JP2771186B2 - Parallel data transfer method and apparatus - Google Patents

Description

Description: BACKGROUND OF THE INVENTION The present invention relates to a parallel data transfer system for restoring data using redundant data and an apparatus therefor.

[Conventional technology]

A conventional device divides data into a plurality of recording media as described in JP-A-62-24481, adds ECC data and records the data, detects a read error based on the ECC data at the time of reading, and corrects the error. I was going.

[Problems to be solved by the invention]

In the above-mentioned prior art, no consideration is given to a storage medium failure, and there is a problem that the transfer rate is deteriorated if a read error occurs during parallel data transfer.

An object of the present invention is to provide a method and an apparatus for avoiding a data transfer delay or stop when a read failure occurs.

It is another object of the present invention to provide a method and an apparatus that do not require a replacement process due to a failure of a storage medium that causes a delay in data transfer.

[Means for solving the problem]

In order to achieve the above object, the occurrence of a read failure is detected as a data delay in parallel data transfer synchronized with data transfer. In addition, when data delay is detected, data restoration is forcibly started from redundant data.

[Action]

In parallel data transfer, data delay does not occur while data transfer is performed correctly. When a read failure occurs at the transfer source of the parallel data, the data transfer starts to be delayed. By detecting the start of the delay and starting to repair the data of the read failure from the redundant data, the data transfer can be continued without increasing the delay.

〔Example〕

Hereinafter, one embodiment of the present invention will be described in detail. FIG. 1 is a block diagram showing an embodiment in which the present invention is applied to an array magnetic disk drive, and FIG. 2 is a detailed block diagram of a parity generator / data corrector 6 in FIG. FIG. 3 is a circuit diagram showing one embodiment of the correction control unit 14 in FIG. 2, and FIG. 4 is a timing chart for explaining the operation of the correction control unit in FIG.

In FIG. 1, reference numeral 2 denotes a 32-bit data bus, which is a data transfer path between the array magnetic disk device and the host system. Reference numeral 3 denotes a bus control unit which executes input / output of data to / from the bus 2, divides the data into four by the selector 5, collects the data from the selector 5, and generates a DATAREADY signal 13
Synchronize data with a. Reference numeral 5 denotes a selector, which passes data from the host to the data buses 110, 111, 112, and 113 when data is input. At the time of data output, the a input or the b input is individually selected by the error signal 13 and sent to the bus control unit 3. If the error signal 13 is true, the b input is selected,
The restoration data 12 is input. Reference numeral 6 denotes a parity generator / data corrector, which generates parity data from the data on the data buses 110, 111, 112, 113 when data is input from the host system, and outputs the parity data from the P output to the data bus 11P.
When outputting data to the host system, the data buses 110 and 11
1, 112, 113, and 11P are input, and when the data needs to be corrected, the repair data 12 is output, the error signal 13 is output, and the repair data is input to the selector 5. Reference numeral 7 denotes a data buffer, which is used to synchronize data input / output.
When outputting data, the data buffers 70, 71, 72, 73, and 7P send the data to the data bus 110, 111, 112, 113, or 11P in response to a read error-free signal from the hard disk controller 8. A hard disk controller 8 controls magnetic recording on the magnetic disk device 9. 9
Is a magnetic disk drive, and the magnetic disk drives 90, 91, 9
2,93,9P synchronizes the index signal, which is the reference of magnetic disk rotation, with the rotation synchronization signal 10, and controls the same sector to pass under the R / W head in each magnetic disk device at the same time. Have been.

In FIG. 2, reference numeral 19 denotes a parity generator which operates only when data is received from the host system, and has inputs 0, 1, 2, 3, and
Is output to the output P. 150,15
1, 152, 153, 15P are data transfer request signals, which indicate data transmission requests from the data buffer 7 of FIG. 1 for each of the data buffers 70, 71, 72, 73, 7P. Reference numeral 14 denotes a correction control unit, which determines whether or not repair is necessary based on the data transmission requests 150, 151, 152, 153, and 15P.
The signal 13a and error signals 160, 161, 162, 163, and 16P are output.
17,170,171,172,173,17P are selectors and error signals
Inputs a and b are selected based on 160, 161, 162, 163 and 16P and output to the parity generator 18. False “0” is always input to the input a. 18 is an odd parity generator, and inputs 0,
An odd parity is generated from 1, 2, 3, and P, and the repair data 12 is output.

In FIG. 3, reference numeral 21 denotes a timer, in which all signals 15a are changed from false "0" to one of true "1".
Triggered by the rising edge of the signal 23a, the counting starts. When the count UP reaches a certain set value, a TimeOUT signal 21a is output and the count UP is stopped.

Also cleared by input of REQREADY24a. Reference numeral 22 denotes an error lock device, to which the output TimeOUT signal 21a of the timer 21 is input as a trigger, triggered at the rising edge, and latches an inverted signal of the data transfer request signal 150, 151, 152, 153, 15P which is an input signal. Further, the latch is cleared by the clear signal 22a, and false “0” is output.

4, DRVO to P show the timing of the sector for each installation of the magnetic disk device 9 shown in FIG. 1, and an example of the timing of the signals in FIGS. 2 and 3 corresponding thereto. Is shown.

Next, the operation of each unit will be described with reference to the drawings. First, an input from the host system to the magnetic disk device will be described. Data from the host system has a data width of 32 bi.
The signal is input to the bus control unit 3 from the bus 2 of t. The data is divided into four data of 8 bit width by the bus control device 3, and is temporarily recorded in the data buffers 70, 71, 72, 73 via the selector 5. When data is recorded in the data buffer, the parity generator / data corrector 6 operates on the data buses 110, 111, 112, 113,
Parity data is generated from the data of
Record to. The data in the data buffer 7 is exchanged into an NRZ signal by the hard disk controller 8 and recorded on the magnetic disk device 9. Magnetic disk units 90, 91, 92, 93,
9P synchronizes the index signal with the rotation synchronization signal 10. READ and WRITE to the magnetic disk devices 90, 92, 93 and 9P are executed synchronously by the synchronization of the index signal.

Next, output from the magnetic disk device to the host system will be described. The data read from the magnetic disk device 9 is converted into 8-bit data by the hard disk controller 8 and recorded in the data buffer 7. If there is no READ error, the data in the data buffers 70, 71, 72, 73 is sent to the bus control unit 3 via the selector 5 and the DATAREADY signal 13a
The data is sent to the bus 2 after being collected into a data width of 32 bits in synchronization with the host system.

The magnetic disk device 9P for recording parity data is used to generate data of a failed magnetic disk device when a failure occurs in any of the magnetic disk devices 90, 91, 92, 93, 9P.

For example, if a failure occurs in the magnetic disk drive 93 and data is
When READ becomes impossible, the data is restored as follows.

The data is written to the magnetic disk devices 90, 91, 92, 93 and 9P. Assuming that the data is D0, D1, D2, D3, and DP, the DP is generated by equation (1) and written.

DP = NOT (D0D1D2D3) (1) DP is an odd parity of D0, D1, D2, D3.

At the time of READ, a failure occurred in the magnetic disk
When READ becomes impossible, the data of D3 can be generated by equation (2).

D3 = NOT (D0D1D2DP) (2) This parity data generation and data restoration are performed by the parity generator / data corrector 6. The repair data is output from the parity generator / data corrector 6 and input to the b input of the selector 5. RE by error signal 13
The selector corresponding to the magnetic disk having the AD failure selects the restoration data and restores the data.

 The following failures can be considered as the READ failure.

Failure 1, Drive unrecoverable failure, Failure 2, Data unrecoverable error, Failure 3, Data recoverable error, Failure 4, Delay due to defective replacement, Failure 5, Temporary rotation synchronization deviation, Failure 1 fixed data recovery By executing, it is possible to continue the READ and WRITE processing to the host system without delay.

In the failure 2, it is sufficient to perform data restoration only on the data of the unrecoverable error, but a delay occurs until the detection of the unrecoverable error, that is, until the retry is completed.

For the failures 3, 4, and 5, data recovery is not required, but delays still occur.

The fault 1 is operable without delay despite the degree of the fault being heavier than the faults 2, 3, 4, and 5.

The parallel data transfer device according to the present invention is provided with a circuit for detecting the start of the delay due to the failures 2 to 5, which is originally delayed, and starts data restoration by detecting the delay.
The characteristic feature is that the data delay by this delay detection allows the data delay to be stopped within the time required for the delay detection. The data recovery by detecting the delay will be described in detail below with reference to FIGS. 2, 3, and 4.

It is assumed that DRV0 to DRV3, P of the magnetic disk device shown in FIG. When reading the sector 1, an ECC error occurs in the DRV3, and the data corresponding to the sector in the data buffer 73 is not transmitted from the data buffer.
3, timing 99 corresponding to sector 1 of P.
Since the LRESET 22af is terminated at the timing 99a, the error lock unit 22 has been cleared and the REQ0 to REQ3,
Timer 21 is also cleared in synchronization with P. Thereafter, the REQ0 signal 150 is output first and passes through the OR circuit 23 in FIG.
Start one.

Thereafter, REQ1, REQ2, and REQP are also output before the Timer 21 outputs the TimeouT signal, but REQ3 is not output because a read error has occurred. For this reason, since the AND circuit 24 cannot satisfy the AND condition, the timer 21 continues to count up without being cleared by the REQREADY signal 24a. At timing 99b, the timer 21 detects TimeouT and outputs a TimeouT signal 21a. With this output, the error locker uses this TimeouT
The inverted signal of REQ0 ~ 3, P at the time is latched and output. The output of this latch is ORed with REQ0-3, P as shown in FIG. 3 to become signal 15a. By this OR, the signal 1 corresponding to REQ3
5a becomes true and all the REQ1-3, P signals become true. As a result, the AND condition can be taken in the AND circuit 24, and the REQREADY signal 2
Let 4a be true. As a result, the timer 21 is cleared and Ti
The meouT signal 21a becomes false. Timer 21 start signal 23a
That is, the output of the OR circuit 23 becomes true fixed by the output of the error lock device 22, and the count UP is
Not resumed.

In this state, the AND condition of the AND circuit 24 is satisfied when the outputs of the other REQ0, REQ1, REQ2, and REQP signals are completed, regardless of the signal of REQ3, and the DATAREADY signal 24a is set to true. At this time, if the error detector 20 has not detected an error, the AND circuit
25 AND conditions are satisfied, and the DATAREADY signal 13a becomes true. The DATAREADY signal 13a is input to the bus control unit 3 in FIG. 1 and serves as a data transfer synchronization signal.

The outputs ERROR0 to ERROR3, P of the error lock unit 22 are output as error signals 160,161,162,163,16P in FIG. 2, and only the selector corresponding to the magnetic disk device in which the error has occurred is operated to select the input a of the selector 17. . The output restoration data 12 of the parity generator 18 shown in FIG. 2 output under this input condition has the value of the above equation (2), and is exactly the DRV3 restoration data.

The outputs ERROR0 to ERROR3 of the error lock unit 22 are output as the error signal 13 in FIG. In the case of the error, since ERROR3 is true, the selector corresponding to the data bus 113 selects the b input and sends the repair data to the bus control unit.

As described above, when TimeouT is detected by the timer 21, the repair data is used for the error data bus and the DATAR
The EADY signal 13a is output. The bus control unit 3 has a DATAREADY
Data is transmitted to the bus 2 by a signal. When the transmission to the host system is completed, the bus control unit 3 notifies the data buffer of the completion of the transmission, and the data buffer resets the data request signal and starts the next data transfer. When all the data in sector 1 in FIG. 4 has been transferred, the clear signal in FIG.
22a is output, and the latch of the error lock device 22 is cleared. Thereby, as shown in FIG. 4, the input signals START23a and ERROR3 of the timer 21 are reset and the error lock state is released.

The error detector 20 in FIG. 3 determines whether or not the data can be restored by the parity data. The error detector 20 has inputs a,
From b, c, d, e, an output Q is output from equation (3).

Q = a ▲ ▼ + b ▲ ▼ + ▲ ▼ c ▲
▼ + ▲ ▼ d + ▲ ▼ e + ▲
▼ (3) From page 1 to page 5 on the right-hand side of equation (3), it is detected that an error has occurred in only one magnetic disk device, and all of the items in paragraph 6 detect no error. This error detector 20 can prevent erroneous restoration of data.

According to this embodiment, data recovery can be started by detecting data delay due to a data error, a replacement sector, a replacement track, an ID read error, a hard disk controller failure, a data buffer parity error, and the like. There is an effect that the read operation can be executed without causing a significant data delay.

In this embodiment, the odd parity is used. However, the parity can be realized by the even parity.

〔The invention's effect〕

As described above, according to the present invention, since data restoration is started by detecting data delay, substantial data delay does not occur.

In the case of a magnetic disk device that is rotationally synchronized, it takes 50 ms to 160 ms to retry a data error, but according to the present invention, a delay can be detected within the range of synchronous rotational deviation, so that 40 μs to 50 μs
The delay can be suppressed to s.

Note that the present invention is not limited to a magnetic disk device, and can be applied to an optical disk device, a magneto-optical disk device, a magnetic tape device, a magnetic drum device, and the like.

Also, the present invention can be applied to a communication device and a communication method for transmitting and receiving data through a plurality of paths.

[Brief description of the drawings]

FIG. 1 is a block diagram of one embodiment of the present invention. FIG. 2 is a block diagram of the parity generator / data corrector 6 of FIG. FIG. 3 is a circuit diagram of the correction control unit 14 of FIG. FIG. 4 is a timing chart for explaining the operation of the circuit diagram of FIG. 5 selector, 6 parity generator, data corrector, 7 data buffer, 8 hard disk controller, 9 magnetic disk device, 17 selector, 14 correction controller, 18 parity generator, 19 parity Generator, 21… Timer, 22… Error lock device.

Claims (8)

(57) [Claims] 1. A storage device group comprising a plurality of storage devices for storing a plurality of divided data and at least one storage device for storing redundant data generated from the plurality of data, and transferring the data to a control unit. A data transfer method for reading data from a storage device, wherein the control unit starts reading data from one of the storage devices in the storage device group, and starts reading data from another storage device in the storage device group. A read failure of the one storage device is detected when it is delayed by a predetermined time or more as compared with the start of reading of data from the device, and data to be read from the one storage device is removed from the one storage device except for the one storage device. A parallel data transfer method, wherein the data is generated from data of a recording device group and redundant data. 2. A method for storing data and redundant data in a storage device group comprising a plurality of recording devices for storing a plurality of divided data and one or more storage devices for storing redundant data generated from the plurality of data. In a parallel data transfer device for reading / writing, a control unit for receiving data from the storage device group, wherein reading of data from one of the storage devices in the storage device group is started by the storage device Means for detecting a read failure of the one storage device when a delay of a predetermined time or more has occurred as compared with the start of reading data from another storage device in the group, and data to be read from the one storage device And a means for generating the data from the recording device group excluding the one storage device and the redundant data. 3. A storage device group comprising: a plurality of recording devices for storing a plurality of divided data; and at least one storage device for storing redundant data generated from the plurality of data; and a storage device group connected to the control unit. A parallel data transfer method in which data is read out and synchronously transferred, wherein the correction control unit starts reading data from one of the storage devices in the storage device group, and starts reading data from another of the storage device groups. When a delay of a predetermined time or more compared with the start of reading data from the storage device is detected, the synchronization shift of the one storage device is detected, and data to be read from the one storage device is stored in the one storage device. A parallel data transfer method, wherein the data is generated from data and redundant data of the recording device group excluding the data. 4. A method for storing data and redundant data in a storage device group comprising a plurality of recording devices for storing a plurality of divided data and one or more storage devices for storing redundant data generated from the plurality of data. In a parallel data transfer device that reads / writes synchronously, a control unit that receives data from the storage device group, wherein reading of data from one of the storage devices in the storage device group is started by: Means for detecting a read failure of the one storage device when a delay of a predetermined time or more has occurred as compared with the start of reading data from another storage device in the storage device group;
Means for generating data to be read from two storage devices from data of the recording device group excluding the one storage device and redundant data. 5. A method for temporarily storing data read from a storage device group including a plurality of recording devices for storing a plurality of divided data and at least one storage device for storing redundant data generated from the plurality of data. Stored in the data buffer,
A data transfer request signal corresponding to each of the storage devices is sent from the data buffer to a data correction unit, and the data correction unit transmits the data transfer request signal of one of the storage devices in the storage device group to the storage device. When the data transfer request is delayed by a predetermined time or more compared with the data request signal of another storage device in the group, a read failure of the recording device of the data transfer request is detected, and data to be read from the one storage device is detected. A parallel data transfer method, wherein the data is generated from data and redundant data of the recording device group excluding one storage device. 6. A storage device group including a plurality of storage devices for storing a plurality of divided data, and one or more storage devices for storing redundant data generated from the plurality of data, and read from the storage devices. A data buffer for temporarily storing data, and a parity generation / data correction unit for receiving a data transfer request signal and data corresponding to each of the storage devices when the data buffer receives data read from the storage device Wherein the parity generation / data correction unit comprises:
When the data transfer request signal of one of the storage devices in the storage device group is delayed by a predetermined time or more compared with the data request signal of another storage device in the storage device group, the data transfer request A correction control unit that detects a read failure of the recording device; and a parity generation unit that generates data to be read from the one storage device from data of the recording device group excluding the one storage device and redundant data. A parallel data transfer device. 7. The parallel data transfer method according to claim 1, wherein the plurality of storage media are rotationally synchronized by an index signal. 8. The parallel data transfer device according to claim 2, wherein the plurality of storage media are rotationally synchronized by an index signal.