JPH10124359A - Cache selection method and data processing system - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To selectively use a volatile cache and a nonvolatile cache. Selectively use each cache in different devices. SOLUTION: Caches a501, b502, c50
3 to obtain volatile / non-volatile information and select volatile cache if the content can be stored in volatile cache;
If the content is to be stored in the nonvolatile cache, the nonvolatile cache is selected. Furthermore, caches a501 and b5
02, c503, the position information on the system configuration is obtained,
Choose a cache that minimizes straddling between buses. [Effect] The required capacity of the nonvolatile cache can be reduced. It is possible to reduce the frequency of transfer over a plurality of buses.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cache selection method and a data processing system.
In a data processing system having a plurality of caches, the present invention relates to a method of selecting an optimum cache for storing contents generated by a certain process, and a system for implementing the method.

[0002]

2. Description of the Related Art One example of a data processing system is a disk array system. Generally, a disk array system includes a host device, a storage control device, and a disk group. In a conventional disk array system,
Each of the host device and the storage control device has a volatile cache or a nonvolatile cache. The cache of the host device is managed exclusively by the host device, and the cache of the storage control device is managed exclusively by the storage control device.

[0003]

In a conventional disk array system, a host device has both a volatile cache and a nonvolatile cache, and a storage controller has both a volatile cache and a nonvolatile cache. Therefore, there was no way to selectively use volatile caches and nonvolatile caches. However, when only a nonvolatile cache is provided, a margin is required for the capacity, and there is a problem that the price of the disk array system is increased (because the nonvolatile cache is more expensive than the volatile cache).

Further, in the conventional disk array system, integrated management of the caches existing in the host device and the storage control device is not performed, and therefore, the cache of the host device and the cache of the storage control device are selectively provided. Neither was used properly. For this reason, there has been a problem that related transfer is frequently performed between the host device and the storage controller, and transfer overhead is increased.

Accordingly, a first object of the present invention is to provide a cache selecting method which can selectively use a volatile cache and a nonvolatile cache.
A second object of the present invention is to provide a cache selection method that can selectively use each cache of different devices. A third object of the present invention is to provide a cache selection method that can selectively use a plurality of caches according to various criteria.

[0006]

According to a first aspect, the present invention relates to a method for storing the contents generated by a certain process in one of a plurality of caches, the volatile / nonvolatile information of each cache. A cache selection method is provided, wherein a cache that stores the content is selected based on the content. In the above-described cache selection method, it is possible to selectively use contents such as storing contents that can be stored in the volatile cache in the volatile cache and storing only contents to be stored in the nonvolatile cache in the nonvolatile cache. As a result, the capacity of the expensive nonvolatile cache can be suppressed to a necessary minimum, so that the cost of the system can be reduced.

In a second aspect, the present invention selects a cache for storing the content based on the attribute information of each cache when the content generated by a certain process is stored in any of a plurality of caches. Provide a cache selection method. The cache selection method is an extension of the cache selection method according to the first aspect. Elements of the attribute include volatile / non-volatile cache,
Examples include cache capacity, cache access speed (transfer speed), and cache multiplexing. A plurality of caches having different attributes can be selectively used, and a system configuration without waste can be achieved. Therefore, the cost of the system can be reduced.

According to a third aspect, the present invention provides a method for selecting a cache for storing the content based on information on the capacity of each cache when storing the content generated by a certain process in any of a plurality of caches. And providing a cache selecting method. The cache selection method is an example of the cache selection method according to the second aspect. A plurality of caches having different capacities can be selectively used, and the plurality of caches can be used efficiently.

According to a fourth aspect of the present invention, when the content generated by a certain process is stored in any of a plurality of caches, the span of the data transfer path of the related transfer path is minimized. And storing the contents by selecting a cache to be provided. In the above-described cache selection method, the caches of different devices can be selectively used so that the span between the data transfer paths is minimized. As a result, transfer over different devices is suppressed, and transfer overhead can be reduced. For example, the transfer time can be reduced. Further, the transfer performed by another device is not hindered.

According to a fifth aspect of the present invention, when the content generated by a certain process is stored in any of a plurality of caches, the content is stored based on positional information on the system configuration of each cache. There is provided a cache selection method characterized by selecting a cache. The cache selection method is an extension of the cache selection method according to the fourth aspect. A plurality of caches having different positions in the system configuration can be selectively used, and integrated cache management can be performed, so that transfer overhead can be reduced.

In a sixth aspect, the present invention provides a method for storing content generated by a certain process in one of a plurality of caches, based on how the cache is used in the process or another process. A cache selection method is provided, wherein a cache for storing contents is selected. In the above-described cache selection method, a plurality of caches are selectively used selectively based on the use of the cache in the process of generating the content to be stored in the cache or the use of the cache in another process, so that an effective use of the cache can be realized.

According to a seventh aspect, the present invention provides a method for storing, when storing a content generated by a certain process into one of a plurality of caches, the cache for storing the content based on information on the free space of each cache. There is provided a cache selection method characterized by selecting. In the above cache selection method, the free space of each cache is used as a criterion,
Since a plurality of caches are selectively used, the cache utilization efficiency can be improved as a whole.

According to an eighth aspect of the present invention, in a data processing system having a plurality of caches, when storing the contents generated by a certain process in any of the plurality of caches, the volatile / nonvolatile of each cache is used. A data processing system comprising: a cache selection unit that selects a cache that stores the content based on gender information. In the data processing system, the cache selection method according to the first aspect can be suitably implemented.

According to a ninth aspect, the present invention provides a data processing system having a plurality of caches, wherein when the contents generated by a certain process are stored in any of the plurality of caches, the attribute information of each cache is A data processing system comprising: a cache selection unit that selects a cache that stores the content based on the content. In the above data processing system, the cache selection method according to the second aspect can be suitably implemented.

According to a tenth aspect, the present invention provides the data processing system according to the eighth or ninth aspect, further comprising an attribute information acquiring means for obtaining volatile / nonvolatile information or attribute information of each cache. A data processing system is provided. In the above data processing system,
Since volatility / non-volatile information or attribute information of each cache is obtained spontaneously, it is possible to flexibly respond to changes in the system configuration.

According to an eleventh aspect, the present invention relates to a data processing system having a plurality of caches, wherein a content generated by a certain process is stored in any one of the plurality of caches. And a cache selection unit for selecting a cache that minimizes the spread between the data transfer paths of the first path and the second path. In the above data processing system,
The cache selection method according to the fourth aspect can be suitably implemented.

According to a twelfth aspect, the present invention relates to a data processing system having a plurality of caches, wherein when the content generated by a certain process is stored in any of the plurality of caches, the system configuration of each cache is reduced. And a cache selection unit for selecting a cache for storing the content based on the location information. In the data processing system, the cache selection method according to the fifth aspect can be suitably implemented.

According to a thirteenth aspect, the present invention provides the twelfth aspect,
In the data processing system according to the aspect of the present invention, there is provided a data processing system characterized by comprising position information obtaining means for obtaining position information of each cache. In the data processing system, since the position information of each cache is obtained spontaneously, it is possible to flexibly cope with a change in the system configuration.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Referring to FIGS.
An embodiment of the present invention will be described. Note that the present invention is not limited by this.

<First Embodiment> FIG. 1 is a block diagram of a disk array system 1 according to a first embodiment of the present invention. The disk array system 1 includes a host system 100, an accelerator board 200, and a disk group 300.

The host system 100 includes an MPU 101
, PCI chipset 102 and main memory a10
3, the MPU bus 104, and the primary PCI bus 10
5 and a system ROM 106. MPU101
The PCI chipset 102, the main memory 103, and the system ROM 106 are connected by an MPU bus 104. The MPU bus 104 and the primary PCI bus 105 are connected by a PCI chipset 102. The main memory a103 is volatile (content is lost due to power failure) and has a read / write request use area 500 and a volatile cache a501.

The accelerator board 200 is a PCI-
PCI bridge 201 and SCSI controller 202
, A transfer device 401, an XOR operation device 402, a memory interface (IF) 203, a nonvolatile memory b204, a volatile memory c205, and a secondary PC.
And an I bus 206. The nonvolatile memory b204
Has a non-volatile cache b502. The volatile memory c205 has a volatile cache c503. The PCI-PCI bridge 201 and the SCSI
The controller 202, the transfer device 401, and the XOR
The arithmetic device 402 and the memory IF 203 are connected by a secondary PCI bus 206. Non-volatile memory b
The 204, the volatile memory c205, and the secondary PCI bus 206 are connected by a memory IF 203. Also,
Secondary PCI Bus 206 and Host System 10
The primary PCI bus 105 is connected by a PCI-PCI bridge 201.

The disk group 300 includes a plurality of disk devices 301 arranged in an array and a SCSI bus 302. The disk device 301 and the SCSI controller 202 of the accelerator board 200
They are connected by a bus 302. That is, the SCSI bus 302 of the disk device 301 and the secondary PCI bus 206 of the accelerator board 200 are connected by the SCSI controller 202.

The MPU bus 104 and the primary P
The CI bus 105, the secondary PCI bus 206, and the SCSI bus 302 can individually transfer data and the like.

The main memory a1 of the host system 100
03 and the non-volatile memory b of the accelerator board 200
The volatile memory 204 and the volatile memory c205 are mapped as accessible areas on the memory space of the MPU 101 of the host system 100. With this mapping, the MPU 101 can manage the main memory a103, the nonvolatile memory b204, and the volatile memory c205.

The cache management software is stored in the system ROM 106 or the disk device 301, and is executed by the MPU 101 as stored, or read out to the main memory a103 as needed. That is, cache management is performed by MPU1
01.

FIG. 2 is a configuration diagram of the management information table 120 for managing each device related to data transfer to the cache and data transfer from the cache. The management information table 120 includes a device name 121, a device type 1
22, a cache attribute 123, a bus name 124 to which the device is connected, and a bus hierarchy number 125 representing the number of buses from the bus to which the device is connected to the MPU bus 104. It is configured.

After the disk array system 1 is started up and the configuration on each bus is changed, the MPU 101 executes at least one time the system configuration of each device related to data transfer to the cache and data transfer from the cache. The position is confirmed, and the position information is stored in the management information table 12.
Hold at 0. Confirmation of the position is realized, for example, by cyclically inquiring about the existence of all devices on each bus. Further, the positions of some or all of the devices in the system configuration are fixed, the fixed position information is set in the management information table 120 in advance, and only the non-fixed position information is collected by the inquiry, and the management information is collected. You may make it hold | maintain in the table 120. It should be noted that another device that knows the position of the device, instead of the device itself, may respond to the inquiry. For example, the memory IF 203, which is a PCI device that knows the existence of the nonvolatile memory b204, may respond to the inquiry about the position of the nonvolatile memory b204 based on the PCI bus protocol.

After confirming the location of each device, the MPU 101 confirms the attribute of each cache at least once, and stores the attribute information in the management information table 120. Elements of the attribute include cache volatility / non-volatility, cache capacity, cache access speed (transfer speed), cache multiplexing, and the like. Confirmation of the attribute is realized, for example, by inquiring all caches about the attribute cyclically. Further, a part or all of the attributes of the cache are fixed, the fixed attribute information is set in the management information table 120 in advance, and only the non-fixed attribute information is collected by the inquiry.
20. Note that other devices that know the attributes of the cache may respond to the inquiry, instead of the cache itself. For example, the memory IF 203, which is a PCI device that knows the existence of the nonvolatile memory b204, may respond to the inquiry about the attribute of the nonvolatile memory b204 based on the PCI bus protocol.

If a change in cache attributes can occur,
Each cache may notify the MPU 101 of the attribute change, and the MPU 101 receiving the notification may change the held attribute information. For example, assuming that the non-volatility in the non-volatile cache b502 is realized by a power supply backup by a battery, the non-volatility is lost when the battery becomes unavailable, and the non-volatile cache b502
Becomes volatile. At this time, the nonvolatile cache b50
2 has the attribute change notification means, it notifies the MPU 101 that it has become volatile. Upon receiving this notice,
The MPU 101 changes nonvolatile in the attribute information of the nonvolatile cache b502 to volatile.

In the disk array system 1 of the present embodiment, when data such as data and parity is stored in the cache in the processing of a read request or a write request, the MP
U101 checks the position of each device in the system configuration, and checks the attributes of each cache. Then, when storing the content in the cache, the MPU 101 determines the attribute to be possessed by the cache from the type of the content and the like, predicts the transfer path, and predicts the cache having the attribute that matches the determination. The cache with the least frequent transfer path spanning multiple buses is selected.

As can be understood from the above description, the position information acquisition means for acquiring the position information on the system configuration of each device including each cache, and the attribute information of each cache such as volatile / non-volatile. Attribute information acquisition means to perform
For example, a cache selecting unit that selects a cache that stores the contents based on the position information so as to select a cache that minimizes a straddle between buses of a related transfer path; The cache selecting means for selecting a cache for storing the contents based on such attribute information is realized by the MPU 101 executing software realizing each of these means.

Next, specific examples of cache selection will be described. << Selection of cache for storing read data >>
FIG. 9 is a flowchart of a process of selecting a cache for storing data read from the disk group 300 to the read / write request use area 500 based on a read request. In step 1010, a criterion for determining the attribute of the cache is extracted from the type of data. This criterion is preset in the cache management software. Here, considering only volatile / non-volatile as the attribute of the cache, if the data type is “data read from the disk group 300 to the read / write request use area 500 based on the read request”, May be extracted ". The reason for this determination criterion is that the data is originally stored in the disk group 300, and there is no problem if the data is lost from the cache. Therefore, the data need not be limited to the nonvolatile cache.

In step 1020, if the content of the retrieved criterion is "either volatile or non-volatile", the process proceeds to step 1030. If the content of the retrieved criterion is “limited to non-volatile” or “limited to volatile”, the process proceeds to another process. In step 1030, the MPU 101 sets all caches as candidates. Here, the volatile cache a501, the nonvolatile cache b502, and the volatile cache c503 are cache candidates.

In step 1040, a transfer route when each of the candidate caches is selected is predicted. This prediction of the transfer path can be realized by providing the cache management software with a transfer path for each data type in advance and searching by the data type. For example, when the volatile cache a501 is selected, the transfer path is [Read /
Write request use area 500]-[Transfer device 401]-
[Volatile cache a501] or [disk device 3]
01]-[Transfer device 401]-[Volatile cache a5
01]. When the non-volatile cache b502 is selected, the transfer path is changed to [read / write request use area 5].
00]-[Transfer device 401]-[Nonvolatile cache b
502] or [Disk device 301]-[Transfer device 4]
01]-[non-volatile cache b502]. When the volatile cache c503 is selected, the transfer path is [Read / Write Request Use Area 500]-[Transfer Device 401]-[Volatile Cache c503] or [Disk Device 301]-[Transfer Device 401]-[Volatile] Sex cache c503].

In step 1050, a cache which becomes a transfer path with the least straddling between buses is set as a candidate.
The span between buses can be calculated by the bus hierarchy number 125 in the management information table 120. For example, when the volatile cache a501 is selected, one transfer path is [Read / Write request use area 500]-[Transfer device 4].
01]-[volatile cache a501], the span between buses is | read / write request use area 50
0 bus hierarchy number “1” −Transfer device 401 bus hierarchy number “3” | + | Transfer device 401 bus hierarchy number “3” −
Bus hierarchy number “1” of the volatile cache a 501 | =
It becomes “4”. Further, since the other transfer path is [disk device 301]-[transfer device 401]-[volatile cache a501], the straddling between buses is: | bus hierarchy number "4" of disk device 301-transfer device 401
Bus hierarchy number “3” | + | bus hierarchy number “3” of the transfer device 401−bus hierarchy number “1” | = “3” of the volatile cache a501. By the same calculation, the span between the buses of the transfer path when the nonvolatile cache b502 is selected is “2” or “1”, and the span between the buses of the transfer path when the volatile cache c503 is selected is “2”. "Or" 1 ". Therefore, the nonvolatile cache b502 and the volatile cache c503 are candidates.

In step 1063, the number of remaining candidates is determined. If the number is singular, the process is terminated. If the number is plural, the process proceeds to step 1070. Here, the nonvolatile cache b5
02 and the volatile cache c 503 remain as candidates, so the process proceeds to step 1070.

In order to facilitate understanding, steps 1010 to 1050 have been described separately.
If the processing results of steps 1010 to 1050 are obtained in advance for each data type, and the results are searched by data type, the same result as described above can be obtained in one step. The same applies to the following steps.

In step 1070, the criterion for determining the attribute of the cache is extracted from the way the cache is used in other processes. This criterion is preset in the cache management software. For example, when there is a write-back type write process as another process, a criterion of “limited to volatility” is extracted. The reason for this criterion is that other processes (write-back write process) require the use of a non-volatile cache, so that this process (store read data in the cache) This is because it is desirable to use a volatile cache.

In step 1071, if the content of the retrieved criterion is "limited to volatile", step 1072
Proceed to. If the content of the retrieved criterion is “limited to nonvolatile” or “can be volatile or nonvolatile”, the process proceeds to another process. In step 1072, the MPU 101 sets the volatile cache among the remaining candidates as a candidate. Here, since the nonvolatile cache b502 and the volatile cache c503 remain, the volatile cache c503
Is a candidate. In step 1073, the number of remaining candidates is determined. If the number is singular, the process is terminated. If the number is plural, the process proceeds to step 1071. Here, the volatile cache c50
Since only 3 remains, the process ends. If a plurality of volatile caches remain as candidates, the process proceeds to step 1080 in FIG.

In step 1080 in FIG. 4, a criterion for determining the cache capacity is extracted from the type of data. This criterion is preset in the cache management software. For example, if the data type is “Read / write request use area 500 from disk group 300 based on read request”
In the case of "data read to", a criterion of "maximum capacity" is extracted. The reason for this criterion is that if a large-capacity cache is used as much as possible, a small-capacity cache can be preserved, and this is provided in case a small-capacity cache must be used. In step 1081, if the content of the retrieved criterion is “maximum capacity”, the process proceeds to step 1082. If the content of the retrieved criterion is “minimum capacity”, the process proceeds to another process. In step 1082, the MPU 101
The cache having the largest capacity among the remaining candidates is set as a candidate. In step 1083, the number of remaining candidates is determined. If the number is singular, the process is terminated.
Go to 0.

In step 1090, a criterion for determining the free space in the cache is extracted from the type of data. This criterion is preset in the cache management software. For example, when the type of data is “data read from the disk group 300 to the read / write request use area 500 based on a read request”, a criterion of “large free space” is extracted. The reason for this determination criterion is that if a cache having a large free space is preferentially used, a free area can be left in each cache.
In step 1091, if the content of the retrieved criterion is “large free space”, the process proceeds to step 1092. If the content of the retrieved criterion is "small free space", the process proceeds to another process. In step 1092, the MPU 101
, A cache having a large free space among remaining candidates is set as a candidate. In step 1093, the number of remaining candidates is determined. If the number is singular, the process is terminated. If the number is plural, the process proceeds to another determination criterion.

The order in which the above criteria are applied is not limited to the above description. For example, the above steps 1070 to 1073
And the above steps 1800 to 1083 may be exchanged. Also, the above steps 1070 to 1073 and the above steps 1090 to 1093 may be exchanged. In general, since the nonvolatile memory is more expensive than the volatile memory, the capacity of the nonvolatile cache is often smaller than the capacity of the volatile cache. For this reason, the free space of the nonvolatile cache is often smaller than the free space of the volatile cache. Therefore, the above step 1070
By exchanging Steps 1080 to 1073 with Steps 1800 to 1083 or Steps 1070 to 1073 and Steps 1090 to 1093, the volatile cache c503 is selected. As a result, the nonvolatile cache b502 can be preserved and can be used effectively. Further, the required capacity of the nonvolatile cache b502 can be reduced.

<< Selection of Cache to Store Data Written in Write-Through Method >> In the write-through method, data is written to the disk group 300 at the same time as data is stored in the cache. , Considering only volatile / non-volatile,
It may be volatile or non-volatile. The data transfer path to the cache is the same as the path for transferring data read from the read / write request use area 500 to the cache based on the read request. Therefore, selection of a cache for storing data written to the disk group 300 from the read / write request use area 500 based on a write request is similar to selection of a cache for storing data read based on the read request. Therefore, the volatile cache c 503 is selected.

If the old contents of the data exist in the cache, the old contents are overwritten by the new contents, or the old contents are stored in another location before the new contents are stored in the cache. Is deleted (invalidated) from.

<< Selection of Cache to Store Parity Generated in Write-Through Method >> When data is written from read / write request use area 500 to disk group 300 based on a write request, the parity is updated. FIG. 5 is a diagram illustrating an example of a parity update operation. Step 2010: Transfer old contents of data from the disk group 300 or the cache to the XOR operation device 402. Step 2020: Read / write request use area 50
From 0, the new contents of the data are transferred to the XOR operation unit 402. Steps 2030 and 2040: XOR operation device 402
Generates and holds a difference by an XOR operation of the old content and the new content of the data. Step 2050: The pre-update value of the parity is transferred from the disk group 300 or the cache to the XOR operation device 402. Step 2060: Generate an updated parity value by XOR operation of the difference and the pre-update value of the parity, transfer the updated parity value to the cache, and write the updated parity value to the disk group 300.

Instead of storing the parity (updated value) in the cache, the difference generated in step 2030 may be stored in the cache. The difference is stored in the cache, and when there is a write request to update the same parity again, the frequency of reading the pre-update value of the parity from the disk group 300 can be reduced by updating the difference. .

FIG. 6 is a flow chart of a process for selecting a cache for storing (updated value) or a difference generated by the XOR operation device 402. In step 3010, the criterion of the attribute of the cache with respect to the parity or the difference is extracted. This criterion is preset in the cache management software. Here, considering only volatile / non-volatile as the attribute of the cache, a criterion of “can be volatile or non-volatile” is extracted. The reason for this determination criterion is that in the write-through method, since data is written to the disk group 300,
This is because there is no problem even if the parity or the difference is lost from the cache, and it is not necessary to limit the parity or the difference to the nonvolatile cache.

In step 3020, if the content of the retrieved criterion is "either volatile or non-volatile", the flow advances to step 3030. If the content of the retrieved criterion is “limited to non-volatile” or “limited to volatile”, the process proceeds to another process. In step 3030, the MPU 101 sets all caches as candidates. Here, the volatile cache a501, the nonvolatile cache b502, and the volatile cache c503 are cache candidates.

In step 3040, a transfer route when each candidate cache is selected is predicted. The prediction of the transfer path can be realized by providing the cache management software with a parity or difference transfer path in advance and searching for it. For example, volatile cache a
When 501 is selected, the transfer path is set to [XOR operation device 40
2]-[volatile cache a501]. When the non-volatile cache b502 is selected, the transfer path is [XOR operation device 402]-[non-volatile cache b5
02]. When the volatile cache c503 is selected, the transfer path is [XOR operation device 402]-[volatile cache c503].

In step 3050, a cache which is a transfer path with the least straddling between buses is determined as a candidate.
The span between buses can be calculated by the bus hierarchy number 125 in the management information table 120. For example, when the volatile cache a501 is selected, the transfer path is [XO
R operation device 402]-[volatile cache a501], the span between buses is | XOR operation device 40
2, the bus hierarchy number “3” −the bus hierarchy number “1” | = “2” of the volatile cache a501. By the same calculation, the span between the buses of the transfer path when the nonvolatile cache b502 is selected is "0", and the span between the buses of the transfer path when the volatile cache c503 is selected is also "0". Therefore, the nonvolatile cache b5
02 and the volatile cache c503 are candidates.

In step 3063, the number of remaining candidates is determined. If the number is singular, the process is terminated. If the number is plural, the process proceeds to step 3070. Here, the nonvolatile cache b5
02 and the volatile cache c 503 remain as candidates, so the process proceeds to step 3070.

In step 3070, a criterion for determining the attribute of the cache is extracted from the way the cache is used in other processes. This criterion is preset in the cache management software. For example, when there is a write-back type write process as another process, a criterion of “limited to volatility” is extracted. The reason for this criterion is that other processes (write-back write process) require the use of a non-volatile cache, so that this process (store read data in the cache) This is because it is desirable to use a volatile cache.

In step 3071, if the content of the retrieved criterion is "limited to volatile", step 3072
Proceed to. If the content of the retrieved criterion is “limited to nonvolatile” or “can be volatile or nonvolatile”, the process proceeds to another process. In step 3072, the MPU 101 sets the volatile cache among the remaining candidates as candidates. Here, since the nonvolatile cache b502 and the volatile cache c503 remain, the volatile cache c503
Is a candidate. In step 3073, the number of remaining candidates is determined. If the number is singular, the process is terminated. If the number is plural, the process proceeds to another process. Here, since only the volatile cache c503 remains, the process is terminated. Note that the other processing when there are a plurality of candidates is the same as in step 1080 and subsequent steps in FIG.

If the pre-update value of the parity or the difference exists in the cache, the pre-update value is overwritten by the update value or after the update value is stored in another location. The pre-update value is deleted (invalidated) from the cache.

<< Selection of Cache for Storing Write Data in Write-Back Method >> FIG. 7 shows a process of selecting a cache for storing data to be written to the disk group 300 from the read / write request use area 500 based on a write request. It is a flowchart. In step 4010,
The criterion of the attribute of the cache is extracted from the data type. This criterion is preset in the cache management software. Here, considering only volatile / non-volatile as the attribute of the cache, if the data type is “data written to the disk group 300 from the read / write request use area 500 based on a write request”, “limited to non-volatile” Is extracted. The reason for this determination criterion is that in the write-back method, the data is only stored in the cache when the processing for the write request is completed, and may not be written to the disk group 300. This is because it is necessary to limit the cache to a non-volatile cache.

In step 4020, if the content of the retrieved criterion is "limited to non-volatility", step 403
Go to 0. If the content of the retrieved criterion is “can be volatile or nonvolatile” or “limited to volatile”, the process proceeds to another process. In step 4030, the MPU 101 sets a nonvolatile cache as a candidate. Here, the nonvolatile cache b502 is set as a cache candidate. Step 4
In step 063, the number of remaining candidates is determined. If the number is singular, the process is terminated. If the number is plural, the process proceeds to another process. Here, since only the non-volatile cache b502 remains, the process ends. Note that the other processing when there are a plurality of candidates is the same as in step 1040 and subsequent steps in FIG.

If the old contents of the data exist in the cache, the old contents are overwritten by the new contents, or the old contents are stored in another location before the new contents are stored in the cache. Is deleted (invalidated) from.

According to the disk array system 1 described above, it is possible to reduce the frequency of transfer of data and parity over a plurality of buses. As a result, the transfer processing time can be reduced, and it is possible to suppress the hindrance of the processing of other devices connected to each bus, thereby improving the performance of the system. In addition, since the cache is used to preserve the nonvolatile cache as much as possible,
The required amount of the nonvolatile memory can be reduced, and the cost of the system can be reduced.

In the case where the free space of the selected cache is insufficient, the cache may be excluded and the above selection method may be applied again. When the same content is stored in some of a plurality of caches for performing cache multiplexing or the like, one cache is selected by applying the above selection method, and then the cache is excluded to remove the cache. Is applied again to select the next cache, and this is repeated to select the required number of caches.

<Second Embodiment> FIG. 8 is a block diagram of a disk array system 2 according to a second embodiment of the present invention. The difference between this disk array system 2 and the disk array system 1 of the first embodiment is that the MPU 10
1 has the functions of the transfer device 401 and the XOR operation device 402, and the accelerator board 200
The point is that no arithmetic unit is provided.

In this embodiment, the cache management is performed by the MPU 101 in the same manner as in the first embodiment. That is, each means related to the present invention is realized by the MPU 101 executing software for realizing each means.

FIG. 9 is a configuration diagram of the management information table 190 for managing each device related to data transfer to the cache and data transfer from the cache. The difference between this management information table 190 and the management information table 120 of the first embodiment is that the bus names of the transfer device 401 and the XOR operation device 402 are “MPU bus” and the bus hierarchy number is “1”. It is.

Next, specific examples of cache selection will be described. << Selection of Cache to Store Read Data >> The process of selecting a cache to store read data from the disk group 300 to the read / write request use area 500 based on a read request is described in the flowcharts of FIGS. It is on the street. Steps 1010 to 1040 in FIG. 3 are the same as those in the first embodiment. In step 1050, a cache which is a transfer path with the least straddling between buses is set as a candidate. For example, when the volatile cache a501 is selected, one transfer path is [read / write request use area 500]-[transfer]. [Device 401] − [volatile cache a501], so that the straddling between buses is the bus hierarchy number “1” of the | read / write request use area 500.
Bus layer number “1” of transfer device 401 | + | transfer device 4
01 bus hierarchy number “1” —volatile cache a501
Bus hierarchy number “1” | = “0”. Another transfer path is [disk device 301]-[transfer device 4].
01]-[volatile cache a501], the span between buses is: | bus hierarchy number “4” of disk device 301−bus hierarchy number “1” of transfer device 401 | + |
The bus hierarchy number “1” of the transfer device 401−the bus hierarchy number “1” | = “3” of the volatile cache a501. By the same calculation, the span between the buses of the transfer path when the nonvolatile cache b502 is selected is “2” or “5”, and the span between the buses of the transfer path when the volatile cache c503 is selected is “2”. "Or" 5 "
Becomes Therefore, the volatile cache a501 is set as a candidate. In step 1063, the number of remaining candidates is determined. If the number is singular, the process is terminated;
Go to 0. Here, since only the volatile cache a501 remains as a candidate, the process ends.

<< Selection of Cache for Storing Data Written in Write-Through Method >> Disk group 3 from read / write request use area 500 based on a write request
The selection of the cache for storing the data written to 00 is performed in the same manner as the selection of the cache for storing the data read based on the read request, as described in the first embodiment. Therefore, the volatile cache a501
Is selected.

<< Selection of Cache for Storing Parity Generated in Write-Through Method >> A cache for storing a parity or a difference generated when data is written from the read / write request use area 500 to the disk group 300 based on a write request. Is selected as shown in the flowchart of FIG. Steps 3010 to 304 in FIG.
Up to 0 is the same as in the first embodiment. Step 3050
In this example, a cache which is a transfer path with the least straddling between buses is set as a candidate.
The transfer path when 501 is selected is [XOR operation device 4
02]-[volatile cache a501], the straddling between buses is | bus hierarchy number "1" of XOR operation device 402-bus hierarchy number "1" of volatile cache a501 | = "0" . By the same calculation, the span between the buses of the transfer path when the nonvolatile cache b502 is selected is “2”, and the volatile cache c503 is selected.
Is "2" across the buses of the transfer path when is selected.
Becomes Therefore, the volatile cache a501 is set as a candidate. In step 3063, the number of remaining candidates is determined. If the number is singular, the process is terminated.
Go to 0. Here, since only the volatile cache a501 remains as a candidate, the process ends.

<< Selection of Cache for Storing Write Data in Write-Back Method >> Based on a write request, the disk group 300
The process of selecting the cache that stores the data to be written to is as shown in the flowchart of FIG. Step 4 in FIG.
The process from 010 to 4063 is the same as in the first embodiment, and the nonvolatile cache b502 is selected.

According to the disk array system 2 described above, it is possible to reduce the frequency of transfer of data and parity over a plurality of buses. As a result, the transfer processing time can be reduced, and it is possible to suppress the hindrance of the processing of other devices connected to each bus, thereby improving the performance of the system. Further, since the cache is used so as to preserve the nonvolatile cache as much as possible, the required amount of the nonvolatile memory can be reduced, and the cost of the system can be reduced. Further, since the accelerator board 200 without the transfer device 401 and the XOR operation device 402 is used, the configuration of the accelerator board 200 is simplified and the price is low.

<Third Embodiment> FIG. 10 is a block diagram of a disk array system 3 according to a third embodiment of the present invention. The disk array system 3 includes a host system 100 and a disk group 300.

The host system 100 includes the MPU 101
, PCI chipset 102 and main memory a10
3, the MPU bus 104, and the primary PCI bus 10
5, a system ROM 106, and a host SCSI controller 107. MPU 101, PCI chipset 102, main memory a103, and system ROM
106 is connected by the MPU bus 104. Also,
The MPU bus 104 and the primary PCI bus 105
Are connected by a PCI chipset 102. Further, the PCI chipset 102 and the host SCSI controller 107 are connected by a primary PCI bus 105. The main memory a103 is volatile, and the read / write request use area 500 and the volatile cache a
501.

The disk group 300 includes a plurality of disk devices 301 arranged in an array and a SCSI bus 302. Disk device 301 and host system 1
00, the host SCSI controller 107
They are connected by a bus 302. That is, the SCSI bus 302 of the disk device 301 and the host system 10
The primary PCI bus 105 is connected by a host SCSI controller 107.

The disk device 301 is divided into a disk interface (IF) unit 310 and a disk main unit 320. The disk IF unit 310
A CSI controller 311, a disk memory interface (IF) 312, a non-volatile memory d313,
It has a volatile memory e314, a transfer device 401, and an XOR operation device 402. The nonvolatile memory d3
13 has a non-volatile cache d504. Further, the volatile memory e314 has a volatile cache e505.
have. The disk SCSI controller 311, the disk memory interface (IF) 312, the transfer device 401, the XOR operation device 402, and the disk main part 320 are connected by a disk internal bus 330. The disk internal bus 330, the nonvolatile memory d313, and the volatile memory e314
They are connected by the disk memory IF 312. In addition, the disk SCSI controller 311
It is connected to the SI bus 302.

The MPU 101 generates a SCSI command and sends it to the disk device 3 via the host SCSI controller 107 and the disk SCSI controller 311.
01 non-volatile memory d313 and volatile memory e3
Access 14. With this access, MPU 10
1 can manage the cache.

The MPU bus 104 and the primary P
The CI bus 105, the SCSI bus 302, and the disk internal bus 330 can individually transfer data and the like.

In this embodiment, the cache management is performed by the MPU 101 in the same manner as in the first embodiment. That is, each means related to the present invention is realized by the MPU 101 executing software for realizing each means.

FIG. 11 is a configuration diagram of the management information table 111 for managing each device related to data transfer to the cache and data transfer from the cache. This management information table 111 is different from the management information table 120 of the first embodiment in that the bus names of the transfer device 401, the XOR operation device 402, the nonvolatile cache d504, and the volatile cache e505 are “disk internal bus” and the bus The point that the layer number is “4” and that the disk main unit 320 is provided instead of the disk device 301, the bus name is “disk internal bus”, and the bus layer number is “4”. .

Next, specific examples of cache selection will be described. << Selection of Cache to Store Read Data >> The process of selecting a cache to store read data from the disk group 300 to the read / write request use area 500 based on a read request is described in the flowcharts of FIGS. It is on the street. Steps 1010 and 1020 in FIG. 3 are the same as those in the first embodiment. In step 1030, the MPU 10
1 makes all caches candidates. Here, the volatile cache a 501 and all the disk devices 301
The non-volatile cache d504 and the volatile cache e505 are cache candidates.

At step 1040, a transfer route when each candidate cache is selected is predicted. For example, when the volatile cache a501 is selected, the transfer path is [read / write request use area 500]-[transfer device 401].
-[Volatile cache a501] or [disk main part 320]-[transfer device 401]-[volatile cache a501]. Here, [disk main part 320]
Is limited to the disk main part 320 of the disk device 301 storing the read data. On the other hand, [Transfer device 4
01] can be used in the transfer device 401 of any disk device 301, but [Disk main part 320]-[Transfer device 4]
01]-[Volatile cache a501], when the transfer device 401 of the disk device 301 different from the [disk main part 320] is used, the straddling between buses obviously increases. Therefore, the [transfer device 401] is the transfer device 401 of the disk device 301 that stores the read data. When the non-volatile cache d504 is selected, the transfer path is [Read / Write request use area 500]-[Transfer device 401]-[Non-volatile cache d504] or [Disk main part 32].
0]-[Transfer device 401]-[Non-volatile cache d5
04]. Here, although the [transfer device 401] and the [non-volatile cache d504] are present in each disk device 301, if both are not in the same disk device 301, the straddling between buses obviously increases. There is. For the above-mentioned reason, the [transfer device 401] is the transfer device 401 of the disk device 301 that stores the read data. Therefore, [transfer device 401] and [non-volatile cache d
504] is the transfer device 401 of the disk device 301 storing the read data and the non-volatile cache d5.
04. When the volatile cache e505 is selected, the transfer path is changed to [Read / Write request use area 50].
0]-[Transfer device 401]-[Volatile cache e50
5] or [Disk main part 320]-[Transfer device 40]
1]-[volatile cache e505]. Here, for the above-mentioned reason, the [transfer device 401] and the [non-volatile cache d504] are the transfer device 401 and the non-volatile cache d504 of the disk device 301 storing the read data.

In step 1050, a cache which is a transfer path with the least straddling between buses is selected as a candidate. For example, when the volatile cache a501 is selected, one transfer path is set to the [read / write request use area 50].
0]-[Transfer device 401]-[Volatile cache a50
1], the span between the buses is: | bus hierarchy number “1” of read / write request use area 500−bus hierarchy number “4” of transfer device 401 | + | bus hierarchy number “4” of transfer device 401 The bus hierarchy number “1” | = “6” of “−volatile cache a501”. Another transfer path is [disk main part 320]-[transfer device 401].
-[Volatile cache a501], the span between buses is: | bus hierarchy number "4" of disk main part 320-bus hierarchy number "4" of transfer device 401 | + | bus hierarchy number of transfer device 401 “4” −the bus hierarchy number “1” | = “3” of the volatile cache a501. By the same calculation, the span between the buses of the transfer path when the nonvolatile cache d504 is selected is “3” or “0”, and the span between the buses of the transfer path when the volatile cache e505 is selected is “3”. "Or" 0 "
Becomes Therefore, the nonvolatile cache d504 and the volatile cache e505 of the disk device 301 storing the read data are candidates. In step 1063, the number of remaining candidates is determined. If the number is singular, the process ends; if the number is plural, the process proceeds to step 1070. Here, the nonvolatile cache d504 and the volatile cache e505
Are left as candidates, the process proceeds to step 1070.

In step 1070, the criterion for determining the attribute of the cache is extracted from the way the cache is used in other processes. For example, when there is a write-back type write process as another process, a criterion of “limited to volatility” is extracted. The reason for this determination criterion is that other processes (write-back type write process) require the use of a non-volatile cache.
This is because it is desirable to use a volatile cache.

In step 1071, if the content of the retrieved criterion is "limited to volatility", step 1072
Proceed to. If the content of the retrieved criterion is “limited to nonvolatile” or “can be volatile or nonvolatile”, the process proceeds to another process. In step 1072, the MPU 101 sets the volatile cache among the remaining candidates as a candidate. Here, the disk device 301 storing the read data
Non-volatile cache d504 and volatile cache e5
05 remains, the volatile cache e505 of the disk device 301 storing the read data is set as a candidate. In step 1073, the number of remaining candidates is determined. If the number is singular, the process is terminated;
Proceed to 1. Here, since only the volatile cache e505 of the disk device 301 storing the read data remains, the processing is terminated. If a plurality of volatile caches remain as candidates, the process proceeds to another criterion.

<< Selection of Cache for Storing Write Data in Write-Through Method >> Based on the write request, the disk group 300
As described in the first embodiment, the selection of the cache that stores the data written to the cache is performed in the same manner as the selection of the cache that stores the data read based on the read request. Accordingly, the volatile cache e50 of the disk device 301 that has written data based on the write request.
5 is selected.

<< Selection of Cache for Storing Parity Generated in Write-Through Method >> A cache for storing a parity or a difference generated when data is written from the read / write request use area 500 to the disk group 300 based on a write request. Is selected as shown in the flowchart of FIG. Steps 3010 and 302 in FIG.
0 is the same as in the first embodiment. In step 3030, the MPU 101 sets all caches as candidates.
Here, the volatile cache a 501 and the nonvolatile cache d 504 and the volatile cache e 505 of all the disk devices 301 are set as cache candidates.

In step 3040, a transfer route when each candidate cache is selected is predicted. For example, when the volatile cache a501 is selected, the transfer path is [XO
R operation device 402]-[volatile cache a501]. Here, [XOR operation device 402] can be used for XOR operation device 402 of any disk device 301,
For the following reason, a disk device (hereinafter referred to as a parity disk device) 3 in which the pre-update value of the parity is stored.
01 XOR operation device 402. That is, the generated parity or difference is written to the disk main part 320 of the disk device 301 other than the disk device to which data is written (hereinafter, referred to as a data disk device) from the viewpoint of data recovery at the time of failure. In consideration of the above, the [XOR operation device 402] also corresponds to the [XOR operation device 40] of the disk device 301 other than the data disk device.
2]. On the other hand, the updated value of the parity is shown in FIG.
As described with reference to, the difference between the old and new contents of the data and the XOR operation of the parity pre-update value are generated. The new content of the data exists in the read / write request use area 500, the old content of the data exists in the data disk device 301, and the pre-update value of the parity exists in the parity disk device 301. Therefore, considering these transfer paths, it is concluded that the [XOR operation device 402] should be the XOR operation device 402 of the parity disk device 301. Next, the non-volatile cache d
When 504 is selected, the transfer path is set to [XOR operation device 40
2]-[Non-volatile cache d504]. Here, [XOR operation device 402]
The XOR operation device 402 of the parity disk device 301 is assumed. The [non-volatile cache d504] can be used in the non-volatile cache d504 of any disk device 301. However, if the non-volatile cache d504 is not in the same disk device 301 as the [XOR operation device 402], the straddling between buses obviously increases. The device 402] is a nonvolatile cache d504 of the parity disk device 301.
Next, when the volatile cache e505 is selected, the transfer path is [XOR operation device 402]-[volatile cache e
505]. The [XOR operation device 402] and the [volatile cache e505] are also referred to as the XOR operation device 402 and the volatile cache e505 of the parity disk device 301 for the above-described reason.

In step 3050, a cache which is a transfer path with the least straddling between buses is selected as a candidate. For example, when the volatile cache a501 is selected, the transfer path is [XOR operation device 402]-[volatile cache a501], the span between buses is |
The bus hierarchy number “4” of the XOR operation device 402−the bus hierarchy number “1” | = “3” of the volatile cache a501. By the same calculation, the span between the buses of the transfer path when the nonvolatile cache d504 is selected is "0", and the span between the buses of the transfer path when the volatile cache e505 is selected is also "0". Therefore, the nonvolatile cache d504 and the volatile cache e50
5 is a candidate. In step 3063, the number of remaining candidates is determined. If the number is singular, the process ends, and if there is more than one, the process proceeds to step 3070. Here, the process proceeds to step 3070 because the nonvolatile cache d504 and the volatile cache e505 remain as candidates.

At step 3070, the criterion for determining the attribute of the cache is extracted from the way the cache is used in other processes. For example, when there is a write-back type write process as another process, a criterion of “limited to volatility” is extracted for the reason described in the first embodiment. In step 3071, if the content of the retrieved criterion is “limited to volatility”, the process proceeds to step 3072. If the content of the retrieved criterion is “limited to nonvolatile” or “can be volatile or nonvolatile”, the process proceeds to another process. In step 3072, the MPU 101 sets the volatile cache among the remaining candidates as candidates. here,
Non-volatile cache d504 and volatile cache e50
Since 5 remains, the volatile cache e505 is set as a candidate. In step 3073, the number of remaining candidates is determined. If the number is singular, the process is terminated. If the number is plural, the process proceeds to another process. Here, since only the volatile cache e505 remains, the process ends. Note that the other processing when there are a plurality of candidates is the same as in step 1080 and subsequent steps in FIG.

<< Selection of Cache for Storing Write Data in Write-Back Method >> Based on a write request, a disk group 300
The process of selecting the cache that stores the data to be written to is as shown in the flowchart of FIG. Step 4 in FIG.
010 and 4020 are the same as in the first embodiment. In step 4030, the MPU 101 sets the nonvolatile cache d504 of the data disk device 301 as a candidate. The reason is that data is finally transferred from the non-volatile cache to the disk main part 320 of the data disk device 301. Therefore, considering the transfer path at that time, it is better to use the non-volatile cache d504 of the data disk device 301. Non-volatile cache d other than disk device
This is because there is clearly less straddling between buses than when 504 is used. In step 4063, the number of remaining candidates is determined, and if the number is singular, the process ends; if the number is plural, the process proceeds to another process. Here, the data disk device 30
Since only one nonvolatile cache d504 remains, the process ends. Note that the other processing when there are a plurality of candidates is the same as in step 1040 and subsequent steps in FIG.

According to the disk array system 3 described above, it is possible to reduce the frequency of transfer of data and parity over a plurality of buses. As a result, the transfer processing time can be reduced, and it is possible to suppress the hindrance of the processing of other devices connected to each bus, thereby improving the performance of the system. Further, since the cache is used so as to preserve the nonvolatile cache as much as possible, the required amount of the nonvolatile memory can be reduced, and the cost of the system can be reduced. Further, it is not necessary to use the accelerator board 200 as in the first and second embodiments.

<Fourth Embodiment> FIG. 12 is a block diagram of a disk array system 4 according to a fourth embodiment of the present invention. This disk array system 4 differs from the disk array system 1 of the first embodiment in that the accelerator board 200 has a memory portable unit 210. The non-volatile memory b204 is in the memory portable unit 210, and is connected to the memory interface (IF) 203 by the memory connection unit 211.

The memory connection means 211 can disconnect and connect the nonvolatile memory b204 and the memory IF 203 without giving an adverse effect on the operation of the nonvolatile memory b204 and the memory IF 203, such as abnormal operation and erasure of the contents. . That is, the memory portable unit 2
10 (ie, the non-volatile memory b204) can be separated from the accelerator board 200.

When a failure such as a power failure or operation stop occurs in the host system 100, the memory portable unit 2
10 is separated from the accelerator board 200, and the memory portable unit 210 is mounted on the accelerator board 200 of the standby disk array system having the same configuration as the disk array system 4, and the nonvolatile memory b204
Can be used in the standby disk array system.

Also in this embodiment, the cache management is performed by the MPU 101 in the same manner as in the first embodiment. That is, each means related to the present invention is realized by the MPU 101 executing software for realizing each means.

According to the disk array system 4 described above, similarly to the first embodiment, it is possible to reduce the frequency of transfer of data and parity over a plurality of buses. As a result, the transfer processing time can be reduced,
Disturbing the processing of other devices connected to each bus can be suppressed, and the performance of the system can be improved.
Further, since the cache is used so as to preserve the nonvolatile cache as much as possible, the required amount of the nonvolatile memory can be reduced, and the cost of the system can be reduced. Further, since the memory portable unit 210 is provided, it becomes possible to make a cache selection judgment by adding “portability” to the attribute of the cache.

[0094]

According to the cache selection method and the data processing system of the present invention, when storing the contents generated by a certain process in any of a plurality of caches, a volatile cache and a nonvolatile cache are selectively used. Can be used properly. In addition, it is possible to selectively use each cache of different devices. Further, a plurality of caches can be selectively used according to various criteria. As a result, the cost of the system can be reduced. In addition, transfer overhead can be reduced, and system performance can be improved. Further, a plurality of caches can be used efficiently with an effective use suitable for each.

[Brief description of the drawings]

FIG. 1 is a block diagram of a disk array system according to a first embodiment.

FIG. 2 is an explanatory diagram of a management information table according to the first embodiment.

FIG. 3 is a flowchart of a process of selecting a cache for storing read data according to the first embodiment;

FIG. 4 is a continuation of the flowchart of FIG. 3;

FIG. 5 is an explanatory diagram showing a parity generation operation.

FIG. 6 is a flowchart of a cache selection process for storing a parity or a difference in the write-through method according to the first embodiment;

FIG. 7 is a flowchart of a cache selection process for storing write data in the write-back method according to the first embodiment;

FIG. 8 is a block diagram of a disk array system according to a second embodiment.

FIG. 9 is an explanatory diagram of a management information table according to the second embodiment.

FIG. 10 is a block diagram of a disk array system according to a third embodiment.

FIG. 11 is an explanatory diagram of a management information table according to the third embodiment.

FIG. 12 is a block diagram of a disk array system according to a fourth embodiment.

[Description of Signs] 1, 2, 3, 4 Disk Array System 100 Host System 101 MPU 103 Main Memory a 104 MPU Bus 105 Primary PCI Bus 106 System ROM 200 Accelerator Board 203 Memory Interface (IF) 204 Nonvolatile Memory b 205 Volatile Memory c 206 Secondary PCI bus 210 Memory portable unit 211 Memory connection means 300 Disk group 301 Disk device 302 SCSI bus 310 Disk interface (I
F) unit 312 Disk memory interface (IF) 313 Nonvolatile memory d 314 Volatile memory e 320 Disk main part 330 Disk internal bus 401 Transfer device 402 XOR operation device 500 Read / write request use area 501 Volatile cache a 502 Nonvolatile Cache b 503 Volatile cache c 504 Nonvolatile cache d 505 Volatile cache e

Continuing on the front page (72) Inventor Yasunori Kaneda 1099 Ozenji Temple, Aso-ku, Kawasaki City, Kanagawa Prefecture Inside the Hitachi, Ltd.System Development Laboratory (72) Inventor Masahiro Takano 2880 Kozu, Kozu, Odawara-shi, Kanagawa Storage System Hitachi, Ltd. Within the business division (72) Inventor Ikuya Yagisawa 1099 Ozenji Temple, Aso-ku, Kawasaki City, Kanagawa Prefecture, Japan Hitachi Systems Ltd.System Development Laboratory

Claims (13)

[Claims] When storing contents generated by a certain process in one of a plurality of caches, a cache for storing the contents is selected based on volatile / nonvolatile information of each cache. Cache selection method. 2. A cache selecting method, wherein when a content generated by a certain process is stored in any of a plurality of caches, a cache for storing the content is selected based on attribute information of each cache. . 3. When a content generated by a certain process is stored in any of a plurality of caches, a cache for storing the content is selected based on information on the capacity of each cache. Method. 4. When a content generated by a certain process is stored in any one of a plurality of caches, a cache which minimizes a span of a data transfer path of a related transfer path is selected. A cache selection method characterized by storing contents. 5. When storing the contents generated by a certain process in any of a plurality of caches, a cache for storing the contents is selected based on positional information on the system configuration of each cache. Cache selection method to be used. 6. When a content generated by a certain process is stored in any one of a plurality of caches, a cache for storing the content is selected based on how the cache is used in the process or another process. A method for selecting a cache, comprising: 7. A cache, wherein when a content generated by a certain process is stored in one of a plurality of caches, a cache for storing the content is selected based on information on the free space of each cache. Selection method. 8. In a data processing system having a plurality of caches, when contents generated by a certain process are stored in any of the plurality of caches, the contents are stored based on volatile / nonvolatile information of each cache. A data processing system, comprising: a cache selecting unit that selects a cache for storing the data. 9. A data processing system having a plurality of caches, wherein when a content generated by a certain process is stored in any of the plurality of caches, the cache stores the content based on attribute information of each cache. A data processing system comprising: a cache selection unit for selecting a data item. 10. The data processing system according to claim 8, further comprising an attribute information acquisition unit for obtaining volatile / nonvolatile information or attribute information of each cache. . 11. A data processing system having a plurality of caches, wherein when a content generated by a certain process is stored in any of the plurality of caches, a data transfer path of a related transfer path is performed. A data processing system comprising: a cache selecting unit that selects a cache that minimizes spread. 12. In a data processing system having a plurality of caches, when storing contents generated by a certain process in any one of the plurality of caches, the contents are stored on the basis of positional information on the system configuration of each cache. A data processing system, comprising: a cache selecting unit that selects a cache for storing the data. 13. The data processing system according to claim 12, further comprising a position information acquisition unit that obtains position information of each cache.