JPH03105541A - Buffer control method for disk devices, etc. - Google Patents

Abstract

PURPOSE:To improve the throughput when the data are read out of a magnetic disk by delaying the start of disconnection of an SCSI (small computer system interface) bus by a fixed time or longer if the remaining quantity of transfer data is smaller than a prescribed data quantity in a continuous reading action mode. CONSTITUTION:The disconnection of an SCSI bus 30 is delayed by a fixed time T3 or longer even if a buffer 111 becomes empty while the data are transferred to a host device 20 as long as the remaining quantity of transfer data is smaller than a prescribed data quantity in a continuous reading action mode. As a result, the remaining data can be continuously transferred with no disconnection of the bus 30. Thus the executing time of a read command is shortened. Then the execution timing of disconnection is controlled with a small remaining quantity of data so as to prevent such a case where the bus 30 is inadvertently occupied and the system performance is deteriorated when the remaining quantity of data is large. Thus the throughput is improved when the data are read out of a magnetic disk.

Description

[Detailed Description of the Invention] [Summary] A buffer for a disk system that has SCSI as an interface and reads/writes while controlling bus disconnect/connect according to the buffer full ratio of the mode select parameter specified in the SCSI specifications. Regarding the control method, when continuous read operation is in progress and the amount of remaining transfer data is less than the predetermined amount of data, the throughput when reading data from the disk is improved by continuing the transfer without disconnecting the SCSI bus. In the buffer control method of a disk system that has SCSI as an interface and reads/writes to the disk section via the buffer according to the SCSI specifications when a read/write command is received from the host device, The timing to disconnect the SCSI bus is detected when the amount of data falls below the specified amount of data, and when the timing to disconnect the SCSI bus is detected, if a continuous read operation is in progress and the amount of data remaining to be transferred is is less than a predetermined amount of data, the SCSI bus disconnection start timing is delayed by a certain amount of time,
The remaining amount of data read within the certain period of time is configured to be transferred to the host device without going through a disconnect.

[Industrial application field]

The present invention has a SCSI interface, and when a read/write command is received from a host device, the SCSI
Disconnects the SCSI bus when reading data from the disk unit where data is being read/written in a disk unit or disk control unit that reads/writes data to/from a disk unit such as a magnetic disk unit or optical disk via a buffer according to the specifications of This invention relates to a Huffer control method for controlling timing.

[Conventional technology]

In recent years, small magnetic disk drives have been using the conventional interface ESDT (Enhanced Sm).
all Device Interface) and ST5
01, the highly versatile S C S I (S
mall computer system
rface).

SCSI is used as a peripheral device interface for small computers, not only for magnetic disk devices but also for peripheral devices such as magnetic tapes, optical disks, printers, etc. as shown in Figure 4.
Can be connected with a real cable.

FIG. 4 shows a system configuration using SCSI. In the figure, 10. ~10, are peripheral devices such as disk devices and printers, and 20. and 20. is a host device of the peripheral devices 10o to 10, and 30 is an SCSR bus that connects each host device and peripheral device.

Each peripheral device 10o to 10s is an I/O controller 11
．． ~11, to the host device 20 via the SCSI bus 30. or 20, to transfer data. 10o in the peripheral device is a disk device such as a magnetic disk device or an optical disk device, and is an I/O device equipped with a bumper 111.
Controller (hereinafter referred to as disk control unit) 11. The host device 20 connects one of the plurality of disk units 12o to 12m to the SCSI bus 30. or 20.

In the host device 20o, 21 is a processor, 22 is a memory, 23 is a DMA controller that controls DMA transfer, 24 is a host bus, and 25 is a host bus 24 and an SCS.
This is a host adapter that controls connection with the I bus 30 and data transfer. This structure also applies to the host device 20.

A system having SCSI as an interface as described above has the following characteristics.

■ Reducing the load on the host bus In other words, the conventional system did not have an SCSI bus 30,
■/○ controller for each peripheral device 11. 115 were configured to be directly connected to the host bus 24 of one host device (20.), which placed a heavy burden on the host bus 24. On the other hand, the SCS
In systems with I, the sc
An si bus 30 is provided to connect each peripheral device 20G and 20. and control each peripheral device from its I/O controller 11. 115, the load on the host bus 24 can be reduced.

■ Many devices can be connected on the bus. Up to 8 devices can be connected on the SCS R bus 24, including each I/O controller and their host devices (in the figure, 2 host devices and 6 peripheral devices). Connectable. Therefore, a much greater variety and quantity of peripheral devices can be connected on the bus than in conventional systems.

@SCSI allows communication between I/O controllers
In addition to being able to communicate between a host device and each I/O controller, the I/O controllers can also communicate with each other without going through a host device.

■ Multitask processing is possible The I/O controller that is executing a command disconnects the SCSI bus 30 between processes, and other I/O controllers can use the SCSI bus 30 to perform processing during that time. .

Next, in systems with SCSI as an interface,
Communication between the initiator and the target is performed through the following eight phases (1) to (2). Initial shake is SCS I bus 3
This is a device that has acquired the right to use 0, and is usually a higher-level device, but in SCSI, an I/O controller can also be an initiator. The target is the device with which the inishke communicates.

■ Bus free ph
) The SCSI bus 30 is not being used by any device.

■ Arbitration phase
ionphase) In response to requests from a plurality of devices to acquire the right to use the SCSI bus 30 in the bus free state, one device (for example, the device with the highest priority) acquires the right to use the bus.

■ Selection phase
(hase) The device that has acquired the right to use the SCSI bus by executing the arbitration process, that is, the initiator, explains the message act phase and selects a specific target.

■ Command phase
) The selected target receives commands from the inishek.

■ Data phase (data phase) Data is transferred in accordance with commands from the inishake.

■ Status phase
) The target that has completed the execution of the command sends a status to the initiator indicating the execution result, such as whether the execution ended normally or abnormally.

■ message in phase
(in phase) Following the status phase, the target notifies the inishek of a message (command complete message) indicating that it has notified a valid status byte.

Also, after the command phase, this phase
onnect message to the inishek and the SC
Notifies that the SI bus will be disconnected.

■Reselection Phase
n phase) The reselection phase is a characteristic phase of scs r. In other words, when the command received from the initiator does not require the result to be immediately returned to the initiator, the target can temporarily disconnect from the initiator, and during this time send the scsr path to another warehouse. 30 will be released.

The reselection phase consists of re-execution of commands interrupted by disconnection and reconnection of the bus (
This is the phase for reconnecting.

Before entering this reseleccigon phase, the aforementioned arbitration phase is executed. SCSI enables multitasking functionality through the disconnect/reconnect function and arbitration function.

As a specification for commands used in SCSI, ANSI
CCS (Common C) proposed by (American National Standards Institute)
command Set) is widely and commonly used.

In the CCS, in order to specify the occupancy rate of the SCS I hash to the minimum necessary for the disk device, the SCS Buffer control parameters are defined to control the usage rate of the I bus.

This parameter is the buffer full ratio (buffe
r full ratio) or buffer empty
mode select command (m
ode selectcomman).

This parameter determines, for example, the disk unit 12. When reading data from and transferring it to the computer main body of the host device (20o), a certain amount of read data has accumulated in the buffer 111, and then the data is transferred on the SCSI bus 30 to the host device 20. DM
A transfer allows data to be transferred at high speed. Thereby, data can be transferred on the SCSI bus 30 in a shorter time than the time required for the disk device 100 to read data.

Further, when writing data transferred from a host device to the disk unit, when the buffer 111 becomes full, the data is immediately transferred to the SCS R bus 30 and the data control unit 10. The SCSI bus 30 is then disconnected to allow other peripheral devices to use the SCSI bus 30. In this way, the disk device 10. The data in the full buffer 111 is transferred to the disk unit 12 to some extent. The disk unit 12 remains in the buffer until it is written to or becomes empty. The next reconnection (reconnection) timing is delayed until the SCSI bus 30 is read from the SCSI bus 30, and other peripheral devices can use the SCSI bus 30 during that time. By doing so, multitasking is enabled as described above, and the SCSI bus can be used effectively.

However, these parameters have meaning only when the transfer speed on the SCSI path is faster than the disk read/write speed.

Next, with reference to FIGS. 5 and 6, if the interface has SCSI and Buffer II is configured according to the SCSI specifications,
Disk device 10. A buffer control method for a disk device or a disk control unit that reads/writes data will be explained.

When reading data from the disk section 12o on which data has been written, a specified number of blocks (recording units) on the disk section l2 are successively read and stored in the buffer 111.

This operation is performed continuously, but generally, the data in one block must be checked for errors using an error detection code attached to the end of the block, and the validity of the data has been confirmed. Data is not output onto the SCSI bus 30 until the reading of the block is completed.

Therefore, when considering the amount of data to be output to the SCSI bus 30 as the amount of data in the buffer 111, it is assumed that after the read of one block is completed, the amount of data in the buffer 111 increases by the amount of data in that block. easier to understand. Therefore, Figures 5 and 6, which will be explained below,
In the figure, after reading of one block is completed, the amount of data in the buffer ll increases by the amount of data of that block.

In addition, as mentioned above, reconnect performs a certain proportion (buffer full) of the number of blocks specified in the read command.
The reconnection (selection phase) starts when data is accumulated in the buffer 111 (ratio), but in the case of FIGS. Assume that the parameters of the mode select command are specified as follows.

First, the buffer control method in an ideal case will be explained with reference to the operation timing chart of FIG. In order to simplify the explanation, processing related to each phase before the selection phase xRS, such as the arbitration phase in which the data control unit 10o acquires the right to use the SCSI bus 30, is omitted in the following explanation.

■ When data reading starts from disk 12o (
At time t0), each time data read for one block is completed, it is considered that storage of one block of read data in the notuffer 111 is completed (see FIG. 5), (C), time t. % l j2 etc.).

t ■ When two blocks of read data with a specified buffer full ratio are stored in the buffer 111, the disk control unit 1l executes a reselection phase RS and selects a predetermined initiator (upper device 20. ) (RS in FIG. 5(a)).

■ When the reselection phase RS ends, the disk control unit 1l sends the message in phase MSG.
The host device 20 executes Ia. The disk unit 12 exchanges messages with the disk unit 12 and reads data from the host device 20o. (MSG Ia in FIG. 6(a)).

In addition, the reselection phase RS and message
Even during the execution of in-phase MSGla, the disk unit 1
2. The read data from is stored in buffer 1 in blocks.
11 (FIG. 5(b)).

■ When the execution of the message-in-phase MSGI a is completed, the disk control unit 11 executes the data phase, and while reading data from the disk unit 12o, transfers the data stored in the buffer 111 to the host device in blocks. 20. Transfer to.

Currently, the host device is 20. The data transfer speed to disk 1 is
2. Assuming that the data read speed is faster than the data read speed from t, the data in the buffer 111 decreases until time t
The data for the predetermined 8 blocks is stored in the disk section 12. When the time t is led, the time t is slightly delayed from the time t3.
4, the data transfer ends (FIG. 5(b)).

■ When the data phase ends, disk controller 1
1 executes the status phase STS and sends status information to the host device 200 to notify the execution result, such as whether the command ended normally or abnormally. Next, the message-in-phase MSGIb is executed and a command complete message is sent to the higher-level device 20. and disconnect from the SCSI bus 30 (see Figure 5).
a) STS and MSGI b).

As described above, the host device 20. The data read command received from the above ends normally.

In FIG. 5, T1 is a time determined by the value specified by the buffer full ratio, and indicates the time from the start of data reading from the magnetic disk until the start of disconnection with the SCSI bus 30. Further, T2 indicates the time from the end of reading data from the magnetic disk to the time when phases on the SCSI, such as the data phase and the status phase, are executed and the transition to the bus free state occurs.

The bumper control method described above is an example of an operation in an ideal case where the buffer 111 becomes empty and the SCSI bus 30 is not disconnected during data transfer to the host device 20o.

Next, the host device 20. buffer 11 during data transfer to
A buffer control method when SCSI bus 30 is disconnected when SCSI bus 30 becomes empty will be described with reference to FIG.

■ When data reading starts from disk 12o (
Time L. Each time data read for L1 blocks is completed, it is considered that storage of read data for one block in the buffer 111 is completed (FIG. 6(b), (C), times t0, . . . , t2, etc.).

t ■ When two blocks of read data at the specified buffer full ratio are stored in the buffer 111, the disk control unit l1 executes the reselection phase RS and performs the process of combining with the host device 20o ( Figure 6 (
a) RS).

■ When the selection phase RS ends, the data control unit 11 starts the message in phase MSG I
Execute step a on the host device 20. The host device 20 exchanges messages with the host device 20. A disk unit 12 performs data reading. (Figure 6 (
a) MSGIa).

Note that even during execution of selection phase RS and message in phase MSGIa, the disk unit 12
. The read data from is sent to the buffer 11 in block units.
1 (FIG. 6(b)).

■ Message-in-phase When the execution of MSGI a is completed, the data control unit 1l executes the data phase, and the data control unit 1l executes the data phase. The host device 20 reads data stored in the huffer 111 block by block while reading data. Transfer to.

The above operations from 1 to 2 are the same as those from 1 to 2 in FIG. 5.

■ As an environmental condition under which the present invention is effective, the host device 20
. The data transfer speed to the disk is 12. Since it is assumed that the data read speed from
Now, disk part 12. The data read from block 7 to block 7 is completed and stored in the buffer 111, and the host device 20 is reading the data of the next block 8. Assume that the data transfer up to block 7 has been completed for (in Figure 6)
,(C),L,).

When the disk control unit 1l detects that the buffer 111 is empty, it sends a message in phase MSG I
c and MSG T d, and sends the save data pointer message and disconnect message to the upper device 20. and performs processing to disconnect the SCS I bus 30 (MSGIc, MSG
Id).

However, there are only a few lead broncs left, so this message in phase MSGlc and MSG
Disk unit 12 while executing I d. The reading of block 8 from
Figures (b), (C), t4).

■ The disk control unit 1l is a disk unit ■2. When the data read from 1 is completed, reselection is performed as shown in t and after in FIG. 6(a) in the same manner as in the above
Phase RS and Message in Phase MSGT
Execute SCSI bus 30 and host device 20.a. ``I don't know what to do.''

■ Host device 20. When connected, the data control unit 11 executes the data phase and transfers the read data of the remaining block 8 stored in the huffer 111 to the host device 20. Transfer to.

■ When the data transfer of block 8 ends at time t6,
The disk control unit 11 receives the status phase STS and the message in the same manner as in the case (2) in FIG.
Run in-phase MSGIb to connect SCSI bus 3
Disconnect from 0.

As described above, buffer control when the buffer 111 becomes empty during data transfer to the host device 200 ends. The meanings of T and T2 shown in FIG. 6 are the same as those of T and T2 shown in FIG.

[Problem to be solved by the invention]

In a buffer control method for a disk device or a disk controller that has SCSI as an interface and reads from the disk device 10o via the buffer 111 according to the CCS specifications, the host device l2. When the buffer becomes empty during data transfer, the SCSI bus is temporarily disconnected, and when a specified amount of data has accumulated in the buffer 111, the process is performed to reconnect it again. Once the SCSI bus 30 is disconnected, there is no guarantee that reconnection will be possible immediately, since other devices will have acquired the right to use the SCSI bus 30 in the meantime. Therefore, the predetermined disk unit 12 is closed for the time required for this reconnection. SCS after finishing reading data from
The time required to execute the phase I and transition to bus free (time T2 in FIGS. 6 and 7) increases, and the time required to process the command increases.

In the conventional buffer control method, the host device 12. In order to prevent disconnection from occurring during data transfer, it is necessary to prevent the buffer 11l from becoming empty during data transfer when reading data. In other words, the parameter TI (the time determined by the value specified by the buffer full ratio; the time TI from the start of data read from the disk unit 12. to the start of connection with the SCSI bus 30, as shown in FIGS. 6 and 6). It is necessary to make the buffer 111 long (see FIG. 7) and start data transfer after enough code data has accumulated in the buffer 111.

This reconnection start time T, is specified by the time required to execute the selection phase RS and message in phase MSG Ia during reconnection, the data transfer rate on the SCSI, the disk unit read/write speed, and the command. It is desirable to consider factors such as the number of read/write blocks and to determine the minimum number of buffers without emptying the buffer during data transfer.

However, because the number of said elements is large and variable,
Since it is difficult to set the reconnection start time T1 to an optimal value so as to realize the ideal state shown in Fig. 5, in reality, it is necessary to take into account fluctuations in each element and set a certain margin for T. I have to.

If there is no margin for reconnection start time T1, the 6th
As explained in the figure, a situation where a huffer entity is detected before all the data of the specified number of blocks is read, and the SCSI bus 30 is disconnected even though there are only a few transfer blocks remaining, is likely to occur. . In this case, compared to the ideal case where disconnection does not occur during data transfer as shown in Figure 6, the time required for disconnection and reconnection, that is, the message in
Phase MSGI c, Message in Phase M
T2 increases by the execution time of each phase of SGI d, reselection phase RS, and message in phase MSGIa, resulting in a problem that the execution time of the read command increases.

On the other hand, T. If T2 is lengthened to provide a sufficient margin for the reconnection start time T, T2 will also be lengthened by the margin of T1, which will also increase the execution time of the read command. Since the increase in T2 due to taking too much margin for T occurs every time a read command is executed, the result is larger than the increase in T2 that sometimes occurs due to insufficient margin for TI, as shown in Figure 6, and the system performance decreases. The problem arises that the impact on Therefore, it is not possible to provide a large margin for T, and in that case, it is necessary to allow the situation shown in FIG. 6 to occur.

In this way, having SCSI as a conventional interface,
In the buffer control method of the disk device or the disk controller that reads/writes to the disk device via the buffer according to the CCS specifications, it is not possible to secure a large margin for the reconnection start time T1, and therefore, the buffer emptying is performed during data transfer. There is a problem in that a situation occurs in which the SCSI bus is disconnected, which increases the execution time of the read command and reduces throughput.

The present invention improves the throughput when reading data from a magnetic disk by delaying the start of SCSI bus disconnection for a certain period of time or more when a continuous read operation is in progress and the amount of remaining transfer data is less than a predetermined amount of data. It is an object of the present invention to provide a buffer control method for a disk device or a disk control unit that is improved.

[Means to solve the problem]

The means adopted by the present invention to solve the above-mentioned problems are as follows:
This will be explained with reference to FIG. FIG. 1 shows a principle diagram of the present invention. FIG. 1(A) is a block diagram showing the basic configuration of a system for implementing the present invention, and FIG. 1(B) is an operation timing chart of the present invention.

In FIG. 1(A), the same reference numerals are given to structures common to those of the system previously explained in FIG. 4.

That is, numeral 10 is a disk device that uses a magnetic disk, an optical disk, etc. as a recording medium, and only one of them is shown.

20 is a host device of the disk device 10, one of which
Only the stand is shown. Reference numeral 30 denotes a SCSI bus that connects each host device 20 and disk device 10.

The disk device 10 connects one of the plurality of disk units 12 (only one is shown) to the SCSI bus 30 by a disk control unit 11 equipped with a buffer 111, and transfers data between it and the host device 20. control.

The present invention is implemented by the system described above, and is constructed as follows.

In other words, when the interface has SCSI and a read/write command is received from the host device (20), the C
In a puffer control method of a disk device (lO) or a disk controller (11) that performs read/write to a disk unit (12) via a buffer (l11) according to CS specifications, (a) inside the buffer (111); The amount of data in the scs r bus (30) has become empty.
(b) When the timing to disconnect the scsi bus (30) is detected, if a continuous read operation is in progress and the remaining transfer data amount is less than a predetermined data amount, The start of disconnection of the SCSI bus is delayed by a certain period of time, and the remaining amount of data read within the certain period of time is transferred to the host device (20).

The predetermined data amount is the time required for disconnection and reconnection, that is, MSGIc and MSG in FIG.
It is a value obtained by dividing the total execution time of each phase of Id, RS, and MSGIa by (1) T3 (referred to as non-transfer time) that occurs for each block, and the unit is the number of blocks.

[For production]

The operation of the present invention will be explained with reference to the operation timing chart of FIG. 1(B).

(a) in Figure 1 (B) is the specification of SCS I, and the host device 2
0 (inishek) and disk mounting 710 (Yuichi get)
RS is the reseleccigon phase, MSGIa and M
SGIb is a message-in phase, and data phase is a phase for data transfer, the contents of which have already been described.

(B) of the same figure shows the state of change in the amount of data in the bumper 111 at each operation timing in reading, and (C) of the same figure (B) shows that data is read from the disk unit 12 in blocks. This shows how it is done.

Note that, as mentioned above, reconnect is performed when the specified buffer is full for the number of blocks specified in the read command.
It starts when the data is accumulated in the buffer 111 by the ratio, but in the case of FIG. 1(B), the number of read blocks is set to 8, and the mode select command is set so that the reconnection is performed when only 2 blocks have been read. Assume that buffer control parameters are specified.

To simplify the explanation, the processing related to the arbitration phase in which the data control unit lO0 acquires the right to use the SCSI bus 30 has a sufficiently shorter execution time than other phases, so the following explanation will not be given. Omitted.

■ Data reading is started from the disk unit 12 (time t), and ■ Each time data reading for a block is completed,
It is assumed that one block of read data has been stored in the buffer 111 ((B) and (C) in FIG. 1(B)).
, time to, tl, t2, etc.).

■ When two blocks of read data at the specified buffer full ratio are stored in the buffer 111, the disk control unit 11 executes the reselection phase RS to perform the process of combining with the host device 20 (the first Figure 1 (B
) of (a) RS).

■ When the reselection phase RS ends, the disk control unit 11 sends the message in phase MSG.
Ia is executed to exchange messages with the host device 20, and the host device 20 is notified of the machine number of the disk unit 12 from which data is to be read, that is, the ID number that identifies the location of the disk unit 12. MSC in Figure 1 (B) (a)
la).

In addition, the reselection phase RS and message
Even during the execution of in-phase MSGIa, the disk unit
Read data from 2 is sent to buffer 11 in block units.
1 ((b) in FIG. 1(B)).

■ When the execution of the message-in-phase MSG Ia is completed, the disk control unit 1l executes the data phase, and while reading data from the disk unit l2, transfers the data stored in the buffer 111 to the upper block in units of blocks. Transfer to device 20.

■ Suppose now that data read from the disk unit 11 to block 7 has been completed and stored in the buffer 111, and while data from the next block 8 is being read, data transfer to the host device 20 up to block 7 has been completed (the Figure 1 (B
), (C), t3).

The disk control unit 11 detects the timing to disconnect the SCSI bus when the amount of data in the buffer 111 is exhausted.

■ When timing to disconnect the SCSI bus is detected, the disk control unit l1 determines whether continuous read operation is in progress and the amount of remaining transfer data is less than the predetermined amount of data, and if both conditions are met, SCSI bus 30
The disconnection is extended for a certain non-transfer time T3 or more (T3 in FIG. 1(B)(b)).

■ When a continuous read operation is in progress, the remaining data of the remaining block 8 is read from the disk unit 12 during the non-transfer time T3 and is stored in the buffer III (Fig. 1 (B)).
(b) tz, t4, Ty).

■ Since the disk control unit 11 is executing the data phase, as soon as the read data is stored in the buffer 111, it resumes data transfer and sends the remaining data of the remaining block 8 to the host device 20 ((Fig. (B) (b
)'s).

■ When the transfer of the remaining data of block 8 is completed at time t5, the data control unit 11 transfers the status phase ST
S and message in phase MSGIb to disconnect from the SCSI bus 30.

As described above, even if the buffer 111 becomes empty while data is being transferred to the host device 20, the SCSI bus 30 will be disconnected from the disk drive if continuous read operations are in progress and the amount of remaining transfer data is less than the predetermined amount. The remaining data is loaded and transferred to the host device 20 without being connected.

In the above description of the operation, the disk control section 11 and the disk section 12 are provided separately, but the buffer control operation is the same even when the disk device 10 is made up of an integrated disk control section 11 and disk section 12.

As described above, even if the buffer 111 becomes empty during data transfer to the host device 20, if continuous read operation is in progress and the remaining transfer data amount is less than the predetermined data amount,
Since the disconnection of the SCSI bus 30 is extended for a certain period of time T3 or more, the remaining data can be transferred continuously without disconnecting the SCSI bus 30, so that the execution time of the read command can be shortened.

Since the timing of disconnection execution is controlled based on the amount of remaining data, it is possible to prevent the SCSI bus 30 from being inadvertently occupied and degrading system performance when there is a large amount of remaining data.

Furthermore, since no margin is required for the value of the reconnection start time T1 specified by the buffer control parameter, the time T2 from the end of data reading from the disk unit l2 to the transition to the bus free state is the minimum value. It is only necessary to determine the parameters so that the data read throughput can be improved.

Furthermore, although it is difficult to determine the optimal buffer control parameter, because the value of parameter T1 is erroneously set to a small value, buffer 111 during data transfer to host device 20
Even if s is empty, if the remaining transfer amount is small, s
Since the remaining data can be transferred continuously without disconnecting the CSR bus 30, the execution time of a read command can be significantly reduced compared to the conventional buffer control method.

[Example] An example of the present invention will be described with reference to FIGS. 1 to 3. FIG. 2 is an explanatory diagram of the structure of an implementation system according to an embodiment of the present invention, and FIG. 3 is an operational flowchart of the embodiment. FIG. 1 has already been explained.

(A) Configuration of the embodiment In FIG. 2, a disk device 10, a disk controller 1
1, the disk unit 12, the buffer Ill, the host device 20, and the SCSI path 30 are as described in FIG. 1(A). Although up to eight disk devices and host devices can be connected to the SCSI bus 30, only one of each is shown in the figure.

In the disk control unit 1l, the buffer ill is F
I F O (first in first out
) format in which reading and writing can be executed in parallel.

112 is a SCSI control unit Tearly consisting of an LSI chip;
It controls disconnection and reconnection with the SCS I path 30, communicates with the host device 20, and receives its commands. Reference numeral 112a denotes a byte counter provided in the wholesale section 112 according to the SCSI system, in which the number of bytes of data to be transferred with a read command is counted at the start of the data phase, and the number of bytes is calculated each time data transfer is performed. be done. Therefore, the count of the byte counter 112a indicates the number of bytes of remaining transfer data.

Reference numeral 113 denotes a read/write control unit consisting of an LSI chip, which controls writing the data stored in the buffer 111 to the disk unit 12 during writing, and stores the data read from the disk unit 12 in the buffer during reading. Take control. 113a is a read/write control unit 113
This is a status display register provided therein, and displays the status of the read/write control unit 113. By reading this status display register 113a, it is possible to determine whether the read/write control unit 113 is in continuous read mode.

114 is a buffer control unit, which controls the buffer 111 as an FI
Performs dual port DMA control for control as F○, and generates an interrupt signal when detecting that the buffer 111 is full or empty, and that the amount of data has reached a threshold according to the mode select parameter. and sends it to processors (MPU) 1 1 5.

The MPtJI 15 controls the operations of each section and the overall operation of the disk device 10, such as command analysis, control of read/write operations, and processing of interrupt signals.

116 is a ROM in which a control program for the MPU 115 is stored, and 117 is a RAM with a work area used by the MPU 115 during control processing. 118 is MPU
115 address/data paths.

(B) Operation of Embodiment The operation of FIG. 2, that is, the operation of one embodiment of the present invention, is
The operation sequence will be explained with reference to the timing chart in FIG. 3(B) and the operation flowchart in FIG. In addition,
As explained earlier, in the case of FIG. 1(B), it is assumed that the number of read blocks is 8 and the huffer control parameter of the mode select command is specified so that reconnection is performed when only 2 blocks have been read. .

To simplify the explanation, processing related to the arbitration phase in which the data control unit 10o acquires the right to use the SCSI bus 30 is omitted in the following explanation.

(2) When reading data, the read/write control unit 113 of the disk control unit 1l starts reading data from the disk unit 12 and stores it in the buffer 111 (time t). Every time data reading for one block is completed, the buffer 11
It is assumed that storage of one block of read data is completed at 1 ((b) and (C) in FIG. 1(B), time t.
t1, tz, etc.).

(2) When the buffer control unit 116 detects that two blocks of read data have been stored in the buffer 111 at the designated opening timing of the selection phase RS, it generates an interrupt signal to interrupt the MPU 115.

■ When the MPU 115 receives this interrupt signal, the
The SI control unit 112 is controlled to start the reselection process RS at the specified time or connection start time T1, and connect it to the host device 20 (RS in (a) of FIG. 1(B)).
).

■ When the reselection phase RS ends, the MP
U115 is the message in phase MSG I
Execute step a to exchange messages with the higher-level device 20, and notify the higher-level device 20 of the disk unit machine number from which data is to be read, that is, the ID number that identifies the disk unit I2 (Fig. 1 (B) ) (a) MSGIa).

In addition, reselection phase RS and menosage
Even during the execution of the in-phase MSGTa, data reading is performed by the read/write control unit 113, and the read data from the disk unit 12 is transferred to the huffer 1 in block units.
11 ((b) in FIG. 1(B)).

■ When the execution of the message in phase MSGIa is completed, the MPU 115 executes the data phase,
It instructs the puffer control unit 114 to start data transfer.

(2) Upon receiving this instruction, the buffer control unit 114 transfers the read data stored in the buffer 111 to the host device 20 by DMA transfer. During this data transfer, the read/write controller 114 is reading data from the disk unit 12 in parallel, and the buffer 11
1 is stored in blocks.

■ Since the data transfer speed to the host device 20 is faster than the data read speed from the disk unit 11, the data read from the disk unit 12 to block 7 has now been completed and is stored in the eight buffer l1l, and the next Suppose that the data transfer to block 7 from the host device 20 is completed while reading the data of block 8 ((b) in FIG. 1(B)).
), (C), 1:+).

When the buffer control unit 114 detects that the amount of data in the buffer 111 is empty, it generates an interrupt signal and issues an empty interrupt to the MPU 115.

Hereinafter, with reference to the processing flowchart of FIG. 3, the processing when the buffer is empty will be explained according to its processing steps.

[Phase] Step S: Upon receiving an empty interrupt, the MPU 115 sets the status display register 113a in the read/write control unit 113.
The read/write control unit 113 determines whether a continuous read operation is in progress.

When the read/write control unit 113 is not in continuous read operation, it is not expected that new data will immediately enter the buffer 111, so the process moves to step S6 and performs the same disconnection process as the conventional one.

(2) When the read/write control unit 113 is in continuous read operation, the MPU 115 reads the value of the bite counter 112a of the SCSI control unit 112, and determines whether the remaining number of broncs Bn is less than or equal to the expected value Bx.

This expected value Bx is determined by the buffer control method of the present invention.
The total non-transfer time T3 that occurs for each block is the time required for disconnection and reconnection (message in 7.zz MSGI c, MS after time t3 in Figure 6).
GI d, the time required to execute the reselection phase RS and the message in phase MSG I a).

If the number of remaining blocks Bn is larger than the expected value Bx, the SCSI bus can be used more effectively by performing disconnection and reconnection, so the process moves to step S6 and performs the same disconnection process as the conventional one.

@Step S3 Even when reading is in progress continuously, the reading process may be interrupted due to factors such as track switching or cylinder switching, or detection of an error in read data. Processing such interruptions generally takes time, so from the standpoint of system performance, the SCSI bus 30 should be opened to other devices. Therefore, in this case as well, the process moves to step S6 and the same disconnection process as in the prior art is performed.

Note that as long as data is being read continuously, a non-transfer time occurs on the SCS R bus 30 for T3 time every time one block of data is transferred. Since it is confirmed in S2, T2 can be shortened by not disconnecting.

@ Step S4 When the MPUI 15 is in continuous read operation and the remaining transfer data amount is less than the predetermined data amount, the MPUI 15 transfers the SCSI bus 30
(Figure 1)
B) (b) T,).

■ Step S5 Since the disk control unit 11 is executing the data phase, the huffer control unit 114 resumes data transfer as soon as the read data is stored in the buffer 111, and sends the remaining data of block 8 to the host device 20. Send ((th
Figure (B) (b) ts).

If there are a plurality of blocks of remaining data, the above-described process is repeated a plurality of times, and the non-transfer time T3 appears every time one block of data is transferred.

Now, to simplify the explanation, let the remaining number of blocks be l,
It is assumed that the transfer of the remaining data is completed after one non-transfer period T3.

■ Step S6 When the remaining data transfer is completed at time t5, the MPU 11
5 executes the status phase STS and the message in phase MSGIb to connect the SCSI bus 30.
Disconnect.

As described above, even if the buffer 111 becomes empty during data transfer to the host device 20, if the continuous read operation is in progress and the remaining transfer data amount is less than the predetermined data amount, the SCS I hash 30 The remaining data is read and transferred to the host device 20 without being disconnected.

Although one embodiment of the present invention has been described above, the present invention is not limited to this embodiment, and various modifications can be made in accordance with the gist of the invention. For example, instead of the buffer control unit 114 generating an interrupt signal when the buffer is empty, the MPU II 5 may detect the empty state of the buffer 111 by monitoring the state of the buffer control unit 114 via the address/data hash 18 using a program. You may also do so.

〔Effect of the invention〕

As explained above, according to the present invention, the following effects can be obtained.

(1) During data transfer to the host device 20, the buffer l11
Even if the SCSI is empty, if continuous read operation is in progress and the remaining transfer data amount is less than the specified data amount, the SCSI
Since the disconnection of the bus 30 is extended for a certain period of time or more, the remaining data can be transferred continuously without disconnecting the SCSI bus 30, so that the execution time of the read command can be shortened.

(2) Since the timing of disconnection execution is controlled based on the amount of remaining data, it is possible to prevent the SCSI bus 30 from being inadvertently occupied and degrading system performance when there is a large amount of remaining data.

(3) Since no margin is required for the value of reconnection start time T1 specified by the buffer control parameter, after data reading from the disk unit 12 is completed,
Parameters may be determined so that the time T2 until transition to bus-free status becomes the minimum value, and throughput during data reading can be improved.

(4) Although it is difficult to determine the optimal buffer control parameters, because the value of parameter T1 was erroneously set to a small value, the buffer 111
Even when the SCSI bus 30 becomes empty, the remaining data can be transferred continuously without disconnecting the SCSI bus 30, so the execution time of a read command can be significantly reduced compared to the conventional buffer control method.

[Brief explanation of drawings]

Fig. 1 is a diagram of the principle of the present invention, Fig. 2 is an explanatory diagram of the configuration of an implementation system of an embodiment of the present invention, Fig. 3 is a processing flowchart when the buffer is empty in the same embodiment, and Fig. 4 is a scs An explanatory diagram of the buffer control method of the conventional system with r, Fig. 5 is an operation timing chart of the buffer control method in the conventional system in an ideal state, and Fig. 6 is an operation timing chart of the buffer control method in the conventional system when the buffer is empty. This is a planning chart. In FIGS. 1 and 2, 10...disk device, 11...disk control unit,
12... Disk unit, 111... Buffer, 20...
...Host device, 30...SCS I bus. Trade L child 11171 Ha Nno Ani〉F゜-F4 Tokimata and Tari Frothe Mart Figure 3

Claims (1)

[Claims] Having SCSI as an interface, when a read/write command is received from a host device (20), the disk unit (1
2) A disk device (10
) or in the buffer control method of the disk control unit (11), (a) detects the timing to disconnect the SCSI bus (30) because the amount of data in the buffer (111) has fallen below the specified amount; (b) When the timing to disconnect the SCSI bus (30) is detected, if continuous read operation is in progress and the amount of remaining transfer data is less than a predetermined amount of data,
A disk device or a disk control unit characterized by delaying the start timing of SCSI bus disconnection by a certain period of time and transferring the remaining data amount read within the certain period of time to a host device (20) without going through the disconnection. buffer control method.