JPH09237165A - Disk array device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve data block reading speed even at the time of having variation in the reading time of a data block for respective storage units. SOLUTION: A reading instruction issuing means 104 issues a reading instruction. Storage devices 201 and 202 which receives the instruction transmit the read storage units to a controller 100. A first judging means 105 judges that the prescribed number of storage units are read within the storage unit number obtained by adding one to a prescribed number. When it is judged the prescribed number of storage units are read, the second judging means 105 judges whether or not parity block is included in the storage unit. When the parity block is included, an instruction means 105 instructs a decoding means 106 to generate a data unit which is not read. The decoding means generates the data unit and generated and outputs the data block.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk array device.

[0002]

2. Description of the Related Art The amount of data handled by information equipment has increased,
A large capacity storage device is required. As such a large-capacity storage device, there has conventionally been a disk array device in which a plurality of single magnetic disk devices are combined. Since such a disk array device is often shared by a plurality of users, it is necessary to improve fault tolerance and ease of maintenance. Therefore, in recent years, for example, RAID (Redundant Arrays of
A system represented by Inexpensive Disks has come to be used for the disk array device.

The concept of RAID was announced by a research team at the University of California, Berkeley, USA (DA
Patterson, et.al, "A Case for Redundant Arrays of I
nexpensive Disks (RAID) ", ACM SIGMOG CONFERENCE (198
8)). According to this, five types of methods have been proposed according to the three combinations of availability indicated by fault tolerance and ease of maintenance, performance indicated by data transfer rate and required rate, and cost. Among them, the method called RAID 5 is generally used. In this method, when writing a data block, the data block is converted into a plurality of storage units each having a sector length of the disk device and a parity block, which are distributed and stored in the plurality of disk devices. on the other hand,
The reading of the data block is performed by reading all of the plurality of storage units corresponding to one data block stored in the disk device and generating the data block by using all the read storage units. It is reading.

[0004]

By the way, due to recent technical innovation, real-time data such as moving images can be handled by other information devices, and further improvement in performance is demanded even in the disk array device. is there.
Especially for applications such as a moving image server such as VOD (Video On Demand), it is necessary to further improve the read speed of the disk array device.

However, the conventional disk array device has a problem in improving the read speed for the following reason. Since the seek time and the rotation waiting time of the magnetic disk device alone, which affects the reading speed of the data block, depend on the storage position of each storage unit of the data block and the magnetic head position, and thus are often different for each magnetic disk device. . Therefore, the reading time of each storage unit of the data block often varies. on the other hand,
The disk array device waits until all the storage units in the data block have been read, and then creates the data block to read the data block. Therefore, if the reading of one storage unit is delayed with respect to the other storage units, the data block to which the storage unit belongs cannot be generated, and the reading of the data block is delayed.

Therefore, it is an object of the present invention to provide a disk array device capable of improving the read speed of a data block even when the read time of each storage unit of the data block varies.

[0007]

In order to solve this problem, the present invention provides a read command issuing means which, when reading a data block, issues a read command which also reads a parity block. Furthermore, it is determined that a predetermined number of storage units have been read from the plurality of storage devices by the control device, and that the first determination unit has read the predetermined number of storage units. In the case where the storage unit includes a parity block, the second determination unit determines whether the storage unit includes a parity block, and when the second determination unit determines that the parity block is included, the reading is performed. The data unit that has not been issued is regarded as a missing data unit, and an instruction unit that instructs the decoding unit to generate the data unit is provided.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The invention according to claim 1 is provided with a plurality of storage devices and a control device for controlling the storage devices, and the control device writes two or more data blocks when writing the data blocks. Divided into a predetermined number of data units,
And a coding unit for performing a coding process for generating a missing data unit generation parity block and further distributing and storing each of the predetermined number of data units and parity blocks as a storage unit in the plurality of storage devices. When reading a data block, a read command issuing means for issuing a read command for reading the predetermined number of data units to the plurality of storage devices, and the predetermined number of read units according to the read command. And a decoding means for performing a decoding process for generating one data block from the data unit, wherein the encoding means and the decoding means include one data unit out of a predetermined number of data units, A data unit in which the encoding means is missing by using a parity block and another read data unit. A disk array device using an encoding method capable of generating a complete data block by generating the read command issuing means, when reading the data block, further reading a parity block. The control device further includes a first determination unit that determines that a predetermined number of storage units have been read by the control device from the plurality of storage devices, and a first determination unit. When it is determined that a predetermined number of storage units have been read, a second determination unit that determines whether or not a parity block is included in the storage unit, and the parity block by the second determination unit When it is determined that the data unit that has not been read out is regarded as a missing data unit, the data is read by the decoding means. Characterized by comprising an instruction unit that instructs to generate a knit,
The following actions are performed. The read command issuing means issues a read command not only to the data unit but also to the parity block. The storage device that has received the read command sends the read storage unit to the control device. The first determination means of the control device determines that a predetermined number of storage units have been read out of the storage units of one number added to the predetermined number stored in the plurality of storage devices. The second determining unit determines whether or not a parity block is included in the storage unit when the first determining unit determines that a predetermined number of storage units have been read. When the second determining means determines that the parity block is included, the instructing means regards the data unit that has not been read as a missing data unit, and causes the decoding means to identify the missing data unit. Instruct to generate. The decoding means that has received the instruction generates the unread data unit, then generates the data block, and outputs it to the outside.

According to a second aspect of the present invention, the storage device is a magnetic disk device, and the encoding means stores a storage position of each storage unit corresponding to one data block stored in the plurality of storage devices. A storage position calculating unit that calculates a different position for each storage device, and a storage unit that stores each storage unit corresponding to one data block in the storage position calculated by the storage position calculating unit. It is characterized by having the following functions. The storage position calculating means calculates the storage position of each storage unit corresponding to one data block to be stored in the plurality of magnetic disk devices so that the storage position is different for each magnetic disk device. Therefore, the storage means stores the storage position of each storage unit corresponding to one data block differently for each magnetic disk device according to the calculated storage position.

According to a third aspect of the present invention, the magnetic disk device includes a cache memory for temporarily storing the storage unit read from the magnetic disk unit before reading the storage unit, and the control device further stores Prior to reading a unit, a cache memory read request issuing means for issuing a cache memory read request for temporarily storing a storage unit in a cache memory, and a predetermined number of storage units temporarily stored in the cache memory of the magnetic disk device. The cache memory read determination means and the read command issuing means determine the magnetic disk at the time when the cache memory read determination means determines that a predetermined number of storage units are temporarily stored in the cache memory of the magnetic disk device. Issue a read command to the device Characterized the door, an action described below. The cache memory read request issuing means issues a cache memory read request prior to reading a storage unit. The cache memory read determination means determines that a predetermined number of storage units have been temporarily stored in the cache memory of the magnetic disk device. The read command issuing means issues a read command to the magnetic disk device when the cache memory read determining means determines that a predetermined number of storage units have been temporarily stored in the cache memory of the magnetic disk device.

According to a fourth aspect of the present invention, the control device further comprises the cache memory read determination means when the cache memory read determination means determines that a predetermined number of storage units have been temporarily stored in the cache memory. The cache memory read request canceling means for canceling the issued cache memory read request for the magnetic disk device in which the memory unit is not stored in the memory is provided, and the following operation is performed.
The cache memory read request canceling unit is provided to the magnetic disk device whose storage unit is not stored in the cache memory when the cache memory read determining unit determines that a predetermined number of storage units are temporarily stored in the cache memory of the magnetic disk device. On the other hand, the issued cache memory read request is canceled.

According to a fifth aspect of the present invention, the disk array device specifies a data block to be read by using an ID number corresponding to the data block in a one-to-one correspondence. Is a pseudo-random number calculating means for calculating a pseudo-random number corresponding to the ID number and the storage area of the magnetic disk device and having a different value for each magnetic disk device; A storage position calculation instruction that instructs the pseudo random number calculation means to calculate a pseudo random number for each magnetic disk device, and uses the calculated pseudo random number as the storage position of each storage unit corresponding to the data block specified by the ID number. It is characterized by having means, and has the following operation. The pseudo-random number calculating means calculates a pseudo-random number that corresponds to the ID number and the storage area of the magnetic disk device and has a different value for each magnetic disk device. The storage position calculation instruction means is an ID
The pseudo-random number calculation means is instructed to calculate a pseudo-random number for each of the plurality of magnetic disk devices based on the number, and the calculated pseudo-random number is set as a storage position of each storage unit corresponding to the data block specified by the ID number. To do.

According to a sixth aspect of the present invention, the pseudo random number calculating means calculates one pseudo random number corresponding to an ID number,
The storage position calculation instructing means is characterized in that the phase of the one pseudo random number is changed to set the storage position of each storage unit corresponding to the data block specified by the ID number, and has the following operation. The pseudo random number calculating means calculates one pseudo random number corresponding to the ID number. The storage position calculation instructing means changes the phase of the one pseudo-random number and sets it as the storage position of each storage unit corresponding to the data block specified by the ID number.

(Embodiment) FIG. 1 is a block diagram of a disk array device according to an embodiment of the present invention. The disk array device is composed of a control device 100 and five magnetic disk devices 201, 202, ... 205. The control device 100 includes an encoding unit 101, a storage position calculating unit 102, a writing unit 103, a reading unit 104, a determination instructing unit 105, and a decoding unit 1.
And 06. The control device 100 writes / writes a data block identified by an ID number from the outside.
According to the read request, each magnetic disk device 201, 2
02, ..., 205 are controlled. Here, the ID number is, for example, a continuous integer value from 0 to the maximum value M-1.
The configuration of the control device 100 is almost the same as the configuration of a general RAID 5 disk array device. The difference is that the functions of the reading means 104 and the decoding means 106 and the determination instruction means 105 are newly provided.

The encoding means 101 divides the data block identified by the ID number received from the outside into four data units. Further, the parity block for generating the missing data unit is calculated by calculating the exclusive OR between the divided data units bit by bit. The size of the calculated parity block is the same as the size of the divided data unit. Therefore, assuming that the data unit is one storage unit, a total of five storage units are generated by adding the parity block to the data unit. The parity block may be either odd parity or even parity. Further, the size of the data block received at the time of a write request from the outside is always fixed and is the size of four data units.

The storage position calculating means 102 is an encoding means 1.
A total of five storage units of the data unit and the parity block obtained in 01 are stored in five magnetic disk devices 20.
The storage position in each magnetic disk device 201 to be distributed and stored in 1, 202, ... 205 is calculated. For example, the storage position in each magnetic disk device is obtained as follows. ID of data block received from outside
From the number X, a value A of an expression for generating a pseudo random number is obtained, the value A is divided by the total number M of ID numbers, and the remainder is calculated as a storage position of the magnetic disk device.

A = (αX + β) ... Here, the factors α and β are integers determined for each magnetic disk device, and at least the factor α has a coprime relationship with the total number M of ID numbers, It is an integer that differs for each magnetic disk device. The writing unit 103 divides the data unit and the parity block divided or generated by the encoding unit 101 into the magnetic disk devices 201, 202, ..., Based on the storage positions calculated by the storage position calculating unit 102.
・ Write to 205.

The reading means 104 is provided with an external I
Based on the read request for the data block of the D number, a read command for not only the data units forming the data block but also the parity block is issued to the magnetic disk devices 201, 202, ... 205. The read means 104 uses the ID number of the data block received from the outside to cause the storage position calculation means 102 to calculate the storage position and issue a read command.

The judgment instructing means 105 is based on the read command, and the magnetic disk devices 201, 202, ... 205.
Out of the 5 storage units, the number of read storage units is 4
Wait until you get together. When the number of read storage units is four, it is determined whether the four storage units are three data units and a parity block or four data units.

The decoding means 106 generates an original data block from the read four storage units and outputs it to the outside. The data block generation is specifically as follows. First, if the determination instructing means 105 determines that the four read storage units are the three data units and the parity block, one unread data unit is read as another read data unit. It is generated using three data units and a parity block.
Next, a data block is generated using four data units including one data unit thus generated. Next, when the determination instruction unit 105 determines that the four read storage units are four data units, the four data units are used as they are to generate a data block.

Magnetic disk devices 201, 202, ... 2
05 stores the data unit and parity block of the data block generated by the encoding means 101. Further, it receives a read command of the data unit or the parity block from the read means 104 and sends the data unit or the parity block to the read control device 100. FIG. 2 shows the disk array device configured as described above.
The operation of reading a data block will be described with reference to the flowchart shown in FIG.

First, the ID of the data block to be read
Based on the numbers, the magnetic disk devices 201, 202, ...
.. The storage position calculating unit 102 calculates the storage positions of the data units and parity blocks stored in 205 (step S1). Next, the reading means 104
Issues a read command of a data unit or parity block to each of the magnetic disk devices 201, 202, ... 205 based on the calculated storage position (step S2). The magnetic disk devices 201, 202, ... 205, which have received the read command, send the data units or parity blocks to be read to the control device 100 in the order in which they can be read. The determination instructing means 105 of the control device 100 sequentially receives the data units of 4 units.
It is determined whether all the items have been collected (step S3). If four data units are not available, then it is determined whether or not any three data units out of the four data units and the parity block are available (step S).
4). Here, when it is determined that the three data units and the parity block are aligned, the decoding means 106
One data unit that has not been read is generated (step S5). Specifically, an unread data unit is generated using the read three data units and the parity block. Next, the decoding means 106
Generates an original data block from four data units including the generated data unit (step S
6) The data block is output to the outside (step S7).

On the other hand, when it is determined in step S3 that four data units have been prepared, the decoding means 106 determines that four data units are available, as in the conventional RAID5 disk array device.
An original data block is generated from one data unit (step SS6), and the data block is output to the outside (step S7). As described above, the disk array device according to the present embodiment is the storage unit having the maximum read delay time from the reception of the storage unit read command to the actual output of data among the magnetic disk devices. Since it is not necessary for reading the data block, the reading time of the disk array device can be shortened as a whole. Of course, when one storage unit is lost, which is the original function of the disk array device,
The availability of operating even if one of the magnetic disk devices fails is not lost in this embodiment.

Further, in the present embodiment, a plurality of storage units corresponding to one data block are made to have different storage positions in each magnetic disk device by using a formula for generating a pseudo random number. Remembered in. Therefore, all storage units in the data block are not uniformly and early read, and all storage units are not uniformly and lately read. Therefore, even when reading a plurality of data blocks, it is unlikely that some of the data blocks will be read extremely quickly or some will be read very slowly, and the time will be almost constant, and when reading a plurality of data blocks. It is possible to suppress variations in read time between data blocks.

The storage position calculating means 102 calculates different storage positions for each magnetic disk device by applying five different values of α and β, that is, for each magnetic disk device and applying a mathematical formula. There is one fixed α,
By using the value of β and changing only the phase, different storage positions may be calculated for each magnetic disk device.

Further, the storage position calculating means 102 calculates the storage position by using an expression for generating a pseudo random number, but the storage position calculation means 102 is not limited to this expression, and may be, for example, within a value range of 0 to M-1. You may use the formula which can generate | occur | produce a pseudorandom number with. Further, the storage location may be calculated using a random number table instead of using the mathematical formula. Further, although the magnetic disk device is used as the storage device in the present embodiment, the present invention is not limited to this, and if the delay time of the response to the read request is not fixed, it may be another type of storage device. Good.

(Modification of Embodiment) Further, the magnetic disk device may have a built-in cache memory. The configuration of the disk array device according to this embodiment in that case will be described below. The read means 104 issues a first read request for the data unit and the parity block based on an external read command for the data block. The first read request is a request for instructing each magnetic disk device to read a storage unit into a cache memory included therein.

Magnetic disk devices 201, 202, ... 2
Upon receiving the first read request, 05 sends a first response to the control device 100 upon prefetching, that is, when the storage unit is read into the internal cache memory. The storage unit read to the cache memory is sent to the control device 100 based on a second read request described later. further,
At the time of receiving a removal request to be described later, the first read command for the storage unit that has not sent the first response to the control device 100 is ignored.

The judgment instructing means 105 judges that four first responses from the magnetic disk devices 201, 202, ... 205 have been received, and at that time, the magnetic disk device 20.
The second read command is issued to 1, 202, ... 205. Furthermore, when four first responses are received,
205 is issued to the magnetic disk devices 201, 202, ... 205 corresponding to the remaining one storage unit that has not been read by the cache memory.

Next, the read operation will be briefly described with reference to FIG. First, the storage position calculating means 102 calculates the storage position of the data unit or the parity block based on the ID number of the data block to be read (step S21). The reading means 104 determines the magnetic disk device 20 based on the calculated storage position.
A first read request for a data unit or a parity block is issued to 1, 202, ... 205 (step S22). Upon receiving the first read request, the magnetic disk devices 201, 202, ... 205 send a first response to the control device 100 in the order in which the data units and parity blocks to be read are read into the cache memory. The determination instruction means 105 of the control device 100 sequentially waits until there are four first responses sent (step S23). If there are four first responses, the determination instruction means 1
If 05 determines, the magnetic disk devices 201, 202,
.. Issue a removal request for the remaining one storage unit to 205 (step S24). Next, the determination instruction means 1
Reference numeral 05 designates the second read request as the magnetic disk device 20.
Issued to 1, 202, ... 205 (step S
25). Subsequent processing (steps S26 to S3)
0) is the same as the above-mentioned embodiment.

[0031]

According to the first aspect of the present invention, when the read time of each storage unit of the data block is different, the storage unit that is read last is required to read the data block to which the storage unit belongs. Therefore, the disk array device can be read at high speed as a whole.

According to the second aspect of the present invention, each magnetic disk device stores each storage unit such that the storage position of each storage unit corresponding to one data block differs for each magnetic disk device. . Therefore, it is unlikely that all the storage units in the data block will be uniformly and quickly read out, or that all of the storage units will be uniformly and slowly read out, and the time will be almost constant, and when reading a plurality of data blocks It is possible to suppress variations in read time between data blocks.

According to the invention of claim 3, even in the disk device having the cache memory, by providing the cache memory read determination means, the same effect as that of the inventions of claims 1 to 3 can be exhibited. You can
According to the invention of claim 4, by providing the cache memory read request canceling means, unnecessary or read requests can be deleted.

According to the invention of claim 5, the pseudo random number is calculated based on the ID number, and the pseudo random number is used as the storage position of each storage unit, so that the storage position can be easily calculated. According to the invention of claim 6, the storage position calculation instructing means can determine the storage positions of all the storage units only by the pseudo random number calculation means calculating one pseudo random number, so that the storage positions can be more easily stored. Can be calculated.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a disk array device according to an embodiment of the present invention.

FIG. 2 is a flowchart of a data block read operation in the disk array device.

FIG. 3 is a flowchart of a data block read operation of the disk array device when the magnetic disk device has a cache memory.

[Explanation of symbols]

 100 Control Device 101 Encoding Means 102 Storage Position Calculating Means 103 Writing Means 104 Reading Means 105 Judgment Instructing Means 106 Decoding Means 201 Magnetic Disk Units 202 Magnetic Disk Units 203 Magnetic Disk Units 204 Magnetic Disk Units 205 Magnetic Disk Units

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G11B 20/18 570 G11B 20/18 570Z 572 572F

Claims (6)

[Claims] 1. A storage device comprising a plurality of storage devices and a control device for controlling the storage devices, wherein the control device divides the data blocks into a predetermined number of data units of 2 or more when writing the data blocks, and Coding means for generating a missing data unit generating parity block, and further performing a coding process for distributing and storing each of the predetermined number of data units and parity blocks as a storage unit in the plurality of storage devices, When reading a data block, read command issuing means for issuing a read command for reading the predetermined number of data units to the plurality of storage devices, and the predetermined number of data units read in response to the read command And a decoding means for performing a decoding process for generating one data block from the Even if one data unit out of a predetermined number of data units is lost, the data conversion unit uses the parity block and the other data unit that has been read out to generate a missing data unit, thereby generating a complete data block. In the disk array device using an encoding method capable of performing, the read command issuing unit issues a read command that also reads a parity block when reading a data block, and the control device, Further, a first determination unit that determines that a predetermined number of storage units has been read by the control device from the plurality of storage devices; and a predetermined number of storage units that have been read by the first determination unit A second judgment means for judging whether or not a parity block is included in the storage unit when the judgment is made, and the second judgment means. When it is determined that the parity block is included in the means, the data unit that has not been read is regarded as a missing data unit, and an instruction means that instructs the decoding means to generate the data unit. A disk array device comprising: 2. The storage device is a magnetic disk device, and the encoding means changes the storage position of each storage unit corresponding to one data block stored in the plurality of storage devices, for each storage device. And a storage unit for storing each storage unit corresponding to one data block in the storage position calculated by the storage position calculation unit. The disk array device according to claim 1. 3. The magnetic disk device comprises a cache memory for temporarily storing the storage unit read from the magnetic disk unit prior to reading the storage unit, and the control device further stores the storage unit before reading the storage unit. A cache memory read request issuing means for issuing a cache memory read request for temporarily storing a unit in the cache memory, and a cache memory read determining means for determining that a predetermined number of storage units have been temporarily stored in the cache memory of the magnetic disk device. And the read instruction issuing means issues a read instruction to the magnetic disk device at the time when the cache memory read determining means determines that a predetermined number of storage units are temporarily stored in the cache memory of the magnetic disk device. Claims characterized by issuing 2 disk array apparatus according. 4. The control device further stores a storage unit in the cache memory at the time when the cache memory read determination means determines that a predetermined number of storage units have been temporarily stored in the cache memory of the magnetic disk device. 4. The disk array device according to claim 3, further comprising cache memory read request canceling means for canceling an issued cache memory read request for a magnetic disk device that has not been issued. 5. The disk array device specifies a data block to be read by using an ID number corresponding to the data block on a one-to-one basis, and the storage location determining unit is configured to identify the data block with the ID number. A pseudo-random number calculating means for calculating a pseudo-random number corresponding to the storage area of the magnetic disk device and having a different value for each magnetic disk device; and a pseudo-random number for each of the plurality of magnetic disk devices based on the ID number. The pseudo random number calculating means is instructed to calculate a random number, and the calculated pseudo random number is provided as a storage position calculation instructing means for setting the storage position of each storage unit corresponding to the data block specified by the ID number. Claim 2 characterized by the above-mentioned.
5. The disk array device according to claim 4. 6. The pseudo-random number calculation means calculates one pseudo-random number corresponding to an ID number, and the storage location calculation instructing means changes the phase of the one pseudo-random number and is specified by the ID number. 6. The disk array device according to claim 5, wherein a storage position of each storage unit corresponding to a data block is set.