JPH09330277A - Service interruption processing system for disk cache system and method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a disk controller by which processing such as write back to be performed in the case of service interruption can be efficiently executed and the power of a power source for backup at the time of service interruption can be saved. SOLUTION: A disk controller 10 has a resource manager 113 for processing the save of data into a flash memory 15 by managing the attribute of data to be stored in a cache memory 14, a voltage check circuit 16 for monitoring the quantity of remaining power at a battery unit 18 and finding the quantity of saving power required for executing the data saving processing based on the amount of data having a write data attribute on the cache memory 14 at the time of power failure, and a switch circuit 17 for receiving an instruction from the voltage check circuit 16 and controlling power supply from the battery unit 18 to the resource manager 113 and the flash memory 15. At the time when the remaining power amount of the battery unit 18 becomes less than the saving power amount in the case of service interruption, the data saving processing of only the data having the write data attribute is started.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data access system and a disk cache system in which a disk cache is arranged in the middle of data transfer. It also relates to power saving of the backup power supply.

[0002]

2. Description of the Related Art A data access system for accessing data stored in a magnetic disk device is generally shown in FIG.
3, a magnetic disk device (DKU: Disk Uni) as a data input / output device for storing data.
t) 30, and a disk controller 10 for controlling data access to the magnetic disk devices 30, which is provided between the host device 20 for accessing the magnetic disk devices 30 and the respective magnetic disk devices 30. There is.

Since the magnetic disk device 30 is accompanied by a mechanical operation, it is difficult to make the access time faster than millisecond, and there is a considerable difference from the access time of the semiconductor memory.

Therefore, there is a disk controller 10 having a built-in cache memory in order to improve the access speed. This is because the access from the higher-level device 20 tends to concentrate on specific data on the magnetic disk device 30, so that data with a high reference frequency is copied to the cache memory, and a request to access the copied data again is made. If there is, the data is read from the cache memory and transferred to the host device 20 instead of directly accessing the data on the magnetic disk device 30. The cache memory used for this disk access is generally called a disk cache. A data access system using this disk cache will be referred to as a disk cache system.

In the disk control device 10 in the disk cache system, the write data from the host device 20 to the magnetic disk device 30 is not directly written to the magnetic disk device 30, but is written to the cache memory at the time of writing to the host device 20. Signals the end of write processing. Therefore, a backup power source such as a battery is provided to prevent problems such as data loss that may occur when the primary commercial power source is cut off due to a power failure or the like.

For battery backup, a method of backing up both the disk controller 10 and the magnetic disk device 30 by an externally connected uninterruptible power supply (UPS), and a battery unit provided inside the disk controller 10 are provided. There is a method of backing up only the cache memory with a battery.

FIG. 14 is a block diagram of a conventional disk control device adopting the former method. In addition to the disk device 10, the host device 20, and the magnetic disk device (DKU) 30 that are provided in the same manner as in FIG. 13, FIG. 14 shows disk control in case the primary commercial power supply is cut off due to a power failure or the like. An uninterruptible power supply (U) having a built-in battery for supplying power to the device 10 and the magnetic disk device 30 for a certain time.
PS) 40 is provided. Inside the disk controller 10, a channel adapter (CAP) 11 that transmits and receives data to and from the host device 20 and a device adapter (CDP) that controls data transmission and reception to and from the magnetic disk device 30.
12, cache memory management, hit / miss hit determination, cache memory and magnetic disk device 30
Resource manager 13 that controls the data transfer of
A cache memory 14 configured by a semiconductor memory for temporarily storing transfer data between the host device 20 and the magnetic disk device 30 is provided.

FIG. 15 is a block diagram of a conventional disk controller which adopts the latter battery backup method. 15, as in FIG. 13, the disk device 10, the host device 20 and the magnetic disk device (DKU).
30 are provided. In addition to the channel adapter 11, the device adapter 12, the resource manager 13 and the cache memory 14 similar to the disk controller 10 described above, the primary commercial power supply is cut off due to a power failure or the like inside the disk controller 10. In preparation for a case, a battery unit 18 connected to the cache memory 14 for holding the contents of the cache memory 14 in the disk control device 10 is provided.

Next, the operation common to both methods in the conventional disk control device will be described.

In the configuration of FIG. 14, when the host device 20 requests the disk controller 10 to write data to the magnetic disk device 30, the resource manager 13 of the disk controller 10 receives the request and the host device 20 Two
The write data from 0 is taken in via the channel adapter 11. The captured data is stored in the magnetic disk device 30.
It is written only to the cache memory 14 without actually being written directly to the cache memory 14. At this time, the attribute of the written data becomes the attribute of the write data and is recorded / managed in the resource manager 13 as attribute information. The resource manager 13 immediately notifies the upper device 20 of the end of the transfer when the writing of the requested data to the cache memory 14 is completed. As a result, it is possible to significantly reduce the transfer processing time as compared with the case where data is directly written to the magnetic disk device 30.

The data written in the cache memory 14 is actually written back to the magnetic disk device 30 under the control of the resource manager 13 in the ascending or descending order of the access frequency synchronously or asynchronously with the access of the host device 20.
Further, the attribute of the data recorded inside the resource manager 13 is changed from the write data attribute to the read data attribute. This write back operation is called write back processing.

When there is a data read request from the host device 20, the resource manager 13 first searches the cache memory 14 for the data requested to be read. If the data does not exist in the cache memory 14, the data is actually read from the magnetic disk device 30 via the device adapter 12 and written in the cache memory 14, and at the same time, the data is written to the host device 20 via the channel adapter 11. Forward. The attribute of the data written in the cache memory 14 becomes the attribute of the read data and is recorded / managed inside the resource manager 13 as attribute information. In response to a data read request from the host device 20, if the corresponding data exists in the cache memory 14, the data is read from the cache memory 14 instead of from the magnetic disk device 30, and the data is read via the channel adapter 1. And transfers it to the host device 20.

As described above, since the data once transferred through the disk control device 10 is written in the cache memory 14, the data read from the second time onward is
As in the case of write data, the direct magnetic disk drive 30
It is possible to realize a significant reduction in transfer processing time as compared with the case of reading from.

The above is the operation for basic data access in the write cache system, and the operation up to this point is the same as in the configuration of FIG.

Next, a conventional operation when a power failure occurs in the above configuration will be described. In the configuration shown in FIG. 14, when the primary commercial power supply is stopped (power failure) due to some reason, an uninterruptible power supply (UPS) 40
Notifies the disk controller 10 that a power failure has occurred, and the battery built in the uninterruptible power supply 40 causes the disk controller 10 to replace the primary commercial power source.
Then, the same electric power supply as before the supply of the primary commercial power supply was stopped is started to the magnetic disk device 30. Upon receiving the power failure notification, the resource manager 13 of the disk controller 10 writes back all the data of the write data attribute written in the cache memory 14 to the magnetic disk device 30. This operation is called shutdown processing.
Then, when there is no write data attribute data in the cache memory 14, the shutdown process ends,
The magnetic disk control device 10 is in an idle state, then the power supply from the uninterruptible power supply (UPS) 40 is cut off, and the primary commercial power supply is on standby.

On the other hand, in the configuration shown in FIG. 15, when the primary commercial power supply is stopped (power failure) for some reason, the power supply to the magnetic disk device 30 is stopped and the write back is performed. No processing is performed, and only the contents of the cache memory 14 are supplied with power from the battery unit 18 and enter the data holding mode.

In order to improve the system performance, one means is to improve the probability that the data to be read by the host device exists in the cache memory 14 (this is called the hit ratio). However, for this reason, the capacity of the cache memory 14 generally requires a considerably large capacity of about 1/100 to 1/10 of the capacity of the magnetic disk device 30. It is considered that the capacity of the cache memory will further increase as the capacity of the magnetic disk device increases in the future.

[0018]

However, FIG.
In the conventional configuration shown in, as the write data capacity of the cache memory is increased in order to improve the system performance, the data having the write data attribute written in the cache memory at the time of power failure is transferred to the magnetic disk device. Since the write-back time becomes long, a large and expensive uninterruptible power supply device with a large battery capacity and a long backup time is required accordingly.

Further, the backup operation is started immediately after a power failure, but in the event of a power failure such as an instantaneous power failure, this backup processing will be continued even after the power is restored, and the original system processing will be immediately started. There was a problem that I could not. As the capacity of the cache memory increases, the data having the write data attribute also increases, so this problem is considered to become more prominent.

Although not shown, there is also a method of using a flash memory which does not require a battery backup and consumes less power during operation as a target device for write-back during power failure instead of a magnetic disk device. It is difficult to use the conventional device configuration as it is in an environment where the number of times of writing is limited and power failure occurs frequently.

Further, in the conventional structure shown in FIG. 15, a battery having a relatively small capacity can be used as compared with FIG. 14, but the entire cache memory is backed up regardless of the amount of data that needs to be backed up. However, there is a problem that the battery is wasted. It is considered that this problem will become more significant as the capacity of the cache memory increases.

The present invention has been made to solve the above problems, and an object thereof is to efficiently perform processing such as write-back performed when a power failure occurs and to provide a backup power source at the time of power failure. Another object of the present invention is to provide a disk control device that saves power.

[0023]

In order to achieve the above object, a power failure processing method in a disk cache system according to the present invention is a disk device for storing data and a host device for accessing the disk device. In a disk cache system having a disk cache for temporarily storing data transferred between the disk cache and backup power supply means for temporarily supplying electric power to at least the disk cache at the time of power failure, the disk cache at the time of power failure A non-volatile storage means to which the data stored in the disk cache is written; a data attribute management means for managing the attributes of the data stored in the disk cache; Write data attributes Data saving means for saving data to the nonvolatile storage means, and power amount monitoring means for receiving power from the backup power supply means and monitoring the remaining power amount of the backup power supply means The data saving means starts the data saving processing according to the remaining power amount of the backup power supply means at the time of power failure.

[0024] Further, there is further provided a power saving amount calculating means for calculating a power saving amount necessary for executing the data saving process based on the number of write data attribute data stored in the disk cache at the time of power failure, The data saving means is characterized in that the data saving processing is started when the remaining power amount of the backup power source means becomes equal to or less than the saving power amount during a power failure.

Furthermore, at the time of power failure, when the remaining power amount of the backup power source means becomes equal to or less than the save power amount, the power from the backup power supply means is transferred to the data save means, the data attribute managing means, and the power saving means. It is characterized in that it has a switch means for supplying to the non-volatile storage means.

Further, there is further provided matrix switch means for supplying the power from the backup power supply means in block units to the disk cache in which the storage area is divided into elements in block units, and the matrix switch means has a power failure. The power from the backup power supply means is supplied only to the block of the disk cache that stores the data having the write data attribute.

The non-volatile storage means is a flash memory.

Further, a power failure processing method in a disk cache system according to the present invention is a disk for temporarily storing data transferred between a disk device for storing data and a host device for accessing the disk device. A disk cache system comprising: a cache; backup power supply means for temporarily supplying power to the disk cache at the time of power failure; and non-volatile storage means for writing data stored in the disk cache at the time of power failure When a power failure occurs, based on the number of data of the write data attribute that is not actually written in the disk device and is stored only in the disk cache, the data saving process of the data to the nonvolatile storage means is executed. The required amount of power to save In the non-volatile storage means, only the data having the write data attribute stored in the disk cache when the power consumption for backup becomes equal to or less than the power consumption for backup according to the remaining power of the backup power supply means. And a data saving step of saving the data.

[0029]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention will be described below with reference to the drawings. The same elements as those in the conventional example are designated by the same reference numerals.

Embodiment 1. FIG. 1 is a configuration diagram showing an embodiment of a power failure processing method in a disk cache system according to the present invention. In this embodiment, a method of providing a battery unit inside the disk control device 10 and performing battery backup is used. The disk control device 10 includes a channel adapter 11, a device adapter 12, a battery unit 18 as a backup power supply means for temporarily supplying power to at least the cache memory 14 in the event of a power failure, a host device 20 and a magnetic disk as in the conventional case. It has a cache memory 14 which is a disk cache for temporarily storing data transferred to and from the device 30. The disk controller 10 according to the present embodiment further includes a resource manager 1
13, a flash memory 15, a voltage check circuit 16 and a switch circuit 17 are provided. The flash memory 15 is a non-volatile storage means in which the data stored in the cache memory 14 is written for saving in case of power failure for a long period of time in case of power failure. The resource manager 113 is a data attribute management unit that manages the attributes of the data stored in the cache memory 14, and performs a data save process for the data of the write data attribute stored in the cache memory 14 to the flash memory 15. It is a data saving means. Of course, it also has the same functions as the conventional cache memory management and data transfer control to the magnetic disk device 30. Voltage check circuit 16
Is an electric power amount monitoring means that receives power from the battery unit 18 and operates even during a power failure and monitors the remaining power amount of the battery unit 18, and write data attribute data stored in the cache memory 14 during a power failure. It operates as a saved power amount calculation means for obtaining the saved power amount necessary to execute the data save process based on the number of
Instruct to save data. The switch circuit 17 receives an instruction from the voltage check circuit 16 and the resource manager 1
Battery unit 1 to 13 and flash memory 15
It is a switch means for controlling the power supply from 8.

A feature of this embodiment is that when the remaining power amount of the battery unit 18 is measured at the time of power failure and the remaining power amount becomes equal to or less than the saving power amount necessary for executing the data saving process. It is to start the data saving process. That is, since the data saving process, that is, the backup of the contents of the cache memory 14 is not immediately started at the time of the power failure, it is possible to immediately shift to the original system processing even in the case of the momentary power failure.

Next, regarding the operation of the present embodiment, the flowchart of FIG. 2 showing the basic operation of the disk controller 10 and the disk controller 10 at the time of detecting a power failure.
The operation will be described with reference to the flowchart of FIG.

First, referring to FIG. 2, the disk controller 1
When there is no access request from the host device 20, 0 stands by in an idle state in which the access request from the host device 20 can be accepted (step 100). Host device 20
When an access request is received from (step 10
1), the content is analyzed, and the resource manager 113
Is the cache memory 14 according to the content of the access request.
A read process or a write process is executed for (step 102).

At this time, the resource manager 113 calculates the number of write data attributes of the data stored in the cache memory 14 (step 103), but manages it using the data count table shown in FIG. When the content of the access request is a write request, when the cache determination result is a miss hit, or when the cache determination result is a hit and the data in the cache memory 14 is the data having the read data attribute, the resource manager 113 uses the resource. The data count number of the write data attribute built in the manager 113 is added (step 103). That is, when the cache determination result is a miss hit, the access target data is read from the magnetic disk device 30 and newly written in the cache memory 14, so that the data is written as a write data attribute. Therefore, in FIG. 3, the value of the miss hit in the case of the write processing is "+1", which means that the data count number is incremented by "+1" in the case of such a miss hit. When the cache determination result is a hit and the data in the cache memory 14 is the data having the read data attribute, the data attribute of the access target data is updated from the read data attribute to the write data attribute. Therefore, in FIG. 3, the value of the read data attribute in the case of the writing process is “+1”.
This means that when the access target data in the cache memory 14 has the read data attribute, the data count number is incremented by "+1". It should be noted that the data number of the read data attribute is not used in the present embodiment and therefore it is not necessary to count it, but if it is to be counted, the count number is subtracted.

When the data to be accessed in the cache memory 14 has the write data attribute, the write data attribute is maintained as it is, and therefore the data count number does not change. Therefore, in FIG. 3, the value of the write data attribute in the case of the writing process is "0". When the write back process is executed, the data count number of the write data attribute is decremented every time the data is written back. Therefore, in FIG. 3, the value of the write data attribute in the case of the write back process is "-1".
When there is no write data attribute data in the cache memory 14, the data count number becomes zero.

Next, the resource manager 113 records the data count of the write data attribute existing in the cache memory 14 in the voltage check circuit 16 (step 104), and then returns to the idle state again (step 10).
0), wait for the next access request from the host device 20.

When the power supply from the primary commercial power source is stopped during the above basic operation, the process shifts to the process at the time of power failure shown in FIG. In FIG. 4, when a power failure occurs (step 110), the battery unit 18 immediately starts supplying power to the cache memory 14 (step 111). As a result, even if the supply of the primary commercial power supply is stopped, the data in the cache memory 14 is held as it is for a certain period of time. FIG. 5 shows the cache memory 1.
The backup time and the battery unit 18 required for the save processing of the data having the write data attribute stored in FIG.
FIG. 4 is a diagram showing the relationship between the discharge voltage of the battery unit 18 and the discharge voltage of the battery unit 18 in the present embodiment without immediately performing data saving processing even if a power failure occurs. While falling to the save start voltage Vs (Tmemory), even if the supply of the primary commercial power supply is stopped, all the data in the cache memory 14 is held as it is. The value of the escape start voltage Vs will be described later.

At this time, the voltage check circuit 16 determines whether the primary commercial power supply is restored or not by the battery unit 18
The power supply terminal voltage to
2), as shown in FIG. 6, after a power failure occurs, the above Tmemo
ry and the time required from power failure to power recovery (Tof
If the primary commercial power source is restored while the relationship of Tmemory> Toff is established with the f), the data saving and the initialization processing of the cache memory 14 are not performed and the step 100 in FIG. (Step 130), the access is immediately accepted from the host device 20.

On the other hand, when the primary commercial power source is not restored, the voltage check circuit 16 causes the battery unit 18 to
Of the output voltage is measured, and it is monitored whether the voltage has dropped to the retract start voltage Vs shown in FIG. 5 (step 113). The save start voltage Vs is the flash memory 15 of all the data having the write data attribute written in the cache memory 14.
Is the limit voltage (saved electric energy) of the electric power sufficient to perform the shelter processing. Therefore, it is not necessary to start the data saving process while the output voltage of the battery unit 18 drops to the voltage Vs. In this embodiment, since the guaranteed limit voltage Ve is set, the value obtained by adding the guaranteed limit voltage Ve to the calculated value calculated based on the data of the write data attribute stored in the cache memory 14 is saved. The starting voltage is Vs. As described above, in the present embodiment, since the flash memory 15 is used as the data save destination of the cache memory 14, the write back time required for the data save processing can be made shorter than writing to the magnetic disk device 30. In addition, the amount of power required for the data saving process can be reduced.

Here, it is judged whether or not the output voltage of the battery unit 18 measured by the voltage check circuit 16 has dropped to the voltage Vs (step 114). If it has not dropped to the limit yet, the process returns to step 112. , Wait for restoration of primary commercial power. When the remaining power of the battery unit 18 drops to the voltage Vs, the voltage check circuit 16
Issues an activation notification to the switch circuit 17 to activate it (step 115).

The switch circuit 17 which has received the start notification starts power supply from the battery unit 18 to the resource manager 113 and the flash memory 15 (step 116). At this point resource manager 1
Since the power supply to 13 and the like is started, the power consumption of the battery unit 18 can be reduced.

The resource manager 113 whose power supply is restarted by the battery reads the data counter written in the voltage check circuit 16 in step 104 in FIG. 2, and whether there is write data attribute data to be saved in the cache memory 14. Is checked (step 117). If there is no data to be saved (data count is 0), the operation proceeds to step 120. If there is data to be saved (the data count is 1 or more), the data corresponding to the data count number is saved from the cache memory 14 to the flash memory 15 (step 118), and then the write data of the data is saved. The attribute is changed to the read data attribute (step 119).

When the saving of all the data having the write data attribute to the flash memory 15 is completed, the resource manager 113 issues a saving completion notice to the voltage check circuit 16 (step 120). Voltage check circuit 16
When the save completion notice is received from the resource manager 113, the resource manager 11
3 and the power supply to the flash memory 15 are stopped (step 121). Then, the disk controller 10 waits for restoration of the primary commercial power supply (step 122).

As described above, according to the present embodiment, even if a power failure occurs, the data saving process is not performed immediately and the discharge voltage of the battery unit 18 drops to the saving start voltage Vs (Tmemory). Holds the data in the cache memory 14 as it is until the supply of the primary commercial power is started. As a result, if the power is restored during the time Tmemory, it is possible to immediately shift to the original system processing. Of course, since the saving start voltage Vs is calculated based on the number of pieces of data having the write data attribute stored in the cache memory 14 at the time of power failure, the data saving process can be kept on standby for a long time, and the data protection is sure Can be done.

Further, since the data saving process is not performed immediately even if a power failure occurs, the number of times the data saving process is performed can be reduced, so that the flash memory 15 can be used. As a result, since only the data having the write data attribute is set as the backup target data, not only the time required for the backup can be shortened but also the power consumption of the battery unit 18 can be reduced. As a result, it is possible to prevent the backup power supply means such as the battery unit 18 from becoming large and expensive.

In the present embodiment, the case where the battery unit 18 is provided inside the disk control device 10 for battery backup has been described, but an uninterruptible power supply device externally connected as backup power supply means. (UPS) is provided to provide a disk controller 10
The method is also applicable to a method of backing up both the magnetic disk device 30 and the magnetic disk device 30. in this case,
Magnetic disk device 30 instead of flash memory 15
It becomes possible to use itself as a non-volatile storage means. In this case, the time required for the data saving process cannot be shortened and some power consumption is required, but the other effects can be obtained.

Embodiment 2. FIG. 7 is a configuration diagram showing another embodiment of the power failure processing method in the disk cache system according to the present invention. Also in this embodiment, a method of providing a battery unit inside the disk control device 10 to perform battery backup is used. The disk controller 10 includes a channel adapter 11, a device adapter 12, a flash memory 15, a voltage check circuit 16, and a switch circuit 1 similar to those of the first embodiment.
7 and a battery unit 18. The cache memory 214 in the present embodiment has a storage area in which elements are divided into blocks, and the disk controller 10 is a matrix that supplies the cache memory 214 with power from the battery unit 18 in blocks. Switch matrix 1 as switch means
9. Accordingly, the resource manager 21
3 manages the cache memory 214 in block units. FIG. 8 is a diagram showing the relationship between the cache memory 214 and the switch matrix 19. Switch units 19-1 and 19- included in the switch matrix 19
2, ..., 19-n are storage areas (cache memory blocks) 214-1, which are element-divided in block units.
214-2, ..., 214-n are provided respectively, and power is supplied from the primary commercial power source (normal power source) to the storage areas 214-1 to 214-n during normal operation to cause a power failure. At times, each built-in switch is opened or closed according to the switch signal sent from the resource manager 213 to control the supply of electric power from the battery unit 18.

A feature of this embodiment is that the cache memory 214 divided into blocks is provided with a switch matrix 19 for supplying the power from the battery unit 18 in blocks, and the write data attribute is provided in the event of a power failure. That is, the power from the battery unit 18 is supplied only to the cache memory blocks 214-1 to 214-n that store the data. As a result, the unnecessary power consumption of the battery unit 18 can be reduced, so that the consumption of the backup power supply means can be reduced and the data guarantee period can be increased. Alternatively, it is possible to guarantee the data of the write data attribute stored in the cache memory 214 for a long time even by using a smaller and cheaper backup power supply means.

Next, regarding the operation of this embodiment, the flowchart of FIG. 9 showing the basic operation of the disk controller 10, the flowchart of FIG. 10 showing the operation of the disk controller 10 when a power failure is detected, and the power recovery. The operation of the disk controller 10 at the time of detection will be described with reference to the flowchart of FIG.

First, referring to FIG. 9, the disk controller 1
When there is no access request from the host device 20, 0 stands by in an idle state in which the access request from the host device 20 can be accepted (step 200). Host device 20
When an access request is received from (step 20
1), analyze the contents, resource manager 213
Is the cache memory 21 according to the content of the access request.
A read process or a write process is executed for 4 (step 202). The processing up to this point is the same as in the first embodiment.
Is the same as

At this time, the resource manager 213 uses the access-specific switch setting table shown in FIG. 12 to determine the open / closed state of each of the switch units 19-1 to 19-n according to the access contents from the higher-level device 20 and control them. (Step 203). When the content of the access request is a write request, when the cache determination result is a miss hit, or when the cache determination result is a hit and the data in the cache memory 214 is the read data attribute data, the resource manager 213 writes the above. As described in the first embodiment, since the data attribute of the data to be accessed is the write data attribute, one of the cache memory blocks 2 in which the data to be accessed is written.
14-1 to 214-n corresponding to the switch unit 19-1 to
Turn on the switch signal to 19-n. Therefore, FIG.
In the case of the write processing in 2, the set value of the miss hit and the read data attribute is "switch ON". The other switch units 19-1 to 19-n are controlled so that the previous states are retained. For example, the writing process from the higher-level device 20 is performed by the cache memory block 2
When the data of the read data attribute 14-3 is hit, the attribute of the data is changed from the read data attribute to the write data attribute. Therefore, the resource manager 21
3 turns on the switch signal 3 to the switch unit 19-3. When the read process from the higher-level device 20 hits the cache memory block 214-3, the attribute of the cache memory 214 remains the read data attribute. Therefore, the resource manager 13 switches the switch unit 19-
The switch signal 3 to 3 is maintained in the previous state. Note that, as shown in FIG. 12, when the block writeback process is performed to perform the writeback process on the entire block, the write data is written to the cache memory blocks 214-1 to 214-n on which the block writeback process is performed. Since the attribute data does not exist at all, the resource manager 213 turns off the switch signal of the cache memory block. The data included in the cache memory blocks 214-1 to 214-n is set to 1
If write back processing is performed one by one, the resource manager 213 determines that the cache memory block 214-
The switch signal of the cache memory block is turned off when all the data having the write data attribute in 1 to 214-n have disappeared. In this way, the resource manager 213 controls the open / closed states of the switch units 19-1 to 19-n, and when the issuance of the switch signal is completed, the resource manager 213 returns to the idle state again (step 200), and the next higher level device 20 sends Waits for access request.

When the power supply from the primary commercial power source is stopped during the above basic operation, the processing shifts to the time of power failure shown in FIG. In FIG. 10, when a power failure occurs (step 210), the battery unit 18 immediately starts supplying power to the cache memory 214 (step 211). At this time, step 20 described above
In the processing of 3, the cache memory block 214-1
Battery power is supplied only to the cache memory block whose switch signal is controlled to be ON among 214-n (step 212). Cache memory blocks 214-1 to 214- that were not powered by the battery
n is either the read data attribute stored data or an unused block. Therefore, even if the data on the cache memory 214 is erased, the original data of the erased data is not lost because it is already stored in the magnetic disk device 30. It should be noted that each process in the present embodiment until the primary commercial power supply is restored is the same as the process between step 112 and step 122 in the above-described first embodiment, and therefore description thereof will be omitted. However, in the present embodiment, the cache memory 214 is divided into elements in cache memory block units, the cache memory blocks 214-1 to 214-n including the data having the write data attribute are detected, and only the cache memory block concerned is detected. Since the backup power is supplied from the battery unit 18, the power consumption of the battery unit 18 can be further reduced. That is, Tmemory shown in FIG.
Since the time of y can be made longer, there is a higher possibility that the data saving processing will not be executed, and the opportunities for using the flash memory 15 can be reduced.

As shown in FIG. 11, when the primary commercial power source is restored (step 220), the resource manager 2
13 reads the setting states of the respective switch units 19-1 to 19-n (step 221) and initializes only the cache memory blocks 214-1 to 214-n which were in the open state, that is, the switch signal was OFF. Do (Step 2
22). In this way, data can be protected for a long time without loss.

As described above, according to the present embodiment, the switch matrix 19 can back up only the cache memory block in which the data of the write data attribute exists in the cache memory 214, so that the battery unit 18 at the time of power failure can be backed up. The power consumption of can be significantly reduced. As a result, the disk control device 10 can be realized with the small-capacity and small battery unit 18.

The present embodiment is an uninterruptible power supply (U
Needless to say, the present invention is also applicable to the method of backing up a battery using PS).

[0056]

According to the present invention, even if the power supply from the primary commercial power supply is stopped for a short time for some reason, the data saving process is not immediately started and the contents of the disk cache are all left as they are. Since the data saving process is started according to the remaining power of the backup power supply means, it is possible to immediately return to system operation when the power supply from the primary commercial power supply is restarted. Becomes In particular, the data saving process is started when the remaining power amount of the backup power supply means falls to the saving start voltage calculated based on the number of write data attribute data stored in the disk cache at the time of power failure. Therefore, the data saving process can be kept on standby as long as possible, and the data protection can be surely performed.

Further, since the number of times the data saving process is performed can be greatly reduced, it is possible to use a flash memory having a limited number of times of writing as the non-volatile storage means.

Further, since only the data having the write data attribute stored in the disk cache is used as the backup target data, not only the time required for the backup can be shortened but also the power consumption of the backup power supply means required for the data saving processing can be reduced. Can also be reduced. As a result, it is possible to prevent the backup power supply means such as the battery unit 18 from becoming large and expensive. It is more effective to use a flash memory or the like that does not require battery backup and consumes less power during operation as the non-volatile storage means.

Further, the switch means is provided so that the power from the backup power supply means is supplied to the data save means and the like when the remaining power amount of the backup power supply means at the time of power failure becomes less than the save power amount. Therefore, the power consumption of the backup power supply means can be reduced.

Further, since the disk cache is divided into blocks and the power is supplied from the backup power supply means only to the block containing the data having the write data attribute at the time of power failure, the backup power supply means It is possible to significantly reduce power consumption. As a result, the time until the data saving process is started is extended, and the number of times the data saving process is executed is likely to be reduced. Therefore, the flash memory can be used less frequently. This also extends the life of the flash memory.

Further, since the power consumption of the backup power supply means can be greatly suppressed, it is possible to realize a disk cache system with a small capacity and a small backup power supply means.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a first embodiment of a power failure processing method in a disk cache system according to the present invention.

FIG. 2 is a flowchart showing a basic operation of the disk control device according to the first embodiment.

FIG. 3 is a diagram showing an example of contents of a data count table used in the first embodiment.

FIG. 4 is a flowchart showing the operation of the disk control device according to the first embodiment when a power failure is detected.

FIG. 5 is a diagram showing a relationship between a backup time required for a saving process of data having a write data attribute stored in a cache memory and a discharge voltage of a battery unit in the first embodiment.

FIG. 6 is a diagram showing a relationship between a backup time required for saving processing of data having a write data attribute stored in a cache memory and a discharge voltage of a battery unit when power is restored in a short time in the first embodiment. Is.

FIG. 7 is a configuration diagram showing a second embodiment of a power failure processing method in a disk cache system according to the present invention.

FIG. 8 is a diagram showing a relationship between a cache memory and a switch matrix according to the second embodiment.

FIG. 9 is a flowchart showing a basic operation of the disk control device according to the second embodiment.

FIG. 10 is a flowchart showing the operation of the disk control device according to the second embodiment when a power failure is detected.

FIG. 11 is a flowchart showing an operation of the disk control device according to the second embodiment when a power recovery is detected.

FIG. 12 is a diagram showing an example of contents of an access-specific switch setting table used in the second embodiment.

FIG. 13 is an overall configuration diagram of a general disk cache system.

FIG. 14 is a configuration diagram of a conventional disk control device that employs a method of battery-backing both the disk control device and the magnetic disk device by an uninterruptible power supply device externally connected.

FIG. 15 shows a battery unit provided inside the disk control device for only the cache memory.
It is a block diagram of the conventional disk control device which employs the backup method.

[Explanation of symbols]

10 disk controller, 11 channel adapter, 1
2 device adapter, 13, 113, 213 resource manager, 14, 214 cache memory, 1
5 Flash memory, 16 Voltage check circuit, 17
Switch circuit, 18 battery unit, 19 switch matrix, 19-1 to 19-n switch section, 2
0 host device, 30 magnetic disk device, 214-1 to 21
214-n cache memory block.

Claims (6)

[Claims] 1. A disk cache that temporarily stores data transferred between a disk device that stores data and a host device that accesses the disk device, and power is supplied to at least the disk cache when a power failure occurs. In a disk cache system having a backup power supply means that is temporarily supplied, a non-volatile storage means in which data stored in the disk cache is written in the event of a power failure, and management of attributes of data stored in the disk cache A data attribute management unit for performing a data saving process for performing a data saving process on the nonvolatile storage unit of the write data attribute data that is not actually written in the disk device and is stored only in the disk cache; Is the backup power supply means And a power amount monitoring unit that monitors the remaining power amount of the backup power supply unit by receiving power supply from the backup power supply unit, and the data saving unit determines the remaining power amount of the backup power supply unit during a power failure. A power outage processing method in a disk cache system characterized by starting data saving processing in response. 2. A saving power amount calculating means for calculating a saving power amount required for executing the data saving process based on the number of write data attribute data stored in the disk cache at the time of power failure, the data 2. The power failure processing method in a disk cache system according to claim 1, wherein the saving means starts the data saving processing when the remaining power amount of the backup power supply means becomes equal to or less than the saving power amount during a power failure. . 3. The power from the backup power supply means is supplied to the data saving means, the data attribute managing means, and the nonvolatile memory when the remaining power amount of the backup power source means becomes equal to or less than the save power amount during a power failure. 3. A power failure processing method in a disk cache system according to claim 2, further comprising a switch means for supplying the data to the property storage means. 4. There is provided matrix switch means for supplying the power from the backup power supply means in block units to the disk cache in which the storage area is divided into blocks, and the matrix switch means is provided at the time of power failure. 4. The power failure processing method in the disk cache system according to claim 1, wherein the backup power supply unit supplies power only to the block of the disk cache that stores the data having the write data attribute. 5. The power failure processing method in a disk cache system according to claim 1, wherein the non-volatile storage means is a flash memory. 6. A disk cache that primarily stores data transferred between a disk device that stores data and a host device that accesses the disk device, and power is supplied to at least the disk cache during a power failure. When a power failure occurs in a disk cache system having backup power supply means for temporarily supplying power and non-volatile storage means for writing data stored in the disk cache at the time of power failure, the data is actually written in the disk device. A saving power amount calculating step for obtaining a saving power amount necessary for executing the data saving process of the data to the nonvolatile storage means based on the number of write data attribute data stored only in the disk cache. And the remaining power of the backup power supply means The data cache step of saving only the data of the write data attribute stored in the disk cache to the non-volatile storage unit when the power consumption becomes equal to or less than the power saving amount according to Blackout processing method in Japan.