiconductor disk device capable of performing the above.

  According to the present invention, there is provided a semiconductor disk to which a plurality of logical drives are allocated, a disk device to which the semiconductor disk is backed up, and to which a plurality of logical drives are allocated, and each logical drive stored in the semiconductor disk. A disk adapter for saving data of the drive to the corresponding logical drive of the disk device and restoring the data saved to each logical drive of the disk device to the corresponding logical drive of the semiconductor disk; Means for holding a first configuration information table indicating a start address and a capacity; means for holding a second configuration information table indicating a start address and a capacity of each logical drive in the disk device; Half of the data saved to the device An instruction unit for instructing restoration to a body disk, and, when the first configuration information table is changed based on a configuration change instruction, using the first configuration information table and the second configuration information table before the change. Means for creating an address conversion table for converting a disk address into a semiconductor disk address for each logical drive.

  According to the present invention, even if the configuration of the semiconductor disk is changed, the data saved in the backup disk device before the configuration change can be correctly restored on the semiconductor disk device after the configuration change, and the user data before the change is invalidated. Can be used without any. Therefore, the configuration of the semiconductor disk can be freely changed as appropriate.

  After the configuration of the semiconductor disk is changed, the instruction means instructs restoration of the data saved in the disk device to the semiconductor disk. When this restoration is instructed, the address conversion table creating means uses the configuration information table of the semiconductor disk and the configuration information table of the disk device to convert the disk address into the address of the semiconductor disk for each logical drive. Create a translation table. At the time of restoration, the disk adapter restores the data saved in each logical drive of the disk device to the corresponding logical drive of the semiconductor disk using the address conversion table. In this way, even if the configuration of the semiconductor disk is changed, the data evacuated to the backup disk device before the configuration change can be correctly restored on the semiconductor disk after the configuration change, and the user data before the change can be used. Can be.

(a-1) Configuration of Semiconductor Disk Device FIG. 1 is a configuration diagram of a semiconductor disk device of the present invention. , 11a-1, 11a-2,..., CPU (upper-level device), 12 is a semiconductor disk device (SSD), 13 is a semiconductor disk control device, 14 is a semiconductor disk, and 15 is a service adapter (described later). This is a maintenance panel (PNL) for taking out and performing maintenance.
The semiconductor disk 14 includes a plurality of semiconductor memory modules (MS: Main Storage) 14a, 14b, 14c... 14n for storing user data, and when an error occurs in the semiconductor memory module, the semiconductor memory module is replaced with a new semiconductor memory module. A spare semiconductor memory module (HS: Hot Spare Memory) 16 serving as a substitute until the memory module is replaced, and a memory interface adapter (ESP: Extended Storage Adapter) 17 for controlling writing / reading of data to / from each semiconductor memory module. It has. Although only one spare semiconductor memory module (HS) is shown, two or more spare semiconductor memory modules (HS) can be provided.

Each semiconductor memory module is divided into a large number of ranges (range 1 to range n) in units of cylinders each having a predetermined size, for example, an exclusive control unit for access from the CPU. Evacuation control to the module 16 is performed, and the size of the range becomes an access control unit.
For example, a replacement procedure when an error occurs in the semiconductor memory module 14a is generally as follows.
(1) The storage contents of the semiconductor memory module 14a in which the error has occurred are saved in the spare semiconductor memory module (HS) 16 in units of access control. Exclusive control is performed in this access control unit.
(2) After the evacuation is completed, the spare semiconductor memory module 16 becomes a semiconductor memory module for storing user data.
(3) After the evacuation of all access control units is completed, the semiconductor memory module 14a in which the memory error has occurred is replaced with a new semiconductor memory module. Thereafter, the replaced semiconductor memory module 14a becomes a spare semiconductor memory module. In order to return the semiconductor memory module 16 to the spare, the contents stored in the semiconductor memory module 16 are restored to the semiconductor memory module 14a in access control units in the same manner as in the procedure (1).

  In the semiconductor disk controller 13, 13a-1, 13a-2,... Have a single or a plurality of interfaces (upper interfaces) with the upper devices (CPUs) 11a-1, 11a-2,. Each of the channel adapters (CA) and 13b has a resource manager (RM) and has an exclusive control table ECT (not shown). When none of the units uses a certain area of a predetermined semiconductor memory module, Upon request, another unit is allowed to access the area of the semiconductor memory module, and if it is in use, exclusive control is performed to prohibit access. In this embodiment, the resource manager manages use / non-use in units of access control and performs exclusive control in units of access control.

  Reference numeral 13c denotes a service adapter (SA) that performs an IML (initial microprogram loading) process, a status monitoring process, a recovery process in the event of a failure, for example, a semiconductor memory module replacement process when a memory error occurs, in each unit. 13d to 13g are control storage units (CS) for storing various control tables CTL and programs. As shown in FIG. 2, the correspondence between the logical address (CCHH) and the physical address of the semiconductor disk is stored in the control table CTL for each range (access control unit) of each semiconductor memory module. The logical address is an address specified by the CPU (for example, a cylinder / head number CCHH when specified by a data address of a magnetic disk), and the physical address is a real address for accessing the semiconductor memory module inside the semiconductor disk device. . Each module converts a logical address into a physical address using the control table CTL.

  A storage unit (TS) 13h stores a copy management table CCT indicating success / failure (valid / invalid) of copying. The copy management table CCT manages copy success / failure in a unit of data written in the semiconductor memory module. If the copy fails (for example, if the data could not be read due to a medium error), the relevant part is invalidated. For example, as shown in FIG. 3, the success / failure of copying is managed for each track. In the case of a semiconductor disk emulating a magnetic disk, if copying is performed for each cylinder (access control unit), even if an error occurs, not all of the cylinders are invalid. Therefore, fine data can be assured by invalidating only the error track for each track using the copy management table CCT. In this case, more detailed data can be assured if the data is managed on a record basis, not on a track basis. The copy management table CCT may be located at a place where each module can refer to / update, and may be arranged in the semiconductor memory module.

(a-2) Outline of First Exchange Control of Semiconductor Memory of the Present Invention As shown in FIG. 4, the semiconductor disk 14 includes three semiconductor memory modules (MS (1) to MS (3)) 14a to 14c. And one spare semiconductor memory module (HS) 16, each of which comprises five access control units. Initially, the contents of the control table CTL of each of the semiconductor memory modules 14a to 14c are as shown in FIG. In the figure, CTLa is a control table portion of the semiconductor memory module 14a (MS (1)), CTLb is a control table portion of the semiconductor memory module 14b (MS (2)), and CTLc is a semiconductor memory module 14c (MS (3)). ) Is a control table part. In the figure, XXXX indicates a logical address, MS (1) -00 indicates an offset address 00 (see FIG. 4) of the semiconductor memory module 14a (MS (1)), and other notations are the same. .

When a memory error occurs in the semiconductor memory module 14b (MS (2)), copying is performed from the semiconductor memory module 14b (MS (2)) to the spare semiconductor memory module 16 in access control units under the control of the service adapter SA. As shown in FIG. 6, the copy process proceeds, the copy of the first and second control units (6) and (7) is completed, and the copy of the third control unit (8) is performed by another module. If the fourth control unit (9) is currently being copied and the last control unit has not been copied yet, the contents of the control table CTL will be as shown in FIG. . That is, the physical addresses of the control units (ranges) (6) and (7) for which copying has been completed are the physical addresses of the semiconductor memory module 16 at the copy destination.
Thereafter, copying is performed for each control unit, and when copying of all control units is completed, the contents of the control table CTL are as shown in FIG. That is, the physical addresses of all control units (ranges) for which copying has been completed are the physical addresses of the copy destination.

(a-3) Exchange Control Process of Semiconductor Memory Module of the Present Invention Copy Process FIGS. 9 and 10 are flowcharts of a copy process for copying the storage contents of the semiconductor memory module to the spare semiconductor memory module.
In response to an instruction from the maintenance panel (PNL) 15 or error detection management, the service adapter (SA) 13c recognizes a semiconductor memory module in which a memory error has occurred (semiconductor memory module 14b) (step 101). Note that the service adapter SA always reads data from each memory module one by one sequentially and monitors whether a memory error has occurred (patrol). Since a one-bit error can be restored to the original data, a memory error is not immediately generated, but when a one-bit error exceeding a set value is detected, it is determined that a memory error has occurred. Also, the two-bit error is notified from the memory interface adapter (ESP) 17 to the service adapter SA.

When recognizing the semiconductor memory module in which the memory error has occurred, the service adapter SA requests the resource manager (RM) 13b to permit the use of the spare semiconductor memory module (HS) 16 (step 102). This is for confirming that the spare semiconductor memory module HS is not being used for replacing another semiconductor memory module MS. This is because, for example, when the semiconductor disk controller 13 is duplicated, the other service adapter SA may be performing a replacement operation.
When the semiconductor memory module HS is in use, the resource manager RM issues a non-permission notification to the service adapter SA (steps 103 and 104). Thus, the service adapter SA waits until the use is permitted. On the other hand, if the semiconductor memory module HS is not in use, the resource manager RM notifies the service adapter SA of the use permission, and writes "the semiconductor memory module HS is in use" in the built-in exclusive control table ECT. (Step 105).
Next, the service adapter SA requests the resource manager RM to permit the use of the first access control unit of the semiconductor memory module in which the error has occurred (step 106).

If usable, the resource manager RM notifies the service adapter SA of the use permission, and writes "the first access control unit is in use" in the exclusive control table ECT (steps 107 and 108). ).
The service adapter SA executes the copy of the control unit according to the use permission notification (step 109). If no error occurs during the copy and the copy of the access control unit is completed (steps 110 and 111), the service adapter SA instructs the modules 13a-1, 13a-2, and 13b to change the control table CTL. (Step 112). That is, the service adapter SA instructs each module to change the physical address in the control unit (6) of the control table CTLb to HS-00.
Each module changes the control table CTL of the control memory CS as instructed, and notifies the service adapter SA of the change completion (step 113). Note that the service adapter SA also changes the control table CTL of the control memory CS.

Upon receiving the change completion notification from all the modules, the service adapter SA notifies the resource manager RM of the end of the access. As a result, the resource manager RM writes "the first access control unit has become unused" (step 114). Next, the service adapter SA checks that copying of all control units of the semiconductor memory module 14b has been completed (step 115).
When the copying of all the control units is completed, the service adapter SA notifies the host device CPU of the completion of the copying via the channel adapter CA, displays the completion of the copying on the maintenance panel PNL (step 116), and ends the copying process. When the copy processing is completed, the maintenance person replaces the semiconductor memory module (MS) 14b in which the memory error has occurred with a new semiconductor memory module.
On the other hand, if the copying of all the control units has not been completed, the copy target is changed to another control unit whose copying has not been completed (step 117), and the processing from step 106 onward is repeated thereafter.
In step 107, when the first access control unit is being used, the resource manager RM issues a use-disapproval notice to the service adapter SA (step 118). Thereby, the service adapter SA jumps to step 115 and repeats the subsequent processing.

If an error occurs in a predetermined track at the time of copying in step 110, the service adapter SA notifies the host apparatus CPU of the occurrence of the error via the channel adapter CA (step 119), and the corresponding track in the copy management table CCT is written. A copy failure (invalid) is entered (step 120). After that, the processing after step 111 is repeated. FIG. 11 is an explanatory diagram of the copy management table CCT indicating copy success / failure, and shows a copy management table portion of the control unit (6) (the first access control unit of the semiconductor memory module 14b). The control unit (6) is composed of three tracks 1 to 3.
FIG. 11A shows a case where copying of all tracks of the control unit (6) succeeds without any error, and “valid” is written in tracks 1 to 3 of the copy management table CCT. FIG. 11B shows a case where an error has occurred during the copying of track 2, in which "invalid" is entered in track 2 and "valid" is entered in tracks 1 and 3 of the copy management table CCT. When data is written to a track where copying has failed, the track becomes "valid" and the copy management table CCT is rewritten.

-Processing after semiconductor memory module replacement FIG. 12 is a processing flowchart after semiconductor memory module replacement.
If the maintenance staff replaces the semiconductor memory module in which the error has occurred after the copy is completed, the service adapter SA detects the completion of the replacement (step 131), and thereafter, the replaced semiconductor memory module is used as a spare semiconductor memory module (step 131). 132). After the replacement is completed, the semiconductor memory module 16 may be returned to the spare semiconductor memory module according to an instruction from the maintenance panel PNL by the maintenance staff. Further, after the replacement is completed, the semiconductor memory module 16 may be automatically returned to the spare semiconductor memory module. FIG. 13 is a processing flowchart in such a case. The service adapter SA senses the completion of the replacement of the semiconductor memory module (Step 141), and copies the data stored in the semiconductor memory module 16 to the replaced semiconductor memory module 14b in the same procedure as the above-described copy processing (Step 142).

Access processing using a copy management table Each module refers to the copy management table CCT when reading a record from a semiconductor memory module. FIG. 14 is a processing flowchart of access with reference to the copy management table CCT.
First, it is determined whether the record is read or written (step 151). In the case of read, the validity / invalidity of the access destination track is checked by referring to the copy management table CCT (step 152). The record is read (step 153), and if invalid, error processing is performed (step 154). On the other hand, in the case of writing, if the access destination track is invalid, it is changed to "valid" (step 155), and thereafter, a record is written to the track (step 156).

As described above, only by adding a spare semiconductor memory module, the storage contents of the semiconductor memory module in which a memory error has occurred can be evacuated without stopping, and a large-scale maintenance device is not required. In addition, since the semiconductor memory module is subdivided into access control units and copying is performed in the access control units, exclusive control can be performed in the control units. It is possible to access the control unit, and there is no adverse effect on the access of the host device.
Further, every time the copy of the access control unit is completed, the service adapter SA changes the physical address in the control table CTL provided in the channel adapter CA or the like to the physical address of the copy destination. The access control unit can immediately access the copy destination (spare semiconductor memory module).
Further, a copy management table for managing whether or not copying has been normally completed for each access control unit is provided, and a module such as the channel adapter CA refers to the copy management table at the time of data reading to access an area (track) to be accessed. If) is normal, access is made. If abnormal, access is not made and an error occurs. As a result, even if a copy error occurs during copying, erroneous data is not read and processed, and a malfunction can be prevented.

(a-4) Outline of Second Exchange Control of Semiconductor Memory of the Present Invention The above is a case where the control table CTL has the configuration shown in FIG. The control table CTL may be configured as shown in FIG. The control table CTL includes (1) a logical address (CCHH), (2) an original physical address of a semiconductor disk, and (3) a physical address of a copy destination for each range (access control unit) of each semiconductor memory module. Is provided. The logical address is an address specified by the CPU or the like (for example, a cylinder / head number CCHH when specified by a data address of a magnetic disk), and the original physical address is a real address for accessing the semiconductor memory module inside the semiconductor disk device. is there. Normally, only the original physical address is associated with the logical address, and the physical address of the copy destination is invalid (control by attaching a mark, etc.).

  The semiconductor disk 14 includes three semiconductor memory modules (MS (1) to MS (3)) 14a to 14c and one spare semiconductor memory module (HS) 16, as shown in FIG. It is assumed that it is configured with five access control units. Initially, the contents of the control table CTL of each of the semiconductor memory modules 14a to 14c are as shown in FIG. In the figure, CTLa is a control table portion of the semiconductor memory module 14a (MS (1)), CTLb is a control table portion of the semiconductor memory module 14b (MS (2)), and CTLc is a semiconductor memory module 14c (MS (3)). ) Is a control table portion. In the figure, XXXX represents a logical address, MS (1) -00 represents an offset address 00 of the semiconductor memory module 14a (MS (1)), and the same applies to other notations.

When a memory error occurs in the semiconductor memory module 14b (MS (2)), copying is performed from the semiconductor memory module 14b (MS (2)) to the spare semiconductor memory module 16 in access control units under the control of the service adapter SA. In this case, prior to copying, the service adapter SA instructs each module to change the control table. That is, the control table is instructed so that (1) the original physical address and (2) the copy destination physical address correspond to each logical address of the semiconductor memory module 14b in which the error has occurred. As a result, the control table CTL of each module is changed as shown in FIG.
Further, the service adapter SA instructs a module such as the channel adapter CA to perform a forking process when accessing the semiconductor memory module 14b. In the forking process, as shown in FIG. 18, data reading reads data from an area specified by an original physical address, and data writing writes data to two storage areas indicated by an original physical address and a copy destination physical address. This is the writing process.

  Thereafter, as shown in FIG. 19, the access control unit is copied. When the copying of all access control units of the semiconductor memory module 14b is completed, the service adapter SA instructs each module to exchange the original physical address of the control table CTL and the physical address of the copy destination. As a result, the control table CTL of each module is changed as shown in FIG. Thereafter, the service adapter SA notifies each module of the end of the forking process, and instructs the address in the physical address entry column of the copy destination to be invalid, and ends the copy process. As a result, the control table CTL of each module becomes as shown in FIG. 21, and the original physical address becomes the copy destination physical address.

(a-5) Exchange Control of Semiconductor Memory Module of the Present Invention Copy Process FIGS. 22 and 23 are flowcharts of a copy process for copying the storage contents of the semiconductor memory module to the spare semiconductor memory module. It is assumed that the control table CTL is configured as shown in FIG.
In response to an instruction from the maintenance panel (PNL) 15 or error management, the service adapter (SA) 13c recognizes a semiconductor memory module (semiconductor memory module 14b) in which a memory error has occurred (step 201).
When recognizing the semiconductor memory module in which the memory error has occurred, the service adapter SA requests the resource manager (RM) 13b to permit the use of the spare semiconductor memory module (HS) 16 (step 202). When the semiconductor memory module HS is in use, the resource manager RM issues a non-permission notice to the service adapter SA (steps 203 and 204). Thus, the service adapter SA waits until the use is permitted. On the other hand, if the semiconductor memory module HS is not in use, the resource manager RM notifies the service adapter SA of the use permission, and writes "the semiconductor memory module HS is in use" in the built-in exclusive control table ECT. (Step 205).

Next, the service adapter SA instructs each module such as the channel adapter CA to start replacing the semiconductor memory module 14b and start forking processing (step 206). The service adapter SA instructs each module to change the control table. Accordingly, each module changes the control table CTL as instructed (see FIG. 17), and notifies the service adapter SA of the end of the change (step 207).
When the completion of the change of the control table is notified from each module, the service adapter SA requests the resource manager RM to permit use of the first access control unit of the semiconductor memory module in which the error has occurred (step 208).

If usable, the resource manager RM notifies the service adapter SA of the use permission, and writes "the first access control unit is in use" in the exclusive control table ECT (steps 209 and 210). ).
The service adapter SA executes the copy of the control unit according to the use permission notification (step 211). If no error occurs during the copy and the copy of the access control unit is completed (steps 212 and 213), the service adapter SA notifies the resource manager RM of the end of the access. As a result, the resource manager RM writes "the first access control unit has become unused" (step 214). Next, the service adapter SA checks that copying of all control units of the semiconductor memory module 14b has been completed (step 215).
If the copying of all the control units has not been completed, the copy target is changed to another control unit whose copying has not been completed (step 216), and the processing from step 208 onward is repeated thereafter.
In step 209, if the first access control unit is in use, the resource manager RM issues a use-disapproval notice to the service adapter SA (step 217). As a result, the service adapter SA jumps to step 215 and repeats the subsequent processing.

In step 212, if an error occurs in a predetermined track at the time of copying, the service adapter SA notifies the host device CPU of the occurrence of the error via the channel adapter CA (step 218), and the copy failure occurs in the corresponding track of the copy management table CCT. Success (invalid) is entered (step 219). Thereafter, the processing of step 213 and thereafter is repeated.
In step 215, when the copying of all the control units of the semiconductor memory module 14b is completed, the service adapter SA sends the control tables related to the semiconductor memory module 14b to the modules 13a-1, 13a-2, 13b, etc. An instruction is issued to exchange the original physical address of the portion with the physical address of the copy destination (step 220). As a result, each module changes the control table CTL as instructed (see FIG. 20), and notifies the service adapter SA of the end of the change (step 221). Next, the service adapter SA notifies each module of the copy completion and the end of the forking process (step 222), and invalidates the copy destination physical address column of the control table portion related to the semiconductor memory module 14b. (Step 223). Each module changes the control table CTL as instructed (see FIG. 21). As a result, the original physical address becomes the copy destination physical address.

Thereafter, the service adapter SA notifies the host device CPU of the end of the copy via the channel adapter CA, displays the completion of the copy on the maintenance panel PNL (step 224), and ends the copy processing. When the copy processing is completed, the maintenance person replaces the semiconductor memory module (MS) 14b in which the memory error has occurred with a new semiconductor memory module.
As described above, by simply adding a spare semiconductor memory module, the stored contents of the semiconductor memory module in which a memory error has occurred can be copied without stopping. Also, when writing data to the semiconductor memory module being copied, the channel adapter or the like writes data to two storage areas indicated by the original physical address and the copy destination physical address in accordance with the forking process. It is not necessary to change the control table every time the copy of the control unit is completed as in the case of the first exchange control process.

Processing when a Write Command Occurs in the Control Unit Area During Copy A write command may be issued from the host CPU in the control unit area during copy. In such a case, the write command can be made to wait until the copy is completed, but the access speed of the host device is reduced. Therefore, control is performed such that copying is temporarily stopped with priority given to the write command. FIG. 24 is a flowchart of a process when a write command is issued to the control unit area during such copying.
When a write command is issued from the host device CPU to the channel adapter CA (step 251), the channel adapter requests access permission to the resource manager RM for the access control unit (step 252). The resource manager checks whether the access control unit is being copied, and if not, notifies the channel adapter CA of access permission (steps 253 and 254). Thereby, the channel adapter CA executes the write command according to the forking process.

On the other hand, if copying is in progress, the resource manager 255 notifies the service adapter SA that a write command has been issued from the higher-level device to the control unit area being copied (step 255). As a result, the service adapter SA stops copying (step 256). Next, the resource manager RM permits access to the channel adapter CA (step 257), and the channel adapter CA executes the write command according to the forking process.
When the writing by the channel adapter CA is completed (step 258), the service adapter SA compares the write access unit (the number of tracks = Aw) with the copy access unit (the number of tracks = Ac, for example, Ac = 1). (Step 259).
If Aw ≧ Ac, the service adapter SA considers that the copy has been completed for the access unit of the write command (step 260), and thereafter resumes copying for the remaining portion.

On the other hand, when Aw <Ac, the copy unit including the access unit of the write command is copied again assuming that the copy has not been completed, or the copy unit other than the access unit is copied (step S1). 261) Then, copying is resumed for the remaining portion.
As described above, since the copy is stopped and the write command of the host device CPU is executed, the access speed of the host device does not decrease. When the access unit is larger than the copy unit, the copy of the access unit can be regarded as completed and the copy speed can be improved.

(a-6) Actual Configuration of Semiconductor Disk Device FIG. 25 is an overall configuration diagram of a semiconductor disk device, which has a duplex configuration, and a module having a suffix 1 is a module on the first semiconductor disk device G0 side. The module having the suffix 2 is the module on the second semiconductor disk device G1 side, and the module without the suffix is a common module. Note that the control storage unit CS is built in each module.
The CA is a channel adapter that controls the interface with the channel of the host device, and various channel adapters corresponding to an electric channel, an optical channel, and an OC link are appropriately connected. An RM is a resource manager that controls processing operations such as exclusive control and logical path management, and performs resource management of the entire subsystem. SA is a service adapter, which becomes a master and manages the status of other units (modules).

  C-BUS is a control bus for each unit to access message communication and control information, D-BUS is a data transfer bus for each unit to exchange data with the semiconductor disk, and S-BUS is a master for the service module. The service bus manages the status of each unit. BH-1 and BH-2 are bus handlers for controlling bus contention and distributing the bus clock, MDK is a magnetic disk device (optional) for temporarily backing up the contents of the memory when a memory failure occurs, and DA is a magnetic disk device. BANK is a semiconductor disk (shared memory) for controlling the interface of the semiconductor memory module, and has semiconductor memory modules MS and HS and a spare semiconductor memory module HS mounted thereon. ESP1 to ESP4 are ports (Extended Storage Ports) for controlling access to the semiconductor disk, and ESA1 to ESA4 are memories for controlling timing between the ESP and the semiconductor memory module MS, refreshing the memory, and correcting data based on error check codes. The interface adapter and PANEL are maintenance panels.

The first and second semiconductor disk devices G0 and G1 are configured symmetrically with respect to a center dotted line, and the upper CPU is symmetrically connected to the channel adapters CA1 and CA2 of the first and second semiconductor disk devices, and their respective ports. ESP2 and ESP3 are connected to the other memory adapters ESA3 and ESA4. Therefore, even if a failure occurs in one channel adapter, the CPU can access the semiconductor disk from the other channel adapter. Further, even if one semiconductor disk fails, the other semiconductor disk can be accessed, thereby improving reliability.
Modules such as the channel adapter CA, the resource manager RM, and the service adapter SA are each configured by a microprocessor, and generally have a configuration shown in FIG. In the figure, 91 is a microprocessor (MPU), 92 is a control storage unit (CS) having a RAM configuration, 93 is a control storage unit (CS) having a ROM configuration, and 94 is a driver / receiver (DV / RV) connected to an internal bus. ) And 95 are bus interface logic (BIL), 96 is a driver / receiver (DV / RV) connected to an external interface, 97 is a buffer or table storage unit (TS), and 98 is an individual LSI (gate array). The number of drivers / receivers (DV / RV) 96 differs depending on the number of external interfaces connected.

(b-1) Entire Configuration When the power of the semiconductor disk device is turned off, its stored contents are lost. For this reason, a backup disk device may be connected to the semiconductor disk device.
FIG. 27 is a configuration diagram of a semiconductor disk device provided with a backup disk device, and the same parts as those in FIG. 1 are denoted by the same reference numerals. 11a-1 is a CPU (upper device), 12 is a semiconductor disk device (SSD), 13 is a semiconductor disk control device, 14 is a semiconductor disk, and 15 is a maintenance panel (PNL) for issuing various instructions to the service adapter SA to execute maintenance. ).

The semiconductor disk 14 includes a plurality of semiconductor memory modules (MS: Main Storage) 14a, 14b, 14c... For storing user data, and when an error occurs in the semiconductor memory module, the semiconductor memory module is replaced with a new semiconductor memory. A spare semiconductor memory module (HS) 16 serving as a substitute for the module is provided. Reference numeral 17 denotes a memory interface adapter.
In the semiconductor disk controller 13, 13a-1 is a channel adapter (CA), 13b is a resource manager (RM), 13c is a service adapter (SA), and 13d, 13f, and 13g are control memories for storing various control tables CTL and programs. (CS). Reference numeral 18 denotes a backup magnetic disk device that stores data stored in the semiconductor disk 14 when the power is turned off, and reference numeral 19 denotes a data storage device that saves data stored in the semiconductor disk 14 to the magnetic disk device 18 and saves the data. This is a disk adapter that reads data from a magnetic disk device and restores the data to the semiconductor disk 14.

(b-2) Logical Drive Configuration of Semiconductor Disk and Magnetic Disk Each of the semiconductor memory modules (MS) 14a to 14c constituting the semiconductor disk 14 is composed of one printed board, and is inserted into a slot of a memory bank. . Each of the semiconductor memory modules (MS) 14a to 14c is divided into a plurality of logical drives of a predetermined size, and the host CPU designates the predetermined logical drive by a Start I / O command, and the logical drive is usable. If there is, a predetermined position of the logical drive is accessed.
FIG. 28 is a configuration diagram of a logical drive of the semiconductor disk 14 and the magnetic disk device 18. The semiconductor disk 14 includes two semiconductor memory modules 14a and 14b. The first semiconductor memory module 14a includes three logical drives 0 to 3, each of which has a size of 10 cylinders. The second semiconductor memory module 14b includes two logical drives 3 and 4 having a size of 10 cylinders.

The configuration of the logical drive in the magnetic disk device 18 has a one-to-one correspondence with the logical drive configuration of the semiconductor disk 14 as shown by the arrow. That is, the logical drive configuration of the magnetic disk drive 18 and the logical drive configuration of the semiconductor disk 14 are the same.
The first configuration information table 21 indicating the configuration of the logical drive in the semiconductor disk 14 holds a start address (cylinder address) and a capacity (the number of cylinders) for each of the logical drives 0 to 4, and stores a predetermined semiconductor memory module ( Master module) 14a. The second configuration information table 22 indicating the configuration of the logical drive in the disk device 18 has the same configuration as the first configuration information table 21 and is stored in the magnetic disk device 18.

When the data stored in the semiconductor disk 14 is saved to the magnetic disk device 18, the disk adapter (DA) 19 reads the first and second configuration information tables 21 and 22. Next, the data of the i-th logical drive (i = 0 to 4) is read from the semiconductor disk 14 by referring to the first configuration information table 21, and the data is read by referring to the second configuration information table 22. The data is stored in the i-th logical drive (i = 0 to 4) of the device 18.
When restoring the data saved in the magnetic disk device 18 to the semiconductor disk 14, the disk adapter (DA) 19 reads the first and second configuration information tables 21 and 22. Next, the data of the i-th logical drive (i = 0 to 4) is read from the magnetic disk 18 with reference to the second configuration information table 22, and the data is read with reference to the first configuration information table 21. Fourteenth i-th logical drives (i = 0 to 4).

(b-3) Problems when the logical drive configuration of the semiconductor disk 14 is changed By the way, there is a case where it is desired to change the logical drive configuration of the semiconductor disk 14 after saving the data of the semiconductor disk 14 to the magnetic disk device 18. FIGS. 29A and 29B are explanatory diagrams of the change of the logical drive configuration. FIG. 29A shows the configuration before the change, and FIG. 29B shows the configuration after the change. The size of the logical drive 0 is increased from 10 cylinders to 20 cylinders. 1 to the case where the start address of the logical drive 4 is changed.
When the configuration of the semiconductor disk 14 is changed as described above, the first and second configuration information tables 21 and 22 of the semiconductor disk 14 and the magnetic disk 18 in the conventional device are as shown in FIG. As a result, when the data saved on the magnetic disk 18 is restored on the semiconductor disk 14, the data is restored as shown by the arrow, and the data cannot be restored properly. That is, the data (enclosed by a dotted line) of the logical drives 0 and 1 saved on the magnetic disk 18 is restored to the logical drive 0 of the semiconductor disk 14, the data of the logical drive 2 is restored to the logical drive 1, and the logical drive 1 is restored to the logical drive 1. The data of the drive 3 is restored to the logical drive 2, and the data of the logical drive 4 is restored to the logical drive 3. As a result, the saved data cannot be used.
The above is the case where the size of the logical drive is changed, but the same problem occurs when the arrangement of the logical drive is changed as shown in FIG.

(b-4) Outline of the Restoration Process of the Present Invention FIG. 32 is a schematic explanatory diagram of the restoration process of the present invention when the logical drive configuration of the semiconductor disk 14 is changed after data is saved. 14 is a semiconductor disk, 18 is a magnetic disk device 19 is a disk adapter. In the semiconductor disk 14, reference numeral 20 denotes a user data storage area, reference numeral 21 denotes a first configuration information table for holding a logical drive configuration of the semiconductor disk, and reference numeral 23 denotes an address conversion table described later. In the magnetic disk device 18, reference numeral 22 denotes a second configuration information table for holding a logical drive configuration of the magnetic disk, and reference numeral 24 denotes a user data storage area. The semiconductor disk 14 and the magnetic disk 18 initially have the configuration shown in FIG.
After the data on the semiconductor disk 14 is saved to the magnetic disk 18, the configuration of the semiconductor disk 14 is changed as shown in FIG. 29B according to an instruction from the maintenance panel (PNL) 15. As a result, the first configuration information table 21 of the semiconductor disk 14 becomes as shown in FIG.

Next, when restoration is instructed from the maintenance panel PNL, the disk adapter (DA) 19 reads the first and second configuration information tables 21 and 22 from the semiconductor disk 14 and the magnetic disk 18, and reads the first and second configuration information tables. An address conversion table 23 (see FIG. 33) for converting a disk address into an address of a semiconductor memory is created for each logical drive using the configuration information tables 21 and 22 of FIG. The address conversion table 23 has (1) the start address and the number of cylinders of the semiconductor disk 14 and (2) the start address and the number of cylinders of the magnetic disk 18 for each logical drive.
Thereafter, the disk adapter DA reads the data of the i-th logical drive (i = 0 to 4) from the magnetic disk 18 with reference to the address conversion table 23 as shown in FIG. In the i-th logical drive (i = 0 to 4) of the semiconductor disk 14 to be used. As described above, the data of each logical drive saved on the magnetic disk 18 can be restored to the logical drive after the configuration change, and the data can be used.
After the restoration is completed, the second configuration information table 22 of the magnetic disk device 18 is matched with the first configuration information table 21 of the semiconductor disk 14.

(b-5) Restoration Process of the Present Invention FIG. 35 is a flowchart of the restoration process of the present invention.
If it is desired to change the logical drive configuration of the semiconductor disk 14 after saving the data to the magnetic disk device 18, a configuration information change command and logical drive configuration information are input from the maintenance panel (PNL) 15 (step 301). Thus, the service adapter SA updates the first configuration information table 21 stored in the semiconductor disk 14 (Step 302). Thereafter, a restore command is input from the maintenance panel PNL to restore the data saved in the magnetic disk device 18 to the semiconductor disk 14 (step 303). When the restore command is input, the service adapter SA instructs the disk adapter (DA) 19 to create the address conversion table 23 (Step 304).

  The disk adapter DA reads the first and second configuration information tables 21 and 22 from the semiconductor disk 14 and the magnetic disk device 18, and uses these first and second configuration information tables 21 and 22 to convert the address into the address conversion table 23 (see FIG. 32) and stored in the semiconductor disk 14 (step 305). Next, the disk adapter DA reads the data of the i-th logical drive (i = 0 to 4) from the magnetic disk 18 with reference to the address conversion table 23 and reads the data of the semiconductor disk 14 indicated by the address conversion table 23. The data is stored in the i logical drive (i = 0 to 4) (steps 306 and 307).

When the restoration of all data to the semiconductor disk 14 is completed, a configuration information matching command is issued from the maintenance panel PNL (step 308). When the configuration information matching command is issued, the service adapter SA instructs the disk adapter DA to change the second configuration information table 22 (Step 309). Thereby, the disk adapter DA reads the first configuration information table 21 from the semiconductor disk 14, and stores the second configuration information table 22 so that the first and second configuration information tables 21 and 22 correspond to 1: 1. The content is created and the content stored in the magnetic disk 18 is updated (step 310).
As described above, even if the configuration of the semiconductor disk is changed, the data saved in the backup disk device before the configuration change can be correctly restored on the semiconductor disk device after the configuration change, and the user data before the change can be used.

(c-1) Configuration of Track Format, Directory, and Count Section Track Format In the semiconductor disk device, a directory (control information section) is provided at the beginning of each track field. When accessing the user data of the designated track field, the channel adapter CA reads the directory of the track field and uses the directory to access the user data.
FIG. 36 shows an example of a track format in a semiconductor disk. A directory (DIRECTORY) is placed at the beginning of each track, a home address (HA) is placed after that, and a variable-length record (following the CKD format) RECORD-0, RECORD-1, ...) are arranged. The home address HA indicates a track address (CCHH), and each record includes a count section C, a key section K, and a data section D.

Directory The directory has 448 bytes as shown in FIG. 37, (1) 1-byte last record number (LAST RECORD NO.), (2) 218-byte sector directory (SECTOR DIRECTORY), (3) 190 bytes Record directory (RECORD DIRECTORY), (4) other areas such as an ID section for confirming validity, and (5) unused areas.
The last record number (LAST RECORD NO.) Indicates the record number of the last record written in the track field. An end mark (EOF) is written immediately after the last record of each track field, and the last record is the record immediately before reading the end mark. Then, the record number of the last record becomes the last record number.

The sector directory (SECTOR DIRECTORY) stores the physical record numbers that can be read first in the set sector processing for each sector value. The physical record number is a value obtained by adding 1 to the logical record number with the home address HA set to 0. The track is divided into, for example, 218 sectors, and sectors are assigned values from 0 to 217, respectively. The sector value 217 indicates the last sector when the track is divided into sectors of a predetermined size, and is the maximum sector value. This sector directory (SECTOR DIRECTORY) is used in the set sector process for the set sector command, and indicates which record can be read first in the sector process. Note that there are actually 0 to 221 sectors. However, records 218 to 221 do not exist in the sector directory because no records come after this due to the record arrangement.
The record directory (RECORD DIRECTORY) indicates an offset address from the track head to each record (record 0 to record 94) in 2 bytes. By adding the offset address of the target record to the head address of the track, the memory address of the count section of the target record can be obtained. The record number 94 indicates the maximum number of records that can be written on the track. This record directory (RECORD DIRECTORY) is used when reading the field of the record number acquired by the sector processing or when reading the record by directly specifying the record number.

As described above, each record is composed of a count section C, a key section K (not necessarily required) in which a key for search is recorded, and a data section D in which user data is recorded. The count section C has 64 bytes as shown in FIG. 38, (1) an 8-byte real data section provided by the host, and (2) a 4-byte ID section added before and after to guarantee data. , (3) A 48-byte control information section for accessing the track field. In the actual data portion, a track address (CCHH), a record number (R), a length of a subsequent key portion (KL), and a length of a data portion (DL) are recorded. In the ID part, a code for identifying the head of the record, the latest time (time stamp) at which the data was written, and the like are recorded. In the control information section, directory restoration data (control information restoration data) is recorded so that a target record can be searched even if the directory is lost, and a part of the directory can be restored. The directory restoration data includes (1) the relative address (offset address) Ai from the head of the own record to the head of the next record, and (2) the sector value Si of the sector where the own record is located.

Since the directory, the home address, and the length of the count portion of each record are constant, the count portions of the directory, HA, and RECORD-0 are located at relative positions determined from the head of the track. As shown in FIG. 39 (a), assuming that an offset address L0 from the head of the track to the home address HA and an off address from the record Ri-1 are Li, these offset addresses Li (i = 0, 1, 2,...) ) Are recorded in a record directory (RECORD DIRECTORY) (see FIG. 37). In this case, assuming that the offset address from the record Ri-1 to the record Ri is Ai, the offset address Ai is recorded in the count section of the record Ri as directory restoration data. The following equation is given between Ai and Li: Li + Ai → Li + 1
Is established.

  Assuming that the relationship between the sector and the record is as shown in FIG. 39B, 0 (HA) is changed to sector values 1 and 2 corresponding to the sector value 0 in the sector directory (SECTOR DIRECTORY). Correspondingly, 1 (record R0) is recorded corresponding to sector values 3 to 6, record 2 (R1) is recorded corresponding to sector values 7 to 13, and 3 (record R2) is recorded. That is, if the record Rj exists in the sector i, the record Rj + 1 is recorded in the sector directory in correspondence with the sector value i. Then, when the record Rj is located in the sectors i to (i + m), the sector values i to (i + m) are recorded as directory restoration data in the count part of the record Rj + 1.

(c-2) Record Read / Write Procedure FIG. 40 is an explanatory diagram of a record read sequence of the semiconductor disk device. The semiconductor disk device has the same configuration as that of FIG.
When a seek command SK is generated from the host device CPU, the channel adapter CA immediately returns an operation end signal to the host device. Upon receiving the operation end signal, the host device CPU issues a set sector command SS. Thereafter, the upper-level device similarly issues a search ID command SID, and the channel adapter CA receives these commands.
Upon receiving the search ID command, the channel adapter CA converts the logical address (track address CCHH) specified by the seek command into a physical address of the semiconductor disk. Since this physical address indicates the head address of the track, in other words, the head address of the directory, the directory is read from the semiconductor disk.

Next, a record number corresponding to the sector value specified by the set sector command is obtained from the sector directory (set sector operation). Thereafter, the address of the record number is obtained from the record directory, and the count part of the record is read from the address. Next, it is checked whether or not the CCHHR of the record included in the count section matches the CCHHR of the target record designated by the search ID command, and if they match, the search ID operation ends. If they do not match, the record of the next record number is read and the same processing is performed.
If the target record is found, the channel adapter CA returns an operation end signal to the host device. As a result, the host device issues a read command (READ CKD), and the channel adapter CA receives the command. Thereafter, the channel adapter CA sequentially transfers the count section C, the key section K, and the data section D to the upper-level device CPU, and finally sends a normal end to end a series of record read processing.

FIG. 41 is an explanatory diagram of a record writing sequence of the semiconductor disk device. The procedure is the same as the record reading procedure up to the search operation.
If the target record is found by the search ID operation, the channel adapter CA returns an operation end signal to the host device. As a result, the host device issues a read command (READ CKD), and the channel adapter CA receives the command. Thereafter, the channel adapter CA obtains directory restoration data (control information restoration data) Ai, Si and inserts them into the counting section, and thereafter records a record.
After the record recording is completed, the channel adapter CA notifies the host device CPU that the writing has been completed. After the notification, the contents of the directory are updated and written back to the semiconductor disk position where the directory was recorded.

(c-3) Processing when a directory read error occurs
(1) Skip search processing When reading or writing a record, if the channel adapter CA cannot read the directory, the target record is searched (skip) using the directory restoration data recorded in the count section of each record. Search). Note that when the directory cannot be read, a 2-bit error occurs, an ID error occurs (write failure of the directory), and the like. When such an error occurs, the channel adapter CA cannot read the directory, You will not be able to access.
FIG. 42 is a flowchart of the skip search process when the channel adapter CA cannot read the directory, and FIG. 43 is an explanatory diagram of the skip search process.
First, the channel adapter sets 1 → i (step 401). Next, the channel adapter CA reads the count part C0 of the record Ri-1 (= R0) from the address Li-1 (= L0) (step 402). The storage position of the record R0 from the beginning of the track is fixed, and L0 is known.

After the reading of the count section, it is checked whether the record number included in the count section is equal to the record number N instructed to access (step 403). If they are equal, the record Ri-1 is the target record, and the record is read or rewritten (step 404).
However, if the record number recorded in the count section is not equal to the record number N instructed to access, the relative address (offset address) Ai to the next record Ri recorded in the count section is obtained (step 405). ).
Then, the following equation Li-1 + Ai → Li
Thus, the offset address Li from the track head of the record Ri is calculated (step 406), and after the Li calculation, i is incremented (i + 1 → i, step 407), and thereafter the processing of step 402 and thereafter is repeated. As described above, the record number recorded in the count section in step 403 finally becomes equal to the record number N instructed to access, and a target record is obtained, and the record is read or rewritten.
In this way, even if a directory cannot be read when reading or writing a record specified by the host device CPU, the channel adapter CA automatically uses the directory restoration data recorded in each record to read the host device. It is possible to search for a designated target record and read or write the record.

(2) Reconstruction of directory When a directory cannot be read, a target record can be searched by the above skip search method. In this case, the directory can be reconstructed using the directory restoration data in parallel with the skip search processing.
Reconstruction processing of record directory FIG. 44 is a flowchart showing the reconstruction processing of the record directory.
First, L0 and L1 (fixed values) are set in the record directory as the home address HA and the offset address of the record R0 (step 421). Thereafter, i is set to 1 and the count section of the record Ri-1 is read (steps 422 and 423).
Next, it is checked whether the read information is an end mark (EOF) (step 424), and if it is not EOF, a relative address (offset address) Ai to the next record Ri is obtained from the counting section (step 425).

If Ai is obtained, the following equation is obtained: Li-1 + Ai → Li
Then, the offset address Li from the track head of the record Ri is calculated (step 426). Next, Li is set in the record directory as an offset address from the beginning of the record Ri (step 427), i is incremented (i + 1 → i, step 428), and thereafter the processing of step 423 and thereafter is repeated.
If the above process is repeated, sometime in step 424, the end mark is detected as EOF. If an end mark is detected, the remainder in the record directory is filled with "FF" (step 425).
Thus, the record directory can be restored.

FIG. 45 is a flowchart showing a process of reconstructing a sector directory.
The home address HA and the position of the count part of the record R0 are fixed. Therefore, H0 and R0 are set corresponding to the first several sector values in the sector directory (step 451).
Next, i is set to 1 and the count part of the record Ri-1 is read (steps 452 and 453). Thereafter, the relative address (offset address) Ai to the next record Ri is obtained from the counting section (step 454).

If Ai is obtained, the following equation is obtained: Li-1 + Ai → Li
Then, the offset address Li from the track head of the record Ri is calculated (step 455). Next, the count part of the record Ri is read (step 456), and it is checked whether the read information is an end mark (EOF) (step 457). If it is not EOF, the leading sector value Si of the record Ri is obtained from the count part of the record Ri (step 458). Next, a record Ri is written in correspondence with the sector value Si in the sector directory (step 459), and thereafter, i is incremented (i + 1 → i, step 460), and the processing from step 454 onward is repeated.

If the above process is repeated, sometime in step 424, the end mark is detected as EOF. If an end mark is detected, the remainder in the sector directory is filled with "FF" (step 461). Thus, the sector directory can be restored.
In this way, the channel adapter CA automatically uses the directory restoration data to read the directory even if an error that prevents the directory from being read occurs when reading or writing the record specified by the host device CPU. Since the rebuilding is performed, the directory can be referred to at the time of the next access, and high-speed access can be performed.

  FIG. 46 is an explanatory diagram of an overall schematic sequence when an error occurs in a directory. When a record read command READ D is issued from the host device (host) CPU, the channel adapter CA receives the command and performs a directory read process. When a directory reading error occurs, the channel adapter CA switches the access method from the normal directory use access method to the skip search method. Thereafter, the target record is skip-searched according to the procedure shown in FIG. 42, and when the target record is found, the data portion is transferred to the higher-level device. In this case, the directory can be reconstructed by performing the reconstruction process of the record directory and the sector directory in FIGS. 44 and 45 in parallel with the skip search process.

(c-4) Another process when an error occurs in the directory The above is the case when an error occurs in the directory, the target record is automatically searched, read / written, and the directory is reconstructed. explained. However, the search for the target record and the reconstruction of the directory may be performed in accordance with an instruction from the host device CPU.
FIG. 47 is an explanatory diagram of a record read procedure in such a case.
When a record read command READ CKD is issued from the host device (host) CPU, the channel adapter CA receives the command and performs a directory read process. In this case, when a directory reading error occurs, the channel adapter CA creates error information and reports the error to the host device.

The host CPU analyzes the error and issues a special dump command and a record number N to the channel adapter CA. The channel adapter CA accepts the special dump command, searches for a target record in accordance with the procedure shown in FIG. 42, and when the target record is obtained, sequentially transfers the count unit, the key unit, and the data unit to the host device.
The above is the case where the processing by the special dump instruction does not accompany the directory reconstruction processing. When the directory rebuilding process is involved, the record directory and the sector directory shown in FIGS. 44 and 45 are rebuilt, the directory is rebuilt, and the data is written at the head of the target track.

(c-5) Processing when Directory Error Occurs During Copying When a memory failure occurs in the semiconductor disk device, the contents of the semiconductor memory module MS in which the error has occurred are copied to the spare semiconductor memory module HS under the control of the service adapter SA. Thereafter, the failed semiconductor memory module is replaced with a new semiconductor memory module (first embodiment, see FIG. 1).
In such a case, the service adapter SA performs copying other than the track in which the error has occurred, and requests the channel adapter CA to copy the track field in which the error has occurred.
This is because the service adapter SA cannot perform processing in units of fields, and performs block copying for each fixed length (64 bytes). For this reason, even if a similar copy process is performed on the copy of the track field in which an error has occurred, the error occurrence information cannot be written in the directory or the additional information of each count unit, and access to user data cannot be guaranteed. is there. Therefore, the service adapter SA requests the channel adapter CA for the copy processing of the track field in which the memory failure has occurred, and the requested channel adapter CA performs the copy processing in order from the directory to the end mark EOF in field units.

  If the channel adapter CA finds a field (counting part, key part, data part) that causes a memory failure in the middle of copying, it invalidates only that field (in the case of the counting part, the entire record) and necessitates the counting part. Rewrite the control information. However, if the directory is invalidated when a memory failure occurs in the directory, all records on the track field are invalidated. Further, since the channel adapter CA cannot read the directory, the format of the track is changed. I can't judge.

  In order to avoid this, the channel adapter CA copies the HA field at the fixed position and the record R0, and copies the fields after the record R1 to the end mark while acquiring the address of the next record field by the skip search method. I do. In other words, by reading the count part of the current record, the offset address to the next record required for reconstructing the directory, the sector value of the current record, the key length of the current record, and the data length can be obtained. Go to the next record. If this is continued until the end mark is detected, all fields except the directory can be copied. Further, since the copy process includes a process of reading directory restoration data of all record parts necessary for directory reconstruction, the directory can be necessarily reconstructed. Therefore, the directory is reconstructed at the same time as the copying process of each field is performed, and the directory reconstructed in the channel adapter CA is finally written in the copy destination directory area, thereby preventing the loss of user data during the memory exchange process. be able to.

FIG. 48 and FIG. 49 are flowcharts of copy processing when an error occurs in the semiconductor memory. It is assumed that the semiconductor disk device has the configuration shown in FIG.
When recognizing the semiconductor memory module MS and the track in which the error has occurred (step 501), the service adapter SA starts copying the semiconductor memory module MS (step 502).
It is checked whether the track to be copied is an error track (step 503). If the track is not an error track, the copy of the track is executed (step 504). Step 505), if completed, perform the processing from step 503 on for the next track.

On the other hand, if the track to be copied is an error track in step 503, a copy of the error track is requested to the channel adapter CA (step 506). The channel adapter CA starts copying in field units and reads the fields (step 507). If an error is detected in the field (step 508), it is checked for an error in the directory (step 509). If the error is not a directory error, the field is invalidated (step 510) and the process jumps to step 507. However, if there is a directory error, the flag F is set to “1” (“1” → F, step 511), and the routine jumps to step 507.
If there is no error in the field in step 508, the field is copied (step 512), and it is checked whether F = "1" (step 513). If F = "1", The directory is reconstructed (step 514).

Thereafter, it is checked that the end mark EOF has been detected (step 515), and if not detected, the process jumps to step 507 to copy the subsequent fields.
If the end mark EOF is detected, it is checked whether F = "1" (step 516), and if F = "0", the copy completion is notified to the service adapter SA (step 517). As a result, the service adapter repeats the copy processing from step 503 on for the next track.
In step 516, if F = “1”, the reconstructed directory is written to the copy destination directory area (step 518), then F is set to “0” (“0” → F, step 519), and the service The copy completion is notified to the adapter SA (step 517). As a result, the service adapter repeats the copy processing from step 503 on for the next track.

As described above, according to the third embodiment, even if the directory cannot be read, access to the target record specified by the higher-level device using the control information restoration data included in each record without using the directory is performed. It becomes possible. Further, since the directory can be reconstructed using the control information restoration data, access can be made using the reconstructed directory. Therefore, the access speed does not decrease.
In addition, even if an error that makes it impossible to read the directory is detected in the copy processing when replacing the semiconductor memory module due to a memory error, the next record position is sequentially determined using the control information restoration data included in each record. Can be copied, and the directory can be reconstructed and written to the copy destination directory area.

(d-1) Overall Configuration FIG. 50 is a configuration diagram of an embodiment of a semiconductor disk device for compressing data and storing the compressed data in a semiconductor disk. The semiconductor disk device has a duplex configuration. The module having the suffix 2 is the module on the side of the disk device G0, and the module having the suffix 2 is the module on the side of the second semiconductor disk device G1.

  CA1 and CA2 are channel adapters that perform interface control with the CPU and channel of the host device, and are connected to the host device (channel). RM1 and RM2 are resource managers that control processing operations such as exclusive control and perform resource management of the entire subsystem. TS1 and TS2 are table storages (table storage units) for storing tables such as an exclusive control table, SA1 and SA2 are service adapters, MDK is a magnetic disk device for temporarily backing up the contents of a semiconductor disk, and DA1 and DA2 are Device adapters BANK1 and BANK2 for controlling the interface with the magnetic disk device are semiconductor disks, on which a plurality of semiconductor memory modules and spare semiconductor memory modules are mounted. ESP1 and ESP2 are ports for controlling access to the semiconductor disk, MT1 and MT2 are semiconductor memory tables for storing the used / unused state of each block when the semiconductor memory is divided into a large number of blocks of a predetermined size, and MCA1 and MCA2. Manages the used / unused state of each block in the semiconductor memory, allocates an unused block according to a block request from the channel adapters CA1 and CA2, sets the block in use, and sets a predetermined state according to a block release request from the channel adapter. Change the used block of to unused.

BUS1 and BUS2 are buses, and include a control bus, a data transfer bus, and a service bus. CBUS is a copy bus for immediately copying the contents of one table storage unit TS1 (TS2) to the other table storage unit TS2 (TS1). CBUS 'is a copy bus for changing the contents of one semiconductor memory table MT1 (MT2). This is a copy bus that immediately copies the data to the other semiconductor memory table MT2 (MT1).
The first and second semiconductor disk devices G0 and G1 are configured symmetrically about a center dotted line, and the upper CPU is symmetrically connected to the channel adapters CA1 and CA2 of the first and second semiconductor disk devices.

(d-2) Configuration of Semiconductor Memory As shown in FIG. 51, the semiconductor memory BANK is divided into a track emulation area 51 and a compressed data storage area 52 for compressing and storing a data portion of a record. The compressed data storage area 52 is divided into a large number of blocks (logical blocks) of a predetermined size, and after the data portion of each record is compressed, it is stored in a required number of blocks continuously or dispersedly.

Each track in the track emulation area 51 has the track format shown in FIG. 36 except that the data part of the record is not stored. As shown in FIG. 52, each track field stores a directory, a home address HA, a count portion and a data portion of record 0, a count portion of each record, and a key portion. The control information for controlling the field, the record number CCHHR, the key length KL, the data length DL, and the like are written in the count section C of each record Ri (i = 1, 2,...). The block number (block address) of the data storage area 52 is written.
To write a record, as shown in FIG. 53, the data portion DT of the record to be written is compressed into compressed data DT ', and then a number of blocks (three in the figure) corresponding to the size of the compressed data DT' Secured, the first to third parts (1) to (3) of the compressed data DT 'are stored in the secured blocks B1 to B3, and the block addresses of these blocks B1 to B3 are inserted into the counting section to perform track emulation. It is stored in the area 51.

(d-3) Configuration of Channel Adapter FIG. 54 is a configuration diagram of a channel adapter that performs compression / decoding processing. Inside the channel adapter, 61 is a large-capacity (for example, one track) first data buffer for storing uncompressed data, 62 is a large-capacity (for example, one track) second data buffer for storing compressed data, 63 Is a data compression mechanism for compressing / decoding data, 64 is a dictionary memory for assisting the compression / decoding operation, 64 is a channel interface protocol controller, and 65 is a selector for selecting data from the first and second data buffers. A selector 66 for selectively outputting data to the first and second data buffers while outputting the data to the first and second data buffers. An internal bus interface control unit (BIL) 66 controls an internal bus connected to another unit in the semiconductor disk device. ), 67 are first counters for counting the number of bytes of uncompressed data, and 68 is a first counter for counting the number of bytes of compressed data. Counters 69 and controls these hardware resources by a microprogram MPU, 70 program control table control storage for other tables such, 71 denotes a bus.

(D-4) Record Read Process FIG. 55 is a sequence explanatory diagram of the record read process.
When a seek command SK is generated from the host device CPU, the channel adapter CA (MPU 69) immediately returns an operation end signal to the host device. Upon receiving the operation end signal, the host device CPU issues a set sector command SS. Thereafter, the upper-level device similarly issues a search ID command SID and a read command (READ CKD), and the channel adapter CA receives these commands.
When receiving the read command, the MPU 69 of the channel adapter CA converts the logical address (track address CCHH) specified by the seek command into a physical address of the semiconductor disk using the control table. Since this physical address indicates the head address of the corresponding track in the track emulation area 51, in other words, the head address of the directory, the directory is read from the semiconductor disk.

Next, a record number corresponding to the sector value specified by the set sector command is obtained from the sector directory (set sector operation). Thereafter, the address of the record number is obtained from the record directory, the count part of the record is read from the address, and the CCHHR of the record included in the count part matches the CCHHR of the target record specified by the search ID command. To check. If they match, the search ID operation ends. If they do not match, the record of the next record number is read and the same processing is performed.
When the target record is found, the count section C and key section K of the record are read out via the internal bus interface control section (BIL) 66. The selector 65 sets the counting unit C and the key unit K in the first data buffer 61, and the channel interface protocol control unit 64 transfers the counting unit and the key unit to the host device.
Next, the MPU of the channel adapter CA sets the block address (not limited to one) written in the count unit C to BIL. As a result, the BIL sequentially reads data from the block indicated by the block address and writes the data to the second data buffer 62 via the selector 65. The data compression mechanism 63 decompresses the compressed data and sets it in the first data buffer 61, and the channel interface protocol control section 64 transfers the data to the higher-level device.

(D-5) Record Rewriting Process FIG. 56 is a flowchart of the record rewriting process. Until a write command (WRITE D) is received, the process is the same as that of the read command.
When receiving the write command (WRITE D) from the higher-level device, the channel interface protocol controller 64 temporarily stores the record in the large-capacity data buffer 61 (step 601). Next, the MPU 69 of the channel adapter CA converts the logical address (track address CCHH) specified by the seek command into the physical address of the semiconductor disk using the control table. Since this physical address indicates the directory position of the corresponding track, the directory is read (step 602).

Thereafter, as in the case of the read command, the address of the target record is obtained, and the count section of the target record is read from the address (step 603). Then, the block address and the number Bo of blocks included in the counting section are stored (step 604). The block address and the number of blocks indicate the address of the block in which the data part of the target record is stored and the number thereof.
When the storage of the block address and the number of blocks Bo is completed, or in parallel with the above, the data compression mechanism 63 compresses the data portion of the record and stores it in the second data buffer 62 (step 605). Since the number of bytes of the compressed data is monitored by the second counter 68, the required number of blocks Bn is obtained by dividing the number of bytes by the capacity of one block (step 606).
If the number of blocks Bn is obtained, the magnitude of the block is compared with the number of blocks Bo so far (step 607). If Bn = Bo, the block address of the past is set in BIL (step 608), and the block address is determined. Write the compressed data to the block designated by (1) (step 609).

  On the other hand, when Bn <Bo, the MPU 69 instructs the (Bo−Bn) unnecessary block addresses to the semiconductor memory management adapter MCT. As a result, the semiconductor memory management adapter MCT changes the contents of the memory table MT so that the designated block address is not used (step 610). Further, the MPU 69 deletes the block address of the block instructed to be released from the count section of the target record (step 611). Thereafter, the MPU 69 of the channel adapter sets the required number of block addresses in the BIL 66 (step 608), and writes the compressed data to the block indicated by the block address (step 609).

  If Bn> Bo, the MPU 69 requests the semiconductor memory management adapter MCT to allocate (Bn-Bo) missing blocks (step 612). Upon receiving the block request, the semiconductor memory management adapter MCT finds the requested number of unused blocks by referring to the memory table MT, changes the used block to be in use, and changes the obtained block address of the unused block to the channel adapter. Notify CA. When receiving the required number of block addresses from the semiconductor memory management adapter MCT (step 613), the MPU 69 of the channel adapter CA adds the block address of the newly allocated block to the count section of the target record (step 614). Thereafter, the MPU 69 of the channel adapter sets the number of block addresses necessary for writing the new record in the BIL 66 (step 608), and writes the compressed data sequentially to the block indicated by the block address (step 609).

As described above, according to the fourth embodiment, the user data is compressed and stored, so that the memory can be effectively used. Further, since the control information (directory, count unit, etc.) is not compressed, the restoration processing is not required, and the host Data access time can be reduced. According to the fourth embodiment, when data is compressed and written into the semiconductor memory, the memory area can be effectively released and allocated to effectively use the memory.
As described above, the present invention has been described with reference to the embodiments. However, the present invention can be variously modified in accordance with the gist of the present invention described in the claims, and the present invention does not exclude these.

FIG. 1 is a configuration diagram (first embodiment) of a semiconductor disk device of the present invention. FIG. 4 is an explanatory diagram of a control table. FIG. 4 is an explanatory diagram of a copy management table. FIG. 2 is a configuration diagram of a semiconductor disk. FIG. 4 is an explanatory diagram of the contents (initial state) of a control table. FIG. 9 is an explanatory diagram of a copy process. FIG. 8 is an explanatory diagram of the contents (intermediate) of a control table. FIG. 9 is an explanatory diagram of the contents (copy completed) of a control table. FIG. 7 is a flowchart (part 1) of a copy process. It is a flowchart (the 2) of the copy process. FIG. 4 is an explanatory diagram of a copy management table. FIG. 11 is a process flowchart after replacement of the semiconductor memory module. FIG. 14 is another flowchart of the process after the replacement of the semiconductor memory module. FIG. 14 is an access processing flowchart using a copy management table. It is another block diagram of a control table. FIG. 8 is an explanatory diagram of the contents (at the initial stage) of a control table. FIG. 9 is an explanatory diagram of the contents of a control table (after a change instruction). It is a forking process explanatory drawing. FIG. 9 is an explanatory diagram of a copy process. FIG. 9 is an explanatory diagram of the contents of a control table (after a physical address exchange instruction). FIG. 9 is an explanatory diagram of the contents of a control table (at the end of copying). FIG. 7 is a flowchart (part 1) of a copy process. It is a flowchart (the 2) of the copy process. FIG. 11 is a processing flowchart when a write instruction is issued to a control unit area during copying. FIG. 3 is an actual configuration diagram of a semiconductor disk device. It is a block diagram of each module. FIG. 11 is another configuration diagram (second embodiment) of the semiconductor disk device of the present invention. This is the configuration of the logical drive. FIG. 4 is an explanatory diagram of a change in a logical drive configuration. FIG. 7 is an explanatory diagram of a conventional problem when a logical drive is changed. FIG. 14 is another explanatory diagram of a logical drive configuration change. It is a schematic explanatory view of the restoration processing of the present invention. FIG. 4 is an explanatory diagram of an address conversion table. FIG. 9 is an explanatory diagram of restoring user data. It is a flowchart of the restoration processing of this invention. FIG. 3 is an explanatory diagram of a track format in a semiconductor disk. It is a block diagram of a directory. It is a block diagram of a count part. FIG. 3 is an explanatory diagram of a configuration of a track field. FIG. 4 is an explanatory diagram of a record reading sequence. FIG. 9 is an explanatory diagram of a record writing sequence. FIG. 14 is a flowchart of a skip search process using counting unit additional information. FIG. 9 is an explanatory diagram of a skip search process using the counting unit additional information. FIG. 14 is a flowchart of a record directory reconfiguration process. FIG. 14 is a flowchart of a process of reconstructing a sector directory. FIG. 7 is an explanatory diagram of an entire sequence when an error occurs in a directory. FIG. 3 is an explanatory diagram of a host access using the track special access method. FIG. 11 is a flowchart (part 1) of a copy process when an error occurs in the semiconductor memory. FIG. 11 is a flowchart (part 2) of a copy process when an error occurs in the semiconductor memory. FIG. 14 is a configuration diagram (fourth embodiment) of a semiconductor disk device. FIG. 4 is an explanatory diagram of a memory map. It is an explanatory view of a track emulation area. FIG. 9 is an explanatory diagram of storing user data in a logical block. It is a block diagram of a channel adapter. FIG. 4 is an explanatory diagram of a record reading sequence. It is a flowchart of the record rewriting process. FIG. 2 is a configuration diagram of a conventional semiconductor disk device.

Explanation of reference numerals

11a-1, 11a-2, host device 12, semiconductor disk device 13, semiconductor disk control device
13a-1 Channel adapter 13b Resource manager 13c Service adapter 14 Semiconductor disks 14a to 14n Multiple semiconductor memory modules 15 Maintenance panel (PNL)
16. Spare semiconductor memory module 18 Backup magnetic disk unit 19 Disk adapters 21, 22 First and second configuration information table 23 Address conversion table

Claims (4)

A semiconductor disk to which a plurality of logical drives are assigned;
A channel adapter that controls writing and reading of data to and from a semiconductor disk;
A disk device to which a plurality of logical drives are allocated while backing up a semiconductor disk;
A disk adapter that saves data of each logical drive stored in a semiconductor disk to a corresponding logical drive of the disk device and restores data saved to each logical drive of the disk device to a corresponding logical drive of the semiconductor disk In the semiconductor disk device having
A first configuration information table for holding a start address and a capacity of each logical drive in the semiconductor disk;
A second configuration information table that holds the start address and capacity of each logical drive in the disk device;
Instructing means for instructing a configuration change of the semiconductor disk and restoring data saved in the disk device to the semiconductor disk
When the first configuration information table is changed based on the configuration change instruction, the disk address of the semiconductor disk is set for each logical drive using the first configuration information table and the second configuration information table before the change. Means for creating an address conversion table for converting to an address,
A disk adapter restores data saved in each logical drive of the disk device to a corresponding logical drive of the semiconductor disk by using the address conversion table when the disk adapter is restored.   2. The semiconductor disk device according to claim 1, wherein the means for creating the address translation table is a disk adapter, and the disk adapter creates the address translation table when restoration is instructed.     3. The semiconductor disk device according to claim 2, wherein the first configuration information table is stored on a semiconductor disk, and the second configuration information table is stored on a disk device. 4. The semiconductor disk device according to claim 3, wherein the disk adapter matches the second configuration information conversion table with the first configuration information conversion table after the restoration is completed.

Images