JP2003316525A - Disk array controller - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a disk array configured of three storage units whose failure resistance and read/write performance is improved, and to provide a disk array having a switching circuit exclusive for backup capable of quickly executing backup or restore processing, and reducing a load on a system. <P>SOLUTION: The disk array controller 50 is provided with a disk array control circuit 51 configuring the disk array of mirroring for three storage units HDD 53, HDD 54, and HDD 55 and a switching circuit 52 for separating or re-connecting one arbitrary storage unit among those tree storage units from or to the disk array. Thus, it is possible to realize the direct connection to a backup device, and to realize the backup or restore processing in an ideal environment. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for performing information processing using a hard disk device or the like for processing a large amount of data, and a disk array called RAID for improving reliability and performance of the hard disk device or the like. An information processing device such as a server device, an OA device, and a measuring device, which includes a storage device having a configuration. At the same time, it relates to an information processing apparatus having a backup device in order to further improve reliability.

[0002]

2. Description of the Related Art In order to record and hold a large amount of data,
In information processing equipment, a small-capacity recording device using a semiconductor such as a non-volatile memory to a large-capacity recording device using a magnetic tape has been used. Although recording with a magnetic tape has a large capacity, the speed of reading and writing is remarkably slow because of the device characteristics that only sequential access is possible. Therefore, it is mostly used as a storage medium dedicated to backup. On the other hand, a disk device, such as a hard disk device, which records and reproduces data by bringing a head, which changes the magnetism of the medium using electromagnetic induction, close to or in contact with a medium coated with a magnetic material called a disk, has a capacity and a price. It occupies the mainstream due to the balance of usability such as reading and writing speed, and is expected to continue to be used as a recording device for information equipment in the future.

In such a recording apparatus, a rotating mechanism is used for rotating the medium, and the probability of failure and the life of the apparatus are remarkably worse than those of a recording apparatus having no rotating mechanism such as a semiconductor. In addition, the head must reach a predetermined position on the medium in order to read and write the target data,
Since the moving time of the mechanism is required, the access speed will be inferior.

In order to solve the above problems, R
A disk array technology called AID has been invented and widely used. FIG. 1 shows a diagram of a standard system configuration using a disk array device. The CPU 1 uses the system controller 2 to access the main memory 4 and the high-speed display circuit 3 such as AGP. As a storage device,
The disk array control device 6 uses the n hard disks 8
To 10 are used to construct a disk array. In order to perform backup, a backup device 11 such as a tape drive is connected to a disk controller 7 such as a SCSI controller and executes backup processing.

In the above configuration, the disk array control device 6 uses a plurality of hard disk devices and the like,
The storage device itself is made redundant. Thereby, at the same time as dealing with a failure such as a failure, reading and writing of data are simultaneously performed in parallel to a plurality of storage devices to improve the performance. Actually, a cache memory is also installed to improve the apparent access speed, and the order of accessing each disk device is determined by the arrangement of data, the position of the disk head, the track configuration of the recording medium, etc. There are also devising algorithms such as optimization.

Although various methods and variations have been devised for disk array technology, RAID is generally used.
0, RAID1, RAID2, RAID3, RAID
4, a system called RAID 5 is a technique generally called a disk array. However, what is actually used is a method called RAID 0 or disk striping explained in FIG. 2, and R explained in FIG.
A method called AID1 or mirroring, and a method called RAID5 or distributed parity described in FIG. Among them, the RAID 1 shown in FIG. 3 is most popular nowadays because the capacity and price of hard disk devices are increasing in recent years. RAID2,
RAID 3 and RAID 4 are technically only capable of constructing a disk array, and when used practically, they are used because of inferior performance or cost increase compared to other methods. There is almost nothing.

Hereinafter, RAID 0,
The configurations and methods of the RAID 1 and RAID 5 disk array technologies will be described.

FIG. 2 shows RAI in the disk array technology.
This is a system called D0 or disk striping. In this method, the storage device has no redundancy and thus has no fault tolerance function. Since the data is distributed and read / written to / from each storage device, the processing performance is improved. When there are four storage devices as shown in FIG. 2, [A], [B],
When handling the data [C] and [D], the HDD 20
To [A], to HDD21 [B], to HDD22 [C],
It can be distributed and processed in parallel at the same time as [D] in the HDD 23. Another feature is that a plurality of storage devices are treated as one logical storage device. However, it may be said that the storage device is not RAID because it does not have redundancy.

FIG. 3 shows RAI in the disk array technology.
This is a method called D1 or mirroring. In this method, two storage devices HDD30 and HDD31 are provided, and the same data is simultaneously written at the time of writing. As a result, since one storage device is redundant, data preservation is realized in the other storage device even when a failure occurs. For example, when handling the data [A] and [B],
HDD30 and HDD31 are exactly the same [A],
Data [B] is recorded. Even if a failure occurs in either one of the recording apparatus therefor, it is possible to maintain the data by the other recording device. However, data integrity cannot be maintained if two recording devices fail simultaneously. Further, although it has two storage devices, since the contents are the same, the total capacity of the disk array remains the capacity of one storage device.

At the time of reading, it is attempted to improve the performance by devising a reading algorithm.
It cannot be said that the performance improvement rate is so large even compared to a normal storage device.

FIG. 4 shows RAI in the disk array technology.
This is a method called D5 or distributed parity. Three
It is a system that is composed of one or more storage devices and distributes read / write data parity to the storage devices. The case of FIG. 4 shows a case where the RAID 5 is configured by four storage devices. Even if one storage device fails due to parity, the data in the failed storage device
The data in the remaining storage device can be used to generate and continue the processing. In the example of FIG. 4, [A], [B],
When handling the data [C], [A] to the HDD 40, [B] to the HDD 41, [C] to the HDD 42, H
Each storage device is accessed as parity data to the DD 43. The storage devices that read and write the parity data are controlled to be distributed, and in the example of FIG. 4, the parity data will be recorded in the HDD 40 in the next access. It is the disk array controller 6 that generates the parity data and inspects the data for abnormalities with the parity data. Further, a method has been devised in which the generation and inspection of the parity data are replaced by software of the system, but it is rarely used because the load on the CPU of the system is large. It is RAID 4 that the parity data is dedicatedly stored in a specific storage device and is read and written. However, since reading for parity data generation and checking is concentrated in the specific recording device, a load is applied and processing is performed. Becomes a speed bottleneck.
In RAID 5, the parity data is also distributed to the storage devices, so that the load of generating and checking the parity data is also distributed, and the processing speed is improved.

Further, unlike RAID 1, three or more storage devices can be handled, and the storage device capacity other than the parity data can be handled for data processing.
It is also possible to increase the total capacity of the storage device for the entire AID5.

However, in recent years, storage devices such as hard disks have become larger in capacity, lower in price and higher in performance, and a method called RAID 1 or mirroring, which has a simple structure and has a fault-tolerant function, is used. It is getting more and more.

The above is a description of a disk array called RAID. The high speed of RAID 0 and the RA
A configuration using four storage devices, RAID0 + 1, which also have the fault tolerance of ID1, may be devised and used.

[0015]

No matter what method of the disk array technology is used, the danger or failure of the storage device is not completely eliminated, but the redundancy of the storage device causes a stopped state or a dangerous state. It only takes defensive measures to reduce or deal with the odds of reaching. Therefore, an information device incorporating a storage device using the disk array technology must always use some kind of backup device such as a tape recording device.

However, if a backup is created in a tape device or the like during processing for the disk array,
Since there is a time difference between the start and the end, the data in the storage device cannot be consistent. Therefore, offline processing is performed by stopping the normal processing of the system for backup. Further, in that case, the system must be stopped for backup, which is inconvenient. Therefore, various restrictions and conditions are set during operation, and then the entire or part of the recording apparatus is backed up. Further, in this case, data processing during normal system operation and read-out for backup are performed in parallel, which imposes a load on each processing. When the disk array technology is introduced for high-speed processing at the same time as fault tolerance, the slow processing due to backup significantly impairs the advantages of the disk array technology.

[0017]

As described above, the method called RAID1 or mirroring is currently the mainstream among the disk array technologies. This is because the same data is simultaneously written in two recording devices such as hard disks, so that even if one recording device fails, the other recording device retains the data. At the time of reading, the reading order from the two recording devices is analyzed so that the reading time is the shortest, and the reading is executed to realize high-speed reading. Paying attention to the advantage of this method, it is possible to further add a recording device such as a hard disk.
In addition, the present invention invented a disk array control device characterized by using three recording devices. FIG. 5 shows a diagram of the configuration of the disk array control device of the present invention. In the disk array control device 50 of the present invention, the same data is simultaneously written to the three recording devices HDD53, HDD54, and HDD55 at the time of writing. At the time of reading, the read data is analyzed in advance, and it has a feature of having a disk array control circuit 51 having an algorithm capable of selecting a read target storage device and giving a read command so that the read data can be read in the shortest time. .

Further, it has a switching circuit 52 for disconnecting or reconnecting any one recording device of the three storage devices to the structure of three disk array devices and two disk array devices. RAID 1 according to the above, and the configuration of one single recording device, 2
It is characterized by being able to dynamically take the street structure.

The disk array control device of the present invention is a disk array control device for improving reliability and performance of a hard disk device. The disk array control device is a RAID 1 or mirroring first and second storage device and a third storage device. It has a storage device and has a reading method that writes the same data to three storage devices at the same time when writing data and maximizes the reading speed when reading data. And a mechanism for connecting the storage device separated from the disk array by the switching circuit to the backup device.

Further, in a disk array control device which comprises three storage devices and constitutes a disk array for recording the same data, when a failure occurs in the first storage device, a second and a third storage device are provided. The storage device constitutes a RAID 1 disk array. Therefore, when a failure also occurs in the second recording device, the failure resistance is improved as compared with RAID 1 in which data is preserved by the third storage device. Further, as compared with reading from two storage devices as in RAID 1, reading from three storage devices results in improved read speed.

Further, when performing backup, it has a switching circuit for disconnecting any one recording device from the disk array and connecting it to the backup device. Therefore, the storage device can be handled exclusively for backup, and data does not change during backup, so that an ideal environment for implementing backup can be realized.

Even during the backup process, the remaining two storage devices form a RAID 1 disk array. Therefore, the configuration of the disk array can be dynamically changed so that the fault tolerance can be maintained even during the backup process.

Due to the switching circuit of the disk array controller, there is no conflict between the normal data processing access of the system to the disk array and the read access to the disk array by the backup processing. Therefore, it is possible to configure a disk array in which neither access to the storage device conflicts nor load is applied to both the data processing and the backup processing.

The recording device is separated from the disk array by the switching circuit of the disk array control device,
When connected to the backup device, since it is locally connected, the backup data can be directly transferred from the recording device to the backup device without transferring the backup data via the system, and the load on the system is increased. Also, the time required for backup processing,
It can be reduced significantly. The restore process is also similar to the backup process.

The recording device is separated from the disk array by the switching circuit of the disk array control device,
After the backup is completed, the recording device connected to the backup device is switched to the disk array again to rebuild the disk array, whereby the disk array with the three storage devices can be dynamically rebuilt. The restore process is also similar to the backup process.

[0026]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. In this description, the storage device constituting the disk array is described as a hard disk, and is referred to as HDD in the figure. This is because a hard disk is almost always used as a storage device of the disk array. Of course, for the effect of the present invention,
It goes without saying that it does not matter whether the storage device is a hard disk or a storage device using another storage medium.

FIG. 5 is a diagram showing the configuration of the disk array control device of the present invention. In the normal state, the disk array configuration is composed of three HDDs, and the switching circuit 52 at that time connects the disk array control circuit 51 and the HDD 55. When writing data, the disk array control device 50 receives the write data transferred from the system bus 5, and the disk array control circuit 51 causes the three recording devices HDD53, HDD54, and HDD5.
Write the same data to 5 at the same time. Since the same data is simultaneously written in the three recording devices HDD53, HDD54, and HDD55 in this way, it can be said that this is an extended form of the RAID1 method. At the time of reading, the amount of data of the read request is analyzed so that the disk array control circuit 51 can read from the three HDDs at the highest speed, and the read data is read while changing the order of the read target HDDs. Is output to the system bus 5 by the disk array control circuit 51.

The disk array control device 50 of the above-mentioned form
In the case of taking a backup with, the switching circuit 52 disconnects one HDD from the three HDDs.
Although the HDD 55 is described as an object in FIG. 5, there is no particular limitation as to which HDD the HDD should be. At this time, the switching circuit 52 disconnects the HDD 55 from the disk array and connects it to the disk controller 7 to which the tape drive 11 which is a backup device is connected. The disk controller 7 is, for example, SCSI.
It's like a controller. With such a connection, the tape drive 11 as a backup device can take a backup from the HDD 55. At that time, the disk array control device 50 uses the normal RAID 1 with the two recording devices HDD 53 and HDD 54.
Thus, the configuration is such that one backup dedicated storage device HDD55 is adopted. The write and read operations by the disk array control device 50 and the HDD failure and backup operations will be individually described below.

FIG. 6 is a schematic diagram of the switching circuit 52 included in the disk array controller 50. In FIG. 6, the case where the disk array control device 50 operates as a disk array is shown by extracting only the HDD 55. Of course, actually, the HDD 53 and the HDD 54 are also connected to the disk array control circuit 51. In the case of FIG. 6, the control signal and data bus of the HDD 55 are connected to the disk array control circuit 51 inside the switching circuit 52. That is, a disk array is composed of three HDDs. This configuration is the configuration of the disk array control device 50 of the present invention in a normal state.

It should be noted that the actual circuit structure inside the switching circuit 52 is a circuit for switching the bus, and when switching the bus, it is a circuit that is connected to another bus after being in a high impedance state. Needless to say. It may be composed of logic components or analog elements.

For writing from the system to the disk array formed by the disk array control device 50, the disk array control circuit 51 is an HDD having three recording devices.
53, HDD 54, and HDD 55 at the same time. The write data is also the same. That is, exactly the same write command and data are sent to the HDD 53, the HDD 54, and the HDD 55.
Send to. As a result, the data written to each HDD is
It has the same content. Therefore, it is the same as having three HDDs. This is why it is called mirroring.

When reading from the system to the disk array constituted by the disk array control device 50, the disk array control circuit 51 judges from the content and amount of the read-requested data so that the read can be executed fastest. A read command is issued to each HDD. It can be simply said that reading from three HDDs simultaneously at the same time is three times faster than reading from one HDD, but in reality, the position of the HDD medium where the data to be read is located, In this case, the read speed of the head in the storage device depends on which head is used. Therefore, the disk array control circuit 51 calculates the time required for the head of each HDD to move to the position where the target read data is recorded, and then the total read time is shortened to each HDD. Execute a read instruction.
As described above, at the time of reading, the commands issued to the respective HDDs are not the same, and the amount of data to be read usually differs depending on the HDD.

When a failure occurs in any of the HDDs, the disk array configuration is changed when the disk array control circuit 51 recognizes the failure. Recognizing a failure means, for example, an abnormal response to a command,
This is the case when an error occurrence report is received from the HDD during reading or writing. Depending on the HDD, S
Some have a failure notification function called a MART function, but this includes a case where such a failure notification is received. In the example of FIG. 5, if it is recognized that the HDD 53 has a failure, then the disk array control circuit 51 is set to H level.
The access to the DD 53 is stopped and the HDD 54 and the HDD 5 are stopped.
Access only 5. This state is the same as the RAID 1 disk array configuration of the HDD 54 and the HDD 55. Although the system operation can be continued in this state, the HDD 53 is replaced with a new HDD at some point in order to recover the failure of the HDD 53. HDD5
Disk array control device 50
When notified by the management program for disk array, the disk array control circuit 51 reconstructs the disk array. As a result of the rebuilding process, the contents of HDD 53 are
When it becomes the same as the above, the disk array configuration of the mirroring by three HDDs before the failure is restored.

If a failure in one HDD is recognized and the failure in another HDD is recognized at the time when the RAID 1 configuration is adopted as described above, the HDD 53 and the HDD 54 will be described with reference to the example of FIG. However, the disk array control circuit 51 changes the configuration of the disk array to the processing by the remaining HDD 55 alone. However, since it is out of the purpose of maintaining fault tolerance, processing such as issuing a warning by the management program for the disk array control device 50, sounding a buzzer, or automatically stopping the system once is performed. Should be done. This is because there is a high possibility that the data recorded in the HDD will be lost if the remaining HDD 55 also fails.
If the disk array control circuit 51 has a faulty HDD
When it is notified that the disk array has been replaced, the disk array is reconstructed as described above and the disk array configuration of the mirroring by the three HDDs before the failure occurs is restored.

FIG. 7 is a schematic diagram of the switching circuit 52 included in the disk array control device 50 when the disk array control device 50 executes the backup process.
The state of FIG. 7 is when the system is trying to back up the data recorded in the disk array. When the disk array control device 50 receives a backup execution command from the system, it changes the bus connection of the switching circuit 52. In the schematic view of FIG. 7, the HDD 5
The control signal and data bus 5 are disconnected from the disk array control circuit 51 inside the switching circuit 52 and are connected to the bus of the disk controller 7. The disk array has a RAID 1 configuration including the HDD 53 and the HDD 54. On the other hand, the disk controller 7 uses the local bus and the HDD 55 to be backed up,
Since the tape drive 11, which is a backup device, is connected, the disk controller 7 can execute backup processing without system intervention. For example, assuming that the disk controller 7 is a SCSI controller, the HDD 55 and the tape drive 11
Are both SCSI units, and if the copy command defined in the SCSI standard is issued, the HDD 5
Backup process from 5 to tape drive 11 SCS
It can be executed locally on the I bus only. At this time, RAID
Access to the disk array having one configuration and backup processing are performed using completely independent buses, and therefore do not affect each other.

When the disk array controller 50 receives a notification from the system that the backup processing has been completed, the bus connection of the switching circuit 52 is changed again. In the schematic diagram of FIG. 7, the control signal and data bus of the HDD 55 are disconnected from the bus of the disk control controller 7 inside the switching circuit 52 and connected to the disk array control circuit 51. After that, the disk array is reconstructed, and the disk array configuration of the mirroring by the three HDDs before the backup processing is executed is restored.

Although the backup processing has been described so far, even in the case of restoration from the tape drive device, the exchange of data between the HDD and the tape drive is read and write; Is almost the same. The difference is the reconstruction of the disk array performed after the processing is completed. In the case of restoration, the HDD 55 in which the restored data is written
Based on the data of the HDD53 and the HDD54,
Rewrite the same as 5.

During backup and restore processing, the disk array is RAI as described above.
Although the processing is performed as D1, even if a failure occurs in any of the HDDs at this time, the disk array control circuit 51 accesses the failed HDD as in the case of the disk array described above. Take action to discontinue.

The embodiment of the present invention has been described above, but the case where a failure occurs in the storage device has been described, and the case where the storage device in which the failure has occurred is replaced at a certain point and restored is explained. . In practice, in order to perform this failure recovery quickly and easily, it is also possible to connect a spare storage device to the bus forming the disk array in advance. It is called hot standby. Even in that case, it can be handled in exactly the same way from the embodiment of the present invention.

[0040]

As compared with RAID 1, the number of recording devices is changed from two to three, so that the fault tolerance of the recording device is improved. In other words, even if there is a failure in one recording device, it is still possible to operate in the RAID1 configuration with two recording devices.
Even if one recording device fails, the data recorded in the recording device is saved by the last one recording device. Even if up to two of the three recording devices become faulty, the recorded data can be guaranteed, so normal R
Fault tolerance is improved compared to AID1.

For RAID 1, from 2 to 3 recording devices
As a result, the read speed is improved as a disk array recording device. Compared to the case of alternately reading from two recording devices at the time of reading, when the read data of the unit capacity is divided into three into three recording devices and the reading is executed at the same time, one recording device takes three minutes. Therefore, it can be said that the performance improvement of up to 3 times can be achieved if only the reading is focused on. However, in actuality, data analysis, distribution / reconstruction processing, etc. are involved, so that the processing speed cannot be increased up to 3 times.

When the backup is performed, since it is performed for the recording device separated from the disk array device, the contents of the target recording device are not updated from the beginning to the end of the backup, and the consistency can be secured. Therefore, there are no additional conditions for backup processing,
You can perform an ideal full backup.

Further, during backup, another two recording devices constitute RAID1. Therefore, the fault tolerance during backup processing is not impaired.

Further, at the time of backup, since the data for backup is read from the recording device set for backup only, the access for the normal access of the system performed for the remaining two recording devices conflicts. The load is gone. Therefore, the processing speed for the recording device is not significantly reduced.

In the above case, backup can be realized by a local bus connection between the target recording device and the backup device. For example, in case of SCSI connection, data transfer by a copy command is applicable. Therefore, the backup process itself does not impose a load on the system side, and the time required for the backup process can be minimized.

When the backup is completed, the disk array of three recording devices is constructed again. As a procedure, the disk array is rebuilt and the contents of the three storage devices are made the same. When the rebuilding of the disk array is completed, the above-mentioned processing efficiency improvements can be obtained again.

Regarding these backup processes, the method of rebuilding the disk array after the backup process is different, and the same applies to the restore process.

[Brief description of drawings]

FIG. 1 is a diagram of a standard system configuration using a disk array device. Only the main components such as CPU and main memory are shown. The tape device used for backup is also described. In this example, there are n HDDs.

FIG. 2 RAID of disk array technology which is a conventional technology
It is a figure of 0. In this example, there are four HDDs.

FIG. 3 RAID of disk array technology, which is a conventional technology
It is a figure of 1. Since it is RAID 1, there are two HDDs.

FIG. 4 RAID of disk array technology, which is a conventional technology
5 is a view of FIG. In this example, there are four HDDs.

FIG. 5 is a diagram of a configuration of a disk array control device of the present invention. The tape device used for backup is also described.

FIG. 6 is a schematic diagram of a switching circuit of the present invention (during disk array operation). Only the switching circuit and the disk controller are shown.

FIG. 7 is a schematic diagram of a switching circuit of the present invention (during a backup operation). Only the switching circuit and the disk controller are shown.

[Explanation of symbols]

1 CPU 2 System Controller 3 High Speed Display Circuit 4 such as AGP 4 Main Memory 5 System Bus such as PCI 6 Disk Array Controller 7 Controller for Controlling Disk such as SCSI Controller 8 0th HDD 9 1st HDD 10 n Second HDD 11 Backup device such as tape drive 20 RAID0 0th HDD 21 RAID0 1st HDD 22 RAID0 2nd HDD 23 RAID0 3rd HDD 30 RAID1 0th HDD 31 RAID1 1st HDD 40 RAID5 0th HDD 41 RAID5 1st HDD 42 RAID5 2nd HDD 43 RAID5 3rd HDD 50 Disk array control device 51 Disk array control circuit 52 of the present invention Replacement circuit 53 0th HDD of the disk array of the present invention 54 1st HDD of the disk array of the present invention 55 2nd HDD of the disk array of the present invention 60 Data flow in the switching circuit during disk array operation 70 Backup Data flow in the switching circuit during operation

Claims (5)

[Claims] 1. A disk array control device for improving reliability and performance of a hard disk device, comprising first and second storage devices of RAID 1 or mirroring, and a third storage device, and at the time of writing data. The same data is written to three storage devices at the same time, and a read method that maximizes the read speed when reading data is provided, and when backup is performed, the storage device to be backed up is switched off from the disk array. A disk array control device comprising: a circuit; and a mechanism for connecting a storage device separated from the disk array by the switching circuit to a backup device. 2. A disk array control device comprising three storage devices and constituting a disk array for recording the same data, wherein when a failure occurs in the first storage device,
A disk characterized in that a RAID 1 disk array is configured by the second and third storage devices, and if a failure occurs in the second recording device, data protection is achieved by the third storage device. Array controller. 3. The disk array control device according to claim 2, wherein three storage devices are used at the time of reading data, and the read speed is controlled to be maximum. 4. The disk array control device according to claim 2, further comprising a switching circuit for disconnecting any one recording device from the disk array and connecting it to the backup device when performing backup. 5. A disk array control device that constitutes a RAID 1 disk array by the remaining two storage devices even during the backup processing according to claim 4.