JPH08221216A - Storage system - Google Patents

Abstract

(57) [Summary] [Objective] In an array type storage device system in which data is distributedly stored in a plurality of storage devices, the purpose is to increase the data storage capacity, or to improve the data transfer rate, or a combination thereof. To add a storage device in units of one storage device, allow the storage system to be added while the system is operating, and make the capacity of the additional storage device available immediately after the addition. [Configuration] In an array-type storage device system, a control device is provided with a data distribution pattern setting function, a data relocation function, a data swap function, a boundary management function, a cache control function, an input / output control function, and redundancy. It has a data generation function and a cache memory.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an array type storage device system used for a computer system, which stores a group of data such as files in a plurality of storage devices in a distributed manner. The present invention also relates to a storage device system that can dynamically add storage devices for the purpose of increasing the capacity without stopping the system.

[0002]

2. Description of the Related Art A disk array system is generally used as an array type storage device system. A disk array system is a computer system that allows multiple disk devices to operate in parallel by
This is a system that achieves high-speed data transfer speed and high reliability between a processing device and an external storage device.

As an example of the configuration of the disk array system, for example, a paper by D. Paterso of Paterson et al.
n, G. Gibson, R. Katz, ”A Case for Redundant Array
s of Inexpensive Disks (RAID) ”, ACM SIGMOD conferen
ce porcedings, 1988, pp.109-116). In this paper, RAID (Redundunt Array of
It presents a form of disk array called Inexpensive Disks. RAID stores normal input / output data in a distributed manner in a plurality of drives (disk storage devices) and records redundant data. The setting of the redundant data has a purpose of enabling the lost data to be restored by using it together with the data restoration function when the data is lost.

The addition of drives in the disk array system is performed by adding a plurality of data-distributed drive groups as a unit. In this method, the additional area can be treated as a single continuous area, the data movement is not essential only by rewriting the address mapping information, and the fetch processing after adding the storage device is easy. In other words, the added area can be taken in by adding it after the maximum address of the storage area before the addition.

In addition, in the case of adding a drive for changing the width of data distribution as an expansion of one storage device, especially for dynamic expansion without stopping the system, only the data stored before the expansion of the storage device is buffered. A method has been proposed which is realized by rearranging using.

[0006]

In the conventional expansion method,
When the storage unit is expanded by one, only the data stored before the expansion is rearranged. That is, relocation of data in the storage area of the added storage device is not considered. Since the relocation cannot be performed on all the data at once, the storage area of the added storage device is overwritten by the relocation and the corresponding storage area becomes unusable. Therefore, the storage area of the added storage device cannot be used immediately after the addition, and only after the data relocation processing is completed.
There was a problem that the expanded storage capacity could not be used. Further, if the expanded storage device had data from the beginning, this data could not be used at all.

When the additional storage device is used without rearrangement, the performance of the data transfer rate cannot be improved, the reliability of the data in the additional portion is lowered, and the logical continuity and physical However, there is a problem that the continuity of different data does not match.

Therefore, the technical problem to be solved by the present invention is as follows.
In solving the problems of the above-mentioned conventional technology,
The purpose thereof is to make it possible to use the expanded storage capacity immediately after the expansion even when the storage devices are individually expanded, and to dynamically realize the expansion without stopping the system.

[0009]

The above-mentioned object is to relocate the data on the added storage device in parallel with the relocation of the data stored before the extension and to further logically determine the position of the data. This can be achieved by performing rearrangement in consideration of continuity.

That is, a cache having a function of receiving a storage device addition request and a data access request from the processing device, a function of recording data read from the storage device and output data from the processing device, and holding the data in the cache. A cache control function, a data distribution pattern setting function that sets a data distribution pattern for a storage device group that includes an additional storage device, and a specified amount of data of n storage devices that existed before the expansion determined by the number of storage devices and redundancy. A function to read using the old data distribution pattern, a function to read the data of the added k storage devices by a specified amount determined by the number of storage devices and the redundancy, and a function to create redundant data according to the new data distribution pattern The data read from the storage device before the expansion is newly added to the area where the data was read. Data distribution pattern writing function, the function to write the data read from the storage device added to the area after this function output with the new data distribution pattern, and control the progress of data relocation to distribute new / old data This can be realized by the function of recording the boundary of the pattern and the function of swapping the data area to the original position after rearranging the data.

Further, in order to dynamically change the data distribution pattern, parallel processing of normal data input / output processing and data distribution pattern changing processing is required, which can be realized by the following means.

That is, when there is a read request from the processing device for the data recorded in the cache, the data is transferred from the cache to the processing device without being read from the storage device and is also recorded in the cache. When a write request is issued from the processing device for the stored data, the cache that processes the access request from the processing device by directly writing the output data from the processing device in the cache without writing it directly to the storage device. This can be realized by the hit function and the input / output control function that inputs / outputs data to / from the storage device by applying the specified data distribution pattern.

[0013]

In the data distribution pattern setting function, the number of storage devices to be added (k) and the number of storage devices originally included in the system (the number of storage devices originally read from the control memory ( n) and system redundancy (r) information are read, a new data distribution pattern is determined and recorded in the control memory. The data distribution pattern before the expansion is recorded as the old data distribution pattern. Then, the data relocation function is activated.

The data relocation function uses the input / output control function to read a specified amount (n + k−r) of data from each storage device including the expanded storage device and record it in the cache.

By using this data and the redundant data creating function, redundant data according to the new data distribution pattern is created and recorded in the cache.

The data relocation function utilizes the input / output control function to output the read data and the newly created redundant data in a new data distribution pattern from the beginning of the relocation area of the storage device.

The boundary management function distinguishes the relocation-incomplete area, the relocation-completed area, and the relocation-in-progress area as the data relocation progresses, and determines which data distribution pattern the I / O control uses to access. Give information to do.

In the relocation processing, the data of the specified value determined by the number of storage devices and the redundancy is processed as a processing unit (relocation group).
It means that the processing is completed within the processing unit.

In a system having a redundancy of r, k is stored in n storage devices.
When adding one storage device, (n + k-
r) is set, and a set of n × (n + k−r) logically continuous data blocks recorded in the n storage devices before the expansion and k × (k (s) of the k storage devices to be added are set. n + k-
r) combined with a set of consecutive data blocks,
Define as a relocation group. The relocation process is completed within this relocation group.

The rearrangement processing of the n × (n + k−r) data blocks recorded in the n storage device groups before the expansion increases the data distribution width from n to (n + k).

The old redundant data block to be deleted is (n + k−r) × r block, the newly created redundant data block is (n−r) × r block, and the total number of data blocks = n × (n + k−r) − ( n + k−r) × r + (n−r) × r = (n + k) · (n−r).

This data has a data distribution width of (n +
k), the number of distribution blocks for each storage device is the same (n−r) for each storage device, so that the data can be relocated and no data for which redundant data is not set does not occur.

Next, k × (n + k−) of the storage device to be added is added.
The rearrangement process of r) data blocks changes the data distribution width from 1 to (n + k).

No old redundant data blocks to delete,
The number of redundant data blocks to be newly created is k × r, and the total number of data blocks = k × (n + k−r) + k × r = (n + k) × k.

This data has a data distribution width of (n +
k), the number of distribution blocks for each storage device is equal to k for each storage device, so that the storage blocks can be rearranged and no data for which redundant data has not been set does not occur.

The total of the above two data blocks is the same as the setting of the relocation group initially set. That is, (n + k) * (n-r) + (n + k) * k = (n + k) * (n + k-r).

A storage device having a data distribution width of (n + k),
The number of data blocks per storage device is (n + k−r). That is, all the data blocks can be rearranged and stored in the original area, and new redundant data can be set for all the data. This is also applicable when r ≧ 2 and r = 0.

The swap function maintains a partial disruption of the logical continuity of data due to the relocation processing in units of relocation groups. The data is searched according to the logical continuity of the data, and when the logical continuity is broken, the data at the correct position and the place are swapped (swapped). The swap function searches according to the logical continuity of the data, and if the data is replaced with the data at the correct position, the situation that the correct position is not corrected or the swap function does not end does not occur.

The cache control function retains the data input by the relocation function in the cache until the relocation function outputs, and the cache hit function processes the access request with the data in the cache. As a result, it is possible to prevent unauthorized data access by directly accessing the storage device in response to an access request for data in the reallocation area.

With these functions, the pattern of data distribution can be changed.

The request accepting function that accepts the access request from the processor requests the cache hit function if there is data in the cache, and requests the input / output control function to execute the access request if there is no data in the cache. .

The input / output control function checks whether or not the data relocation is being processed based on the information provided by the boundary management function. If the data position is not in the relocation and is before the relocation area, If a new data distribution pattern is selected and the data position is not in the rearrangement and is after the rearrangement area, the old data distribution pattern is selected and the input / output processing with respect to the storage device is performed. Data in the relocation area is processed by the cache hit function.

[0033]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a configuration diagram of an array type storage device system according to an embodiment of the present invention applied to a computer system. The system of FIG. 1 includes a processing device 1001 (including a CPU) that issues an input / output request, a disk array control device 1010 that receives the input / output request, and a plurality of drive devices (storage devices) 1021 to 1100 that record data.
It is composed of 025. This embodiment is applicable to a system having four drive devices 1021 to 1024 before expansion.
This is an application example when a single drive device 1025 is added.

The disk array control device 1010 is a cache I / F (interface) 1011 for controlling input / output with the processing device (host device) 1001, and a cache memory (cache) for temporarily storing input / output data and the like.
1012, a control memory 1 for storing various control data, etc.
003 and a drive I / F (interface) 1014 that controls input / output with each drive device, and includes various functional units to be described later. The disk array control device 1010 performs various processes such as processing of access requests from the processing device 1001, input / output processing with the drive device, generation of redundant data, and restoration of data using redundant data.

In this embodiment, data is initially recorded in the four drive devices 1021 to 1024. The data is distributed to each drive device in block units, and a parity data block of one block is attached for every three blocks. This parity data (redundant data) is generated from the data of the target 3 blocks, and has the degree of redundancy that can be restored even if any 1 block of data is lost. P in Figure 1
The blocks of redundant data indicated by the symbol are stored in a distributed manner in the plurality of drive devices 1021 to 1024 in order to avoid concentration of access to a specific drive device.

FIG. 2 shows a disk array controller 1010.
It is a block diagram which shows the detail of FIG. As shown in the figure, the disk array control device 1010 includes a data distribution pattern setting function 2010, a data relocation function 2020, a redundant data creation function 2030, and a data restoration function 2040.
Data distribution pattern boundary management function 2050, cache control function 2060, access request reception function 2070
A cache hit function 2080, an input / output control function 2090, a swap function 2100, and a host I / F 1
001, cache 1012, and control memory 100
3 and a drive I / F 1014.

FIG. 6 is a diagram showing boundary control associated with data rearrangement. The rearrangement completion pointer 6012 indicates the final address of the rearrangement completion area 6020 where data rearrangement has already been completed and data can be accessed by the data mapping information 11000 (FIG. 11) after the drive device is added.
The relocation non-execution pointer 6011 indicates the start address of the relocation non-execution area 6040 that has not yet been relocated. This area 6040 can be accessed by the data mapping information 10000 (FIG. 10) before the drive device is added. These pointers are the control memory 10
Record at 13. Relocation area 603 sandwiched between relocation completion pointer 6012 and relocation unexecuted pointer 6011
0 is in the reallocation process, and all access requests from the processing device 1001 to this area 6030 are processed by accessing the cache 1012 and do not access the drive device. The address in FIG. 6 corresponds to the logical address in the data mapping information.

FIG. 8 shows the pointer status in the swap processing. The swap pointer 8011 indicates the address of the swap processing target data. Swap destination pointer 801
Reference numeral 2 is a pointer indicating the swap destination address 10030 of the data mapping information 11000. In the swap processing, the data 8021 and 8022 at the address indicated by the swap pointer 8011 and the swap destination pointer 8012 are exchanged. The address in FIG. 8 corresponds to the logical address in the data mapping information.

FIG. 10 is a diagram showing a part of the data mapping information of the data distribution pattern before the expansion, and FIG.
FIG. 6 is a diagram showing a part of data mapping information of a data distribution pattern after expansion. These are set on the control 1013 memory by the data distribution pattern setting function 2010. The logical address 10010 is an address in which data blocks stored in the disk array control device 1010 are sequentially numbered. However, with respect to the parity data, the numbering of a system different from that of the data other than the parity data is performed. The physical address 10020 indicates in which part of which drive device the data block of the logical address 10010 is actually stored. The column number 10021 is a number numbered in ascending order from the first data block in the drive device.
2 is a number for uniquely identifying the drives 1021 to 1025 managed by the disk array control device 1010. The swap destination address 10030 is the column number 10021 of the physical address that finally stores the data in the swap processing.

Next, referring to FIG. 3, the drive unit 102
Data rearrangement when 5 is added will be described.

Since data is recorded also in the added drive unit 1025, the drive units 1021 to 1024
The data already stored in the drive unit 1025 and the data recorded in the drive device 1025 are stored in the rearrangement groups 4010, 4
020, ...
Rearrange 21 to 1025.

The data distribution pattern setting function 2010 which has accepted the addition of the drive device sets a data distribution pattern with a data distribution width of 5 drives and a redundancy of 1 and stores it in the control memory 1013. Also, the relocation group 40
11, 4021 ... (FIG. 4) are set, and thereafter control is passed to the data relocation function 2020.

The relocation processing of the data relocation function 2020 will be described below with reference to the processing flow of FIG.

First, at step 5010, the data distribution pattern setting function 2010 acquires the data mapping information 10000 before the expansion, the data mapping information 11000 after the expansion and the information of the relocation group size recorded in the control memory 1013. .

At the initialization step 5020, the relocation non-execution pointer 6011 and the relocation completion pointer 6012 are set to the head address of the drive.

Read out relocation group step 5030
Then, from each drive device including the additional drive device 1025, the data 3011 (FIG. 3) of four consecutive blocks from the address indicated by the relocation unexecuted pointer 6011 is read and recorded in the cache 1012.

In boundary adjustment step 5040, step 5
The relocation non-execution pointer 6011 is advanced to the final address of the data read at 030.

In the redundant data deletion step 5050, the cache 1012 is read from the data read in step 5030 based on the data mapping information 10000 before the expansion.
The parity data recorded in is deleted from the cache.

In the parity data creating step 5060, new parity data 3021 (see FIG. 3) is created from the data 3031 (FIG. 3) remaining after the redundant data is deleted in step 5050 and the data mapping information 11000 after the expansion using the redundant data creating function 2030. Create 3). The created parity data 3021 is stored in the cache 1012.

In the data output step 5070, the input / output control function 2090 outputs the data 3031 and the parity data 3021. Wait for the output to complete before proceeding to the next step.

At the boundary adjustment step 5080, the relocation completion pointer 6012 is set to the final address of the data output at step 5070, the data mapping information 10000 is retrieved from the logical address of the relocated data, and the physical address 10020 and the swap destination are swapped. Address 10030
Set.

At the end determination step 5090, if the relocation completion pointer 6012 matches the final address of the drive device, the relocation is finished. If they do not match, the process returns to step 5030 and the above-described processing is repeated.

FIG. 4 is a diagram showing in detail the data arrangement before and after the rearrangement. The relocation is performed by the relocation group 4010,
Since it is executed in units of 4011, logical address "9" which should originally be continuous on the drive device, as shown in the leftmost column of the data arrangement before and after the rearrangement in (2) of FIG.
And the data of "13" exists the data of the logical address "a". Further, the data of the logical addresses "a" and "e" are not continuous. This situation is solved by swap processing.

Next, the swap processing will be described with reference to the flow of FIG.

First, in step 7010, the data mapping information 11000 after the addition is acquired.

At step 7020, the swap pointer 8
011 is set to the start address of the drive.

At decision step 7030, data mapping information 1 of the address indicated by swap pointer 8011
Physical address 10020 by referring to 1000 entries
The column number 10021 of the above is compared with the swap destination address 10030. If they match, no swap is required, and the process advances to step 7080. If they do not match, swapping is necessary, and the process advances to step 7040.

At step 7040, the data at the address indicated by the swap pointer 8011 is read and stored in the cache 1012.

In step 7050, the swap destination address 10030 is set in the swap destination pointer 8012,
The data at that address is read out and cache 1012
To memorize.

In step 7060, the data of the address indicated by the swap pointer 8011 is output to the address indicated by the swap destination pointer 8012, and the data of the address indicated by the swap destination pointer 8012 is output by the swap pointer 8.
Output to the address indicated by 011.

In step 7070, the physical address 10020 and swap destination address 10030 changed by swap are written in the data mapping information 11000.

In step 7080, the swap pointer 8011 is advanced to the next logical address.

At decision step 7090, if the swap pointer 8011 matches the final address of the drive device, the swap ends. If they do not match, step 7
The above-mentioned processing is repeated from 030.

FIG. 9 is a diagram showing how the data position changes according to the flow of swap processing. In step 1, the data from column number “1” to “3” is
There is no need for swapping, so it remains the same. Since the swap processing is performed in the order of the logical addresses, the data of the column number “5” becomes the swap processing target after the column number “3”.
The data 9015 of the column number “5” moves to the position of the column number “4” which should be originally stored. As a result, column number "4"
The data 9014 which has been moved to the position of the column number “5”. In step 2, the data 9016 of column number “6”
And the data 9014 of the column number “5” are swapped.
In step 3, the data 9017 having the column number “7” and the data 9014 having the column number “6” are swapped. Furthermore,
The data of the column numbers “7” and “8” are left as they are because there is no need for swapping.

As described above, according to this embodiment, when one drive is added, the data existing before the addition and the data of the added drive device are simultaneously rearranged, and the storage location is corrected by the swap process. By doing so, it is possible to dynamically add one drive unit, and the storage capacity of the added storage device can be used immediately after the addition.

Further, since the data of the rearrangement group are collectively read out to the cache, the number of times of input / output with the drive unit accompanying the data rearrangement processing is reduced, and the time required for the rearrangement processing can be shortened.

Further, when the drive device is expanded, the parity data in the data distribution pattern before the expansion is also read out. Therefore, when a part of the data to be read cannot be read out due to the total failure or partial failure of the drive device. The data in the cache can be used to recover the data without the need to input / output new data.

[0068]

As described in detail above, according to the present invention, it is possible to dynamically relocate data, and one storage device can be used instead of adding a plurality of storage device groups for distributing data. The unit can be expanded, and the cost associated with the expansion of storage capacity can be reduced. Also, by adding storage devices,
Since the number of storage devices for distributing data can be changed, it is possible to improve the data transfer rate by adding storage devices.

Furthermore, the capacity of the added storage device can be used without waiting for the completion of the data relocation processing.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an array type storage device system according to an embodiment of the present invention applied to a computer system.

FIG. 2 is a block diagram showing a configuration of a disk array control device in FIG.

FIG. 3 is an explanatory diagram schematically showing a data flow in a data rearrangement process when a drive device is added in one embodiment of the present invention.

FIG. 4 is an explanatory diagram showing an example of data arrangement before and after rearrangement in one embodiment of the present invention.

FIG. 5 is a flowchart showing a rearrangement processing flow in the embodiment of the present invention.

FIG. 6 is an explanatory diagram showing an example of area division by a rearrangement process according to an embodiment of the present invention.

FIG. 7 is a flow chart diagram showing a swap processing flow in one embodiment of the present invention.

FIG. 8 is an explanatory diagram showing an example of a swap pointer and a swap destination pointer in one embodiment of the present invention.

FIG. 9 is an explanatory diagram showing an example of how a data position is changed by swap processing in an embodiment of the present invention.

FIG. 10 is an explanatory diagram showing an example of data mapping information before expansion in one embodiment of the present invention.

FIG. 11 is an explanatory diagram showing an example of data mapping information after expansion in one embodiment of the present invention.

[Explanation of symbols]

 1001 Processing device 1010 Disk array control device 1012 Cache memory (cache) 1013 Control memory 1021 to 1025 Drive device 2010 Data distribution pattern setting function 2020 Data relocation function 2030 Redundant data creation function 2040 Data restoration function 2050 Data distribution pattern boundary management function 2060 Cache control function 2070 Access acceptance function 2080 Cache hit function 2090 I / O control function 2100 Swap function 6011 Relocation unexecuted pointer 6012 Relocation completion pointer 6020 Relocation completion area 6030 Relocation in progress area 6040 Relocation unexecuted area 8011 Swap pointer 8012 Swap destination pointer 10000, 11000 Data mapping information 10010 Logical address 1 020 physical address 10021 column number 10022 drive number 10030 swap destination address

Claims (3)

[Claims] 1. A plurality of storage devices are controlled by a control device, data and redundant data from a processing device are distributed and stored in one or more storage devices, and in response to a data read request from the processing device, If necessary data is read from one or more storage devices and transferred to the processing device, and if the data on the storage device cannot be directly accessed due to a failure of the storage device or the like, the control device restores the inaccessible data using redundant data. In an array-type storage device system that has the function to increase the data storage capacity or increase the data transfer rate, add one or more storage devices under the control device and import the added storage devices. , Read a specified amount of data from the storage device that stores data before the expansion and a specified amount of data from the expanded storage device to the control device Data rearrangement by means, means for newly creating redundant data according to the redundancy of the storage device system, and means for writing the read data and the new redundant data in all the added storage devices. The logical position of the data and the means for managing the logical position of the data and the position recorded in the storage device, and the means for exchanging (swap) the data contents between the data recorded in the storage device. For the access request from the processing device, the means for storing the write position, the means for comparing the access position of the access request with the write position, and the data access based on the comparison result. And a means for determining a data distribution pattern used for the storage device system. 2. A control device controls n (n ≧ 1) storage devices, and data from a processing device and a redundancy r (n> r ≧).
The redundant data of 0) is distributed and stored in one or more storage devices, and in response to a data read request from the processing device, necessary data is read from the one or more storage devices and transferred to the processing device. When the data on the storage device cannot be directly accessed due to a failure of the storage device or the like, in the storage device system of the array type having the function of restoring the inaccessible data by using the redundant data in the control device, k under the control device is used. Regarding the expansion of (k ≧ 1) storage devices, a total of n × (n + k−r) (n + k−r) from each of the n storage devices that store data before the expansion.
And a means for handling a total of k × (n + k−r) pieces of data from each of the data pieces and the added k pieces of storage equipment and treating the data as a rearrangement group, and a storage device. According to system redundancy (n +
means for newly creating r redundant data from (k−r) data, and the above-mentioned n in the (n + k) storage devices after expansion.
Means for relocating data by means of writing × (n + k−r) pieces of data and the new redundant data of r pieces, and managing the logical positional relationship of the data and the position recorded in the storage device. And a means for exchanging (swapping) the data contents between the data recorded in the storage device, the logical positional relationship of the data is adjusted, and the write position is stored for the access request from the processing device. And a means for comparing the access position of the access request with the write position, and a means for deciding a data distribution pattern used for data access from the comparison result. 3. The storage device system according to claim 1, wherein the reading of redundant data is suppressed when reading the data stored before the addition of the storage device.