JP3457394B2 - Information storage device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information storage device in which only dirty data is duplicated to improve reliability and clean data is unified without duplication.

[0002]

2. Description of the Related Art Conventionally, in order to improve the reliability of a storage device, the storage medium is duplicated so that data is not lost even if a failure occurs in one of the storage media, and a spare storage medium is restored by a recovery process. There is one that copies data from a normal storage medium and maintains duplication.

In Japanese Patent Laid-Open No. 5-27918, information on whether a file is duplicated is recorded in an area called a file label on a storage medium, and when a failure occurs, the duplicated file is recorded. There is something that will be restored. However, this method has a problem in that a file that has not been duplicated is lost when a failure occurs.

Regarding the restoration processing, the purpose is not to stop normal access during the restoration processing, and the concept of the boundary between the restored portion and the unrestored portion is used, and a block for which access is requested during the restoration processing is performed. A method for determining whether or not to recover (Japanese Patent Laid-Open No. 5-35413),
For the purpose of further shortening the recovery time without stopping the normal access, when the normal access is a write, the data is directly written to the spare storage medium without generating the data from the normal storage medium (Japanese Patent Laid-Open No. 5-88988). There is.

As a general method for accessing data on a low-speed storage medium having a large capacity at a high speed, there is a method of using a high-speed recording medium having a small capacity as a cache. A medium that is slower than the cache is called secondary storage. Data that is frequently accessed by the host is copied to the cache, and at the time of reading, high-speed access is possible if there is data in the cache (this is called hitting the cache). At the time of writing, the host is notified of the end when the data is written to the cache, and the writing speed is improved. The state of the data on the cache is called dirty, and is written to the secondary storage after the host is notified of the end. The state of data when the contents of the cache data and the contents of the secondary storage are the same as this data is called clean.

[0006]

However, when the cache is duplicated in order to improve its reliability, there are the following problems. First of all, if all the data in the cache is duplicated, the clean data is also duplicated, and if the original data in the secondary storage is combined, it becomes tripled, resulting in an extra data area. In order to reduce waste, it is necessary to coexist a duplicated file and a non-duplicated file as in the above-mentioned Japanese Patent Laid-Open No. 5-27918.
It is also necessary to recover the data of files that are not duplicated. Also, the cache requires management in block units smaller than files.

Secondly, if the access is not stopped during the recovery from the failure, and the order of the data to be recovered is not related to the access frequency, there is a high possibility that the data will be recovered after the access is requested. Depending on the time required for, the effect as a cache becomes weak.

In addition, it is usually necessary to prepare and place a spare storage medium in case of a failure. Otherwise, at the time of failure, it will not be possible to recover to the duplicated state and data will be lost if the remaining cache fails. That is, there is a problem that a storage amount three times as large as the cache data is required for the cache for both the duplication and the backup.

The present invention has been made in view of the above points, and an object of the present invention is to make a cache only a dirty data and to make a clean data a single one without duplication. An object of the present invention is to provide an information storage device that can effectively use the area of a small-capacity high-speed storage medium.

[0010]

[0011]

An information storage device according to a first aspect of the present invention is an information storage medium for storing data from an external device , and first to third caches used as caches for the information storage medium. A memory, and the first and second
When data is stored in the first cache memory, the data from the information storage medium is first stored in one of the first and second cache memories.
And the data from the external device when storing the data in the first and second cache memories.
It is detected that a failure has occurred in the second processing means for storing the same data in the first and second cache memories and in one of the first and second cache memories. The detecting means and, when the detecting means detects that a failure has occurred in the cache memory, among the data stored in the failed cache memory, the data from the information storage medium. Reads the data from the information recording medium again and stores the data in the third cache memory, and the data from the external device is the same data stored in the cache memory in which no failure has occurred. Data is stored in a third cache memory.

[0012] Information storage apparatus according to claim 2, claim
The information storage device according 1, data of the cache memory where the failure has occurred - a recovery unit to recover the data in the third cache memory in blocks, and ordered list of the blocks in the order of access frequency, the and means for creating a list, said recovery means is characterized by recovering a high block access frequency by referring to the list.

[0013] Information storage apparatus according to claim 3, claim
1. The information storage device according to 1, further comprising a recovery unit for recovering the data of the cache memory in which the failure has occurred to the third cache memory in block units, and the recovery unit restores clean data from dirty data. wherein the Ru prioritize than repair.

[0014] Information storage apparatus according to claim 4, claim
The information storage device according to 1, having a recovery means for recovering the data of the key <br/> Yasshu memory where the failure has occurred in the third cache memory in blocks, said recovery means of the dirty data recovery Than repairing clean data
Is also characterized by prioritizing

[0015]

[0016]

[0017]

According to a first aspect of the present invention, the first information storage medium is used as a cache for storing data from an external device.
And when storing data in the second cache memory,
The data from the information storage medium is stored in one of the first and second cache memories by the first processing means, and the data is stored in the first and second cache memories. When storing the data, the second processing means stores the same data in the first and second cache memories, and detects that the cache memory has failed. If the third
Among the data stored in the cache memory in which the failure has occurred by the processing means, the data from the information storage medium is read again from the information recording medium and the data is stored in the third cache memory. As the data from the external device, the same data stored in the cache memory in which no failure has occurred is stored in the third cache memory.

According to the second aspect of the present invention, the restoration of the data to the third cache memory is performed from the block having a high access frequency. In claim 3 , the restoration means gives priority to the restoration of clean data over the restoration of dirty data.

In claim 4 , the restoration means gives priority to the restoration of dirty data over the restoration of clean data .

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An information storage device according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a system configuration diagram of an information storage device. In FIG. 1, reference numeral 1 is a host I / O (input / output) unit for exchanging data with a host (not shown) to which the device of the present invention is connected. This I / O department
A buffer 2 is provided in 1 for temporarily storing the read or write data with the host.

Further, 3 is a data cache unit comprising caches 3a, 3b, 3c of a high speed storage medium such as three magnetic storage disks. Here, cache numbers 0, 1, and 2 are assigned to the caches 3a, 3b, and 3c, respectively.

Reference numeral 4 is a secondary storage device as a highly reliable low-speed large-capacity storage device in which a plurality of low-speed large-capacity media such as MO are used and a redundant code is added to data to improve reliability. Here, a media number is attached to the MO, and the block address of the secondary storage device 4 is such that the higher order means the media number and the lower order shows the block address in the medium. Then, when the host makes an access, it is designated by the block address of the secondary storage device 4. Further, in this embodiment, the block size of the secondary storage device 4 and the block size of the cache are the same, but there is no problem if one is a positive multiple of the other.

Further, reference numeral 6 is a control unit for integrally controlling this apparatus. The control unit 6 includes a CPU 7, ROM8, RA
It comprises M9, a management table 10, a counter 11, a cache status table 12, a repair mode 13, and a failure occurrence flag 14.

The control unit 6 reads commands from the host I / O unit 1, transfers data between the host I / O unit 1, the cache 3 and the secondary storage device 4, manages the cache 3,
Performs failure recovery processing for cache 3.

The CPU 7 executes a program stored in the ROM 8 and controls the entire device. RAM9 is an area for executing a program. Management table
Reference numeral 10 is a table for managing the cache 3 (consisting of two tables of management table number 0 and management table number 1). The counter 11 is a counter that counts the number of accesses to each cache 3a, 3b, 3c. The cache status table 12 is a table for storing the states of the caches 3a, 3b, 3c. Repair mode storage unit 13
Is a memory area for storing the three set repair modes (reliability-oriented, performance-oriented, and automatic).

The management table 10, the counter 11, the cache status table 12, the repair mode 13, and the failure flag 14 are the NVM backed up by the battery 15.
There is on (Nonvolatile Memory).

Next, the cache status table 12 will be described in detail with reference to FIG. The cache status table 12 has a management table number 22, a phase 23 and a mode 24 for each of the caches 3a, 3b and 3c.

First, the management table number 22 is the cache
Indicates the number of the management table that stores and manages the information of 3a, 3b, and 3c. Phase 22 has 5 caches: "in use, out of order, standby, recovering, no cache" 3
Takes the states of a, 3b, and 3c. Here, "in use" means that the corresponding medium is used as a cache, "in failure" means that the medium has a failure and is not in use, and "standby" means that it is for failure recovery. Yes Usage is stopped, "Recovering" means that data is being recovered to replace the failed cache from the reserved state (transition to in-use when recovery is complete), "Cache" “None” indicates that there is no medium corresponding to the system when the medium in failure is being replaced.

The mode 24 has four modes of "clean only, dirty only, mixed, prohibited use". Here, "clean only" means storing only clean data (data can be restored from secondary storage even if a cache failure occurs), and "dirty only" means dirty data (when the cache fails. If it happens,
Since data cannot be recovered from secondary storage, it is necessary to store only that data in another cache (duplication), and "mixed" means to store both. ing.

When both phases of the cache having the same management table number are in use, the two caches are all duplicated, and the data of the block having the same block address is duplicated (same as the conventional duplicated). Is).

If two caches in use have different management table numbers, not all the data on these caches are duplicated,
Blocks that are duplicated and blocks that are not duplicated coexist in each block.

A storage medium whose phase 23 is "restoring" or "reserved" is called a reserve cache. According to the cache status table of FIG.
The cache 3a with the cache number 0 is managed by the management table 0, is in the "mixed" mode, and has no abnormality in the medium.

Further, the cache 3b having the quiche number 1 is managed by the management table 1, is in the "mixed" mode, and there is no abnormality in the medium. Next, the management table 10 will be described in detail with reference to FIG. The management table 10 is composed of two management tables, a management table 30a having a management table number of 0 and a management table 30b having a management table number of 1. One stage of the management table 10 corresponds to one cache block and is arranged in the order of block addresses. As a result, the block address is associated with the stage in the management table.

The secondary storage block address 31 is an address on the secondary storage device 4 when storing cache data in the secondary storage device 4. Here, when the secondary storage block address is "-1", it indicates that the cache block of the block address corresponding to this stage is not used (called an empty block). At the time of initialization, all the secondary storage block addresses 31 in the table are set to "-1".

The backup block address 32
Stores "-1" if it is not a duplicated block,
In the case of a duplicated block, the block address of the duplicated partner block is stored. In the cache method, it is necessary to manage whether the data on the cache is clean data or dirty data.However, in the present invention, clean blocks are not duplicated and dirty blocks are always duplicated. By the backup block address 32 of this management table, it can be managed.

When the dirty block is cached out, the data of this block needs to be copied to the secondary storage device 4, but when the clean block is cached out, the data is copied to the secondary storage device 4. Since the same data as the block remains, there is no need to copy it.

Further, the access frequency list 34 connects all the data on the caches in the order of high access frequency, and the previous connection block 33 manages the block one order before (the access frequency is high). The table management table number 33a and block address 33b are shown, and the post-connection block 34 is one after the other (access frequency is low).
It represents the management table number 34a and the block address 34b of the management table that manages the block. Here, the blocks to be cashed out are selected in the order of low access frequency. In this embodiment, the access frequency is set to LRU (Least Rece
ntly Used) is used. LRU is a method in which the block with the oldest access time is the one with a low access frequency.

The status 36 is "data invalid".
(Data is invalid due to cache failure),
It takes three values: "redundant data invalid" (failure in the corresponding duplicated block) and "no error".

Next, the access frequency list information 40 will be described in detail with reference to FIG. In FIG. 4, the access frequency list information 40 stores the table number and the block address of the head of the access frequency, that is, the block with the highest access frequency, and the access block number 41, which is the head of the access frequency list. The table number and block address of the access frequency end block number 42 are stored. At the time of initialization, “−1” is stored in the access frequency list top block number 41 and the access frequency end block number 42. (It is -1 when there is no data in the cache.) Next, the management table (table number 0) and the management table (table number 1) in FIG. 3 are shown. The specific content will be described. As described above, according to the cache status table 12, the management table with the table number 0 and the management table with the table number 1 are cache 0 respectively.
And a table corresponding to cache 1. Since the management table manages one cache at a time, when duplicated, the blocks in cache 0 and cache 1 are duplicated.

According to the management table 30a, the cache 0
The block having the block address 5 is the data stored in the block address 100 of the medium having the medium number 0, and the backup block address is "-1". Therefore, it is not a duplicated block and there is no abnormality. Further, the block with the block address 10 of the cache 0 is the block address 20 of the medium with the media number 1.
The data that is stored in the second block is the block of the block address 30 of the cache 1 corresponding to the other table, and there is no abnormality. Further, the block of the block address 50 of the cache 0 is the data stored in the block address 15 of the media number 4, is not a duplicated block, and has no abnormality.

According to the management table 30b, the block of the block address 10 of the cache 1 is the data stored in the block address 10 of the medium number 2, is not a duplicated block, and has no abnormality. Further, the block of the block address 30 of the cache 1 is the data stored in the block of the block address 20 of the medium of the medium number 1, and the duplicated block of the block address 10 of the cache 0 corresponding to the other table. It is a block and there is no abnormality.

When the blocks with high access frequency are arranged in order, the access frequency list information 40 and management table shown in FIG.
From blocks 30a and 30b, block of block address 5 of cache 0 → block of block address 10 of cache 1 → block of block address 10 of cache 0 → block of block address 100 of cache 0 → ... block address of cache 0 400 Block → the block of cache 0 has a block address of 50. That is, when a block is flushed from the cache, the block of the block address 50 of the cache 0 indicated by the table number of the access frequency end block number 42 and the block address is flushed first.

Next, the operation of the embodiment of the present invention constructed as above will be described. Below, "Lead",
The "write", "processing at the time of failure", "block recovery", and "recovery" processing will be described with reference to FIGS.

First, the data read from the secondary storage device 4 is read.
The read (read) processing will be described. The host (not shown) to which this storage device is connected is the host I / O unit 1
The read process is started by sending a read command to the. The read command includes the secondary storage block address of the first block of the data to be read and the data length (block unit). Upon receiving this command, the host I / O unit 1 sends a signal with a command to the control unit 6, and the control unit 6 receives the read command from the I / O unit (ST501).

Then, the control unit 6 takes out the secondary storage block address of the received head block and stores it in the RAM 9 as a read block address, and adds the data length to the head block address to obtain the final block address and stores it in the RAM 9. Yes (ST502).

Next, it is checked whether the data indicated by the secondary storage block address hits the cache 3 (ST503). Therefore, the management tables 30a, 30b
Is searched for in the secondary storage block address.
If it exists, it exists in the cache 3, and the block address in the cache 3 and the cache number of the cache 3 storing this data are stored in the RAM 9,
If it does not exist in cache 3, S
Branch to T507 and secure a block on cache 3. The block reservation on the cache 3 will be described later with reference to FIG.

Here, the block address is the number of rows in the table in which the secondary storage block address is found, and the cache number is obtained from the cache status table. There may be two cache numbers instead of just one.

If it is in the cache 3, it is checked from the status of the management table whether there is a block failure (S
T504). If the data is invalid or the duplicated data is invalid, the block is restored (ST50
From 5), the process proceeds to ST506. The method of block recovery will be described with reference to FIG.

By the way, if the cache 3 does not exist in ST 503, the block allocation processing is performed by setting the argument to "-1".
The cache number and the block address are obtained in the RAM 9 (ST507), and in ST508, the data specified by the secondary storage block address is copied to the cache 3 secured from the secondary storage device 4, and the secured block is secured. ST506 stores the secondary storage block address in the management table corresponding to
Proceed to.

In ST506, the data obtained in the cache 3 is transferred to the I / O unit 1 and the read block is inserted at the head of the access frequency list 34, so that the block is designated by the access frequency list head block number 41. The table number and block address of the read block of cache 3 are stored in the front connection of the management table 10 corresponding to the block, and the access frequency list is stored in the rear connection of the management table corresponding to this read block of cache 3. The block number of the first block number and the block address are stored. Then, it is determined whether the transferred block is the final block by comparing it with the final block address as the read block address on the RAM 9 (ST50).
9).

When it is determined that the block is not the last block in ST509, the read block address on the RAM 9 is incremented by "1" (ST510), the process returns to ST503, and the processes after ST503 are repeated.

In this ST509, "is the final block"
If it is determined, the host I / O unit 1 is notified of the transfer end. In response to this notification, the host I / O unit 1 transfers the data sent from the cache 3 to the host (not shown) under the control of the control unit 6. And the control unit 6
When the transfer end is received from, the transfer end is transmitted to the host (not shown). In this way, the read processing of the secondary storage device 4 is performed.

Next, the writing (writing) process to the secondary storage device 4 will be described. A host (not shown) to which the storage device is connected sends a write command to the host I / O unit 1 to start the write processing.

Here, the write command includes the secondary storage block address of the leading block of the data to be written and the data length (block unit). When the host I / O unit 1 receives the write command, the control unit 6
To the I / O unit 1, and the control unit 6 receives the write command from the I / O unit 1 (ST601).

The control unit 6 takes out the secondary storage block address of the received head block and stores it in the RAM 9 as a read block address, and adds the data length to the head block address to obtain the final block address RA
It is stored in M9 (ST602).

Next, it is checked whether the data indicated by the secondary storage block address hits the cache 3 (ST
603). Therefore, the management table is searched for this block address. If it exists, it exists in the cache 3, and the block address in the cache 3 and the cache number of the cache 3 that stores this data are stored in the RAM. If not, it does not exist in the cache 3. , ST607 secures a cache. The securing of this cache will be described later with reference to the flowchart of FIG.

By the way, the block address is the number of rows of the table in which the secondary storage block address is found,
The cache number is the cache status table 12
Ask from. There may be two cache numbers instead of just one.

If it is in the cache 3, whether the block is a duplicated block from the backup block address of the management table (if the backup block address is not "-1", the block is duplicated and the backup block address is "-1"). , It is not duplicated), and it is checked whether the cache is clean only (ST604).

If it is determined in ST604 that the data is not duplicated (that is, clean), the data becomes dirty due to the write, so that the cache 3 is duplicated in preparation for a failure, the other table is not the cache hit table number. The reservation process is called by using the table number of (1) as an argument (ST605), and the process proceeds to ST606.

If the cache of the hit table number is clean only, the block is set as an empty block, removed from the access frequency list, and ST
Proceed to 607. The method of removing from the list is the same as the method described in Securing blocks.

If the cache is not found in ST603,
In ST607, the cache is secured as follows. To find the caches that are write-enabled, look for caches with dirty or mixed modes in the cache status table 12. It should be possible to find two corresponding caches. If the caches have the same table number, the table number is used as an argument to call the securing process, and if the table numbers are different, the two table numbers are used as arguments. The securing process is performed twice (ST607). The secured cache number and cache address are RA
It is stored in M9. After securing, proceed to ST606.

In ST606, data is transferred from the host I / O unit 1 to both of the two secured caches, and the written block is inserted at the beginning of the access frequency list 34, so that it is indicated by the access frequency list top block number. To the connection of the management table corresponding to the block
Stores the table number and block address of the written cache block, stores the block number and block address of the access frequency list head block number in the back connection of the management table corresponding to this written cache block, and stores it in that cache block. Set the status of the corresponding management table to normal.

After the transfer, whether or not the transferred block is the final block is determined in the same manner as the read (ST60
8) If not final, add "1" to the read block address and return to ST603. If it is the final block, the host I / O unit 1 is notified of the end of transfer. Further, the host I / O unit 1 buffers the data sent from the host, transfers the data to the cache under the control of the control unit 6, receives the transfer end from the control unit 6, and
Send the end of transfer to the host. The write processing to the secondary storage device 4 is completed as described above.

Next, FIG. 7 shows how to secure the cache area.
Will be described with reference to. This area reservation is first called with the table number as an argument and the processing is started (ST7).
01). If the block can be secured from which cache, "-1" is used as the argument value.

First, in ST701, an empty block (block whose secondary storage block address is "-1") is searched from the management table designated by the table number. If the argument is "-1", a free block is searched for in a table with a small counter value, and if not, a free block is searched for in another table. If there is an empty block, ST706
Then, store the table number and block address in RAM9 and return.
The procedure proceeds to ST702 with the bull number as an argument.

When it is determined in ST702 that there is no free space, the block to be cached out is determined in ST703 below. The management table is searched from the number indicated by the access frequency end block 42 in the access frequency list information in the order of lowest access frequency, and the block with the table number specified by the first argument found is the victim block (exited from the cache). Block). The block address and cache number of the victim block are read out to RAM9. The cache number is obtained from the cache status table 12 as the cache block number having the table number of the victim block and being used or restored by the phase. There are one or two cache numbers. Even if there are two cache numbers, since the two caches are duplicated as a whole, the block addresses are the same and only one block address is required.

The process of removing the victim block based on the access frequency will be described below. If this victim block is at the end of the list, the table number and block address of the previous connection block (the second least frequently accessed block) of the management table corresponding to the victim block are stored in the access frequency end number, and the previous connection block is stored. Write "-1" in the connection block after the management table corresponding to. If the victim block is the access frequency top block, the table number and block address of the subsequent connection block (block with the second highest access frequency) of the management table corresponding to the victim block are stored in the access frequency top number, and the subsequent connection block is stored. Write "-1" in the previous connection block of the management table corresponding to. If the victim block is neither the end nor the beginning, copy the value of the back connection block of the management table corresponding to the victim block to the back connection block of the management table corresponding to the front connection block of the management table corresponding to the victim block, The value of the front connection block corresponding to the victim block is copied to the front connection block of the management table corresponding to the rear connection block of the management table corresponding to the victim block.

If the block is a duplicated block by the operation of this access frequency list, the contents of the access frequency list are copied to the duplicated block. Next, it is determined whether the determined victim block is a duplicate block from the backup block address of the management table corresponding to the victim block (ST704),
If the block is a duplex block, the process proceeds to ST705, and if it is not the duplex block, the process proceeds to ST706.

Since the duplicated block is dirty, the data of the victim block is copied to the block of the secondary storage designated by the secondary storage block address of the management table 10, and the management corresponding to the victim block and the duplicated block after copying is performed. The two secondary storage block addresses in the table are set to "-1". Then, it progresses to ST706.
In ST706, the routine for securing this cash is copied.
Return to the processing of the flow that has been executed. As described above, the cache block is secured.

Next, the processing when a failure occurs will be described with reference to FIG. In FIG. 8, if a failure occurs during access to cache 3, cache 3
Then, an interrupt occurs in the control unit 6, and the control unit 6 detects a cache failure and identifies the cache number in which the failure occurred. Then, the phase of this cache number is set to the failure state from the cache status table 12, the management table number is read, this number is stored in the cache status table reserved by the phase, and the phase is changed from reserve to recovery (ST802). ).

Next, if all the statuses in the management table of the table number being restored are invalidated and the modes of the failed cache number of the cache status table 12 are mixed, the duplicated block of the other table is The status of (blocks other than "-1" for the mirror) is made invalid (ST803). Then, the failure occurrence flag is set and the interrupt processing is ended (ST804).

Next, block recovery processing (ST5 in FIG. 5)
05 and ST1006a, ST1006b, S in FIG.
(Called at T1006c) will be described with reference to the flowchart of FIG. This block recovery is called by using the table number of the block to be recovered and the block address as arguments, and the CPU 7 of the control unit 6
Executed by. First, it is checked from the corresponding management table whether the block to be restored is duplicated (ST
902). If it is determined in ST902 that the block is not a duplicated block (the backup block address is "-1"), the data in the secondary storage block indicated by the secondary storage block address is copied to the reserve cache block ( ST903). The block address of the reserve cache is the same as the block address of the argument block.

On the other hand, if the determination in ST902 is "YES", that is, if it is determined to be a duplicated block (when the backup block address is not "-1"), the duplicated block is copied to the block of the reserve cache ( ST904). A cache with a duplicated block is obtained as follows. Check whether there is a cache in use by the phase that has the table number given in the argument from the cache status table, and if it exists, it is the cache containing the duplicated block, and if it does not exist, specify the other table number. The cache used by the mochi phase is a cache including duplicated blocks. The block address at the time of copying is determined as follows. The duplicated block that is the copy source is the value of the backup block address, and the block address of the reserve cache that is the copy destination is the block address given as an argument.

After copying, the status of the block in the management table indicated by the argument is made normal in ST905, and if the restored block is a duplicated block, the status in the other corresponding management table is also made normal. Then, in ST906, the process returns to the calling process,
ST505 of FIG. 5 and ST1006a, ST1 of FIG.
Return to 006b and ST1006c.

Next, the recovery process will be described with reference to FIG. First, when the failure occurrence flag is set, the control unit 6 executes a failure recovery process as described below. First, the repair mode 13 is read, and it is determined whether the repair mode is automatic determination, reliability-oriented, or performance-oriented.

If the repair mode 13 places importance on reliability, the process branches to ST1005a, if performance emphasizes, to ST1005b, and if it determines automatically, to ST1003. When the repair mode is automatic determination, the remaining life is obtained by subtracting the counter value of the cache in use from the counter value of the cache in which the failure has occurred (ST1003). This remaining life is RA
If the remaining life is shorter than the specified value of M9, ST1
005a, if the remaining life is more than the specified value, ST100b
(ST1004). This specified value is RO
It is stored in M8 and loaded from ROM8 when the power is turned on. The specified value on the RAM 9 can be changed by the host.

Then, in ST1005a, the host I /
If there is a request from the O section 1, the above-mentioned read / write processing is performed. Next, the list is traced from the top of the access frequency list, and the unrestored redundant block (dirty block)
If there is a table number and block address as arguments, call the block recovery process (Fig. 9) and proceed to the next step. If the end of the access frequency list is not found, the next Proceed to the step (ST1006a). When the search is performed up to the end of the access frequency list, the process proceeds to ST1005c.

In ST1005b, if there is a request from the host I / O unit 1, the above-mentioned read / write processing is performed.
Next, follow the list from the top of the access frequency list,
Search for a block (clean block) that is not an unrestored duplex block, call the block restoration process with the table number and block address if any, and proceed to the next step. If the end of the access frequency list is searched If not found, go to the next step (S
T1006b). ST when searching to the end of the list
Proceed to 1005c.

In ST1005c, if there is a request from the host I / O unit 1, the above-mentioned read / write processing is performed.
Next, follow the list from the top of the access frequency list,
Search the unrestored block and call the restoration process of the block with the table number and the block address if any, and proceed to the next step. If the end of the access frequency list is not found and the block is not found, go to the next step. Proceed (ST1006c). If the end of the list has been searched, the process proceeds to ST1008. In ST1008, the failure occurrence flag is reset.

Incidentally, ST10 of the flow chart of FIG.
The specified value of 03 can be changed by the host sending a specified value change command and the specified value. In this case, the control unit 6 reads the specified value from the host I / O unit 1 and stores it in the RAM 9. Similarly, in the repair mode 13, the host can change the repair mode 13 by transmitting the repair mode change command and the mode value.

[0081]

As described in detail above, according to the present invention, only dirty data is duplicated to improve reliability, clean data is not duplicated, and a large-capacity low-speed medium which is a secondary storage medium at the time of failure Read from and recover. This makes it possible to effectively use the capacity of the cache and prevent the data from being lost due to the failure of the cache.

Further, by recovering from a frequently accessed one at the time of recovery from a failure, the probability of a cache hit when an access is made from the host is increased, and even when a failure is recovered, the access to a low-speed medium can be seen at a high speed. It does not impair the cache function.

When high speed is important, clean data is prioritized in the order of restoration over dirty data. This is because the recovery of clean data requires a slow read from the secondary storage, and the recovery takes longer than the dirty that can be recovered by reading from the duplicated cache.

On the other hand, when importance is placed on reliability, restoration of dirty data is prioritized. This means that if a failure occurs again during failure recovery, the dirty data will be lost even if it is duplicated. Therefore, the dirty data can be restored first, and the time in a dangerous state can be shortened.

Further, the spare medium can be effectively used by using the spare storage medium as a cache of clean data and duplicating the two storage media as the cache of dirty data.

[Brief description of drawings]

FIG. 1 is a block diagram showing a system configuration of an information storage device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a cache status table according to the embodiment.

FIG. 3 relates to the embodiment, FIG. 3 (A) shows a management table (table number 0), and FIG. 3 (B) shows a management table (table).
FIG. 3 is a diagram showing the stored contents of Bull number 1).

FIG. 4 is a diagram showing an access frequency list according to the embodiment.

FIG. 5 is a flowchart showing a read process according to the embodiment.

FIG. 6 is a flowchart showing a write process according to the embodiment.

FIG. 7 is a flowchart showing a cache securing process according to the embodiment.

FIG. 8 is a flowchart showing a fault occurrence interrupt process according to the embodiment.

FIG. 9 is a flowchart showing a block restoration process according to the embodiment.

FIG. 10 is a flowchart showing a recovery process according to the embodiment.

[Explanation of symbols]

1 ... I / O section, 2 ... Buffer, 3 ... Quiche, 3a ... Cache, 3b ... Cache, 3c ... Cache, 4 ... Secondary storage device, 6 ... Control section, 7 ... CPU, 8 ... ROM, 9 ... R
AM, 10 ... Management table, 11 ... Counter, 12 ... Cache status table, 13 ... Repair mode, 14 ... Failure flag, 15 ... Battery, 21 ... Management table number, 22 ... Phase, 24 ... Mode, 31 ... Secondary block address.

Claims (4)

(57) [Claims] 1. An information storage for storing data from an external device.
And憶媒body, first乃used as a cache of the information storage medium
Data is stored in the third to third cache memories and the first and second cache memories.
When the data is read from the information storage medium,
And one cache memo in the second cache memory
First processing means for storing in memory and data in the first and second cache memories
Data from the external device,
A second cache memory that stores the same data
Processing means and one of the first and second cache memories.
Detecting means to detect the failure of the cache memory
And the detection means causes a failure in the cache memory.
Failed cache when it is detected that
Among the data stored in the memory, the information storage medium
The data from the body is read again from the information recording medium.
Storing data in the third cache memory,
The data from the external device is stored in the cache that has not failed.
The same data stored in memory is stored in the third cache.
An information storage device comprising: a third processing unit that stores the data in a memory. 2. The cache memory in which the failure has occurred
Data is transferred in blocks to the third cache memory.
The recovery means includes a recovery means for recovering, a list in which the blocks are arranged in the order of access frequency, and a means for creating the list, and the recovery means refers to the list to increase the access frequency.
2. The block is restored from a block that is not present.
Of the information storage device. 3. The cache memory in which the failure has occurred
Data is transferred in blocks to the third cache memory.
The recovery means has a recovery means for recovery, and the recovery means restores clean data to dirty data.
2. The repair is prioritized over the repair of the
Of the information storage device. 4. The cache memory in which the failure has occurred
The third cache memory data in block units of
It has a recovery means to recover, and the recovery means cleans dirty data to clean data.
2. The repair is prioritized over the repair of the
Of the information storage device.