JPH06105445B2 - Input/Output Control System - Google Patents

Description

Description: FIELD OF THE APPLICATION OF THE ART The present invention relates to an input/output control system, and more particularly to a method for performing input/output device start-up control, which is performed prior to data transfer processing, offline from a control device.

BACKGROUND OF THEINVENTION

Conventional magnetic tape devices are equipped with a vacuum column, which allows for high-speed start-up of the magnetic tape. In contrast, cartridge-type magnetic tape devices do not include the vacuum column, which is essential for high-speed start-up of the magnetic tape device, in order to reduce weight and size.
Therefore, the startup process is performed offline, and after startup,
When data transfer becomes possible, the magnetic tape device is reconnected to the control device. In this way, for conventional cartridge-type magnetic tape devices, a control method exactly similar to that for disk devices has been proposed. That is, in conventional I/O device control methods that perform startup processing offline, when the transfer path is free, a startup request is issued to all I/O devices that need to perform data transfer, and when startup processing is completed, the device is reconnected to the control device. Note that when startup processing is completed, if a path that is transferring data with another I/O device is occupied, restart startup processing will be started. In the case of disk devices, processing for one rotation will be started. Therefore,
The data transfer path was not always utilized efficiently (see Japanese Patent Laid-Open Publication No. 60-142456).

OBJECTS OF THEINVENTION

The object of the present invention is to provide an input/output control system which improves on the above-mentioned conventional problems and can efficiently use the data transfer path between the input/output device and the control device when the start-up process prior to the data transfer process is performed offline.

In order to achieve the above object, an input/output control system of the present invention comprises a plurality of input/output devices each of which starts startup processing offline and requests data transfer at a known data transfer rate, a central processing unit, a channel device connected to the central processing unit and selected by the central processing unit,
a control unit connected to the channel device, controlling data transfer between the channel device and the input/output devices, issuing a request to start a start-up process to any one of the input/output devices in response to a request from the channel device, and based on a known start-up time, a known data transfer rate, and a known data transfer amount,
The system is characterized by having an input/output control device including a calculation means for calculating the transfer end time of the input/output device, and a data transfer path connected between the input/output device and the input/output control device in order to transfer data of only one input/output device at a time, and further comprising means in the input/output control device for responding to a request from a channel device so that the start-up process of an input/output device other than the input/output device is completed at the data transfer completion time of the input/output device currently transferring data.

[Examples of the invention]

The principles and embodiments of the present invention will be described in detail below with reference to the drawings.

In the case of a disk device, since the disk is constantly rotating, it is difficult to control the time when data transfer starts. In contrast, in an I/O device such as a magnetic tape device, the start-up process is performed from the I/O device's stopped state until it reaches a constant speed, and this time is almost constant, so it is possible to control the time when data transfer starts.

Furthermore, the data transfer time of an input/output device has a certain distribution, and the amount of data to be transferred can generally be known in advance, so it is possible to predict the time at which the data transfer will be completed.
Therefore, the transfer process of the input/output device currently transferring data is
It is possible to control the start-up time for an I/O device so that another I/O device requiring I/O processing can start data transfer just at the time of completion, thereby making it possible to efficiently utilize the data transfer path.

To explain this operation in more detail, it is assumed that the input/output device, that is, the cartridge type magnetic tape device with the vacuum column removed, has the following characteristics:

(a) The time when data transfer starts can be controlled.
(b) The start-up processing start time can be calculated by counting backwards from the data transfer start time (because the time required for the start-up processing can be known in advance at the start time of the start-up processing).

The reason why the above characteristics can be obtained is that the start-up time of the cartridge type magnetic tape device is almost constant, as mentioned above. In this case, it is acceptable for the start-up times to differ for each start-up process.
This time must be known before starting the startup process. However, in the following explanation, for simplicity, it is assumed that the startup time is constant for all devices. In general, the data transfer time of an I/O device has a constant distribution, and the amount of data transfer is known at the start of the data transfer, so the data transfer completion time of the I/O device performing the data transfer can be calculated in advance. Therefore, the above (a) and (b)
For input/output devices having the above characteristics, start-up processing can be started so that data transfer for one input/output device can start at the time when data transfer for another input/output device is completed.

Here, the cartridge type magnetic tape device without the vacuum column is simply abbreviated as MT. If the time when data transfer ends and the data transfer path becomes free is T, and the start-up time of MT is t, then in order to start data transfer of a certain MT at time T, the start-up process must be started at time (T-t). However, at time (T-t),
In the case of a MT, the data transfer path may be occupied by another MT that is transferring data, so the MT cannot unconditionally issue a startup request. For this reason, the MT is equipped with a timer, which is set up to start the MT when the data transfer path is free.
By issuing a request for starting the start-up process after a certain number of hours, the start-up process of MT is started at time (T-t).

FIG. 2 is a block diagram of an input/output control system to which the present invention is applied.

Reference numeral 10 denotes a CPU, 11 denotes a main memory device, 12 denotes a channel, 13 denotes a control device, 14 denotes a data transfer path, 15 denotes an MT, and 16 denotes a timer.

In the case of a magnetic tape device without a MT (vacuum column), the startup time is long, so multiple blocks of data are read ahead and written in one I/O process. For this reason, a buffer is built into the control device 13, and the read ahead data is stored in this buffer, and the data received from the channel 12 is also stored in this buffer, and then written in a batch to the MT 15 group. However, since the startup control method that is the subject of this invention is not directly related to the presence or absence of a buffer, the case where no buffer is provided and no read ahead and no batch writing are performed will be described here. The I/O control method of this invention can also be applied to the case where a buffer is provided and read ahead and batch writing are performed.

FIG. 3 is a detailed block diagram of the control device shown in FIG.
The figure shows the internal structure of the queue shown in FIG.
This figure shows the internal structure of the MT management information section in FIG.

As shown in Fig. 3, the control device 13 is composed of a processor 20 which performs data transfer processing and control processing, a block 21 which indicates the current time, a queue 22, and an MT management information section 23. As shown in Fig. 4, the queue 22 has an array of identifiers (MTiDs) 30 of the MTs which have issued startup requests. The MTiDs are input to the processor 20 in order starting from the left side,
As shown in FIG. 5, the MT management information section 23
Each piece of information, such as data transfer bus idle time 40, MT data transfer rate 41, MT start-up time 42, and MT block length 43, is held.
These pieces of information are about each MT 16 in the MT group. The data transfer path free time 40 indicates the time when the data transfer path 14 is free. The MT data transfer rate 41, the MT start-up time 42
Each indicates the transfer speed and startup time of each MT in the MT group. Since this information is common to each MT, each is listed one by one. MT block length 43 indicates the block length of each MT in the MT group.

In this case, the data transfer path 14 is used not only for data transfer but also when issuing a start-up request. At this time, information on the unit time after which the start-up process will start is also transferred. All MTs 15 in the MT group do not have vacuum columns, and since the start-up process takes a long time, the start-up process is performed offline from the control device 13. Once the start-up process is complete, the running speed of MT 15 becomes constant, and data transfer becomes possible, it is reconnected to the control device 13.

FIG. 1 is a process flow chart in a processor of a control device showing one embodiment of the present invention.

In this case, since the processor 20 directly executes the data transfer process, the processor 20 starts execution when it receives an I/O request from the channel 12 (step 10).
1), or when a start-up completion notification is received for any of the MTs in the MT group (step 100). First, when a start-up completion notification for MT 15 is received, a predetermined procedure is carried out with channel 12 (step 102), and then data transfer is carried out (step 103). When the data transfer is completed, MTiD 30 at the head of queue 22 is removed from queue 22. Next, a check is made to see if there is another MT whose start-up has been completed (step 100).

Next, when an I/O request is received from the channel 12, in this case a start-up request is immediately issued to the MT (step 106). However, since it is necessary to specify how many units of time the start-up process should be started after, the start-up process start time is calculated prior to the issuance (step 105). Next, the data transfer path idle time is updated (step 107), and the process returns to step 100.

The start-up process start time is calculated as follows.

First, S defined by the following formula is calculated.

S=(data transfer bus idle time 40)--current time obtained from the clock 21)--
(MT startup time 42) ... (1) If S is equal to or less than 0, a request is issued to the corresponding MT 15 in the MT group to immediately enter startup processing. If not, a request is issued to start startup processing after S time (the processing up to this point is carried out in step 105).

The data transfer path free time 40 is updated as follows.
This is because the data transfer path 14 has been further reserved by the MT 15. The data transfer time of this MT 15 can be calculated by dividing the MT block length 43 corresponding to this MT 15 by the MT transfer speed 41. If the calculated value is u, the data transfer path 14 will be newly occupied for the time plus overhead time V.
If is greater than 0, the data transfer path free time is updated by the following formula.

(Data transfer path free time 40) ← (Data transfer path free time 40) + u + V ……
(2) If S is equal to or less than 0, this means that the MT startup process has started immediately, so the data transfer path idle time is updated according to the following formula:

(Data transfer path idle time 40)←(Current time obtained from the clock 21)+(MT start-up time 42)+u+V (3) Furthermore, the identifier of this MT 15 is registered as MTiD 30 at the end of the queue 22 (up to this point is step 107).
After this, the process returns to step 100.

In this way, in the conventional method, the devices that were the subject of input/output processing were started up in parallel, and data was transferred from the device that was ready for transfer. However,
If another device is currently transferring data when data transfer becomes possible, the start-up process must be performed again, resulting in a very poor use of the data transfer path. On the other hand, in this embodiment, the use of the data transfer path is extremely efficient, so performance can be expected to improve.

FIG. 6 is a block diagram of an input/output control system showing another embodiment of the present invention.

In Fig. 6, most of the devices are the same as those in Fig. 2, except that a control information path 17 is newly provided in addition to the data transfer path 14, and that the timer 16 is no longer necessary. Also, since the internal structure of the control device is different, the control device is designated here as 13A.

FIG. 7 is a diagram showing the internal configuration of the control device 13A shown in FIG.
This control device 13A includes a processor 20 for controlling data transfer processing, a data transfer device 28 for transferring data between the channel 12 and one MT 15 in the MT group, a clock 21 for indicating the current time, a timer completion flag 25, a timer device 24 for turning on the timer completion flag 25 after a specified unit time has elapsed, a queue 22 for queuing input/output requests, and an MT management information section 2
3, and a data transfer flag 27 indicating that the data transfer device 28 is in the process of transferring data. When the data transfer device 28 completes the transfer, it turns the data transfer flag 27 OFF.
do.

FIG. 8 is a block diagram of the queue 22 of FIG.

The queue 22 stores an identifier (MTiD) 30 for identifying an MT 15 in the MT group and a launch flag 31 indicating whether the launch process for the MT has been performed or not.
(When ON, it indicates that startup is in progress, and when OFF, it indicates that startup has not yet begun) is stored.

The internal structure of the MT management information section 23 in Fig. 7 is the same as that shown in Fig. 5. The data transfer path free time 40 indicating the time when the data transfer path 14 is free, the transfer speed 41 of the MT,
The start-up time 42 and the MT block length 43 are stored.

Data transfer path 14 and control information path 17 can each transfer data to only one MT 15 in the MT group at a time. Even in this case, since MT 15 is an MT without a vacuum column, the start-up process takes a long time, and therefore the start-up process of MT 15 is performed offline with control device 13A. Once the start-up process is complete, the running speed of MT 15 becomes constant, and data transfer becomes possible, it is reconnected to control device 13A.

9 and 10 are process flow charts of the processor of the control device of FIG.

The processor 20 starts processing when the start-up process of any of the MTs 15 in the MT group is completed (step 500), when the data transfer is completed (i.e., when the data transfer flag 27 is turned OFF, step 501), when the timer completion flag 25 is turned ON (step 502), or when an I/O request is received from the channel 12 (step 503). As is clear from the figure, the check as to whether the start-up process of the MT 15 is completed (i.e., step 500) has the highest priority. The processing executed in each case will be described below.

When the start-up process of any one of the MTs 15 in the MT group is completed, data transfer process between that MT 15 and the channel 12 is started. That is, after a predetermined process is performed with the channel 12 (step 504), the data transfer device 28 is started,
A data transfer process is executed (step 505).

Next, when the data transfer process is completed, that is, when the data transfer flag 27 is turned OFF, the data transfer flag 2
7 is turned ON (step 505A), and the MTiD 30 at the head of the queue 22 is removed from the queue 22 (step 50
6).

Next, when the timer completion flag 25 is turned ON, it is necessary to issue a wake-up processing request to one of the MTs 15 in the MT group. This is done by checking the wake-up flag 31 from the head of the queue 22, and when the first one with the wake-up flag 31 OFF is found, the corresponding MTiD 30 is the MT 15 to be wake-up processed. That is, the MTiD 30 with the wake-up flag 31 OFF is found (step 507), and this wake-up flag 31 is turned ON (step 508). Then, a wake-up request is issued to the corresponding MT 15 via the control information path 17 (step 509). The wake-up flag 31 is searched from the head of the queue 22, and an OFF one is found. If this is not found, a jump is made to step 513 (step 501). When it is found, it is necessary to prepare the wake-up of the corresponding MTiD 30. First, it is checked when this wake-up processing should be performed (step 511). The time S can be calculated as follows.

S = (data transfer path idle time 40) - (MT startup processing time 42) ... (4) After that, the timer device 24 counts down from the current time to a predetermined time (S').
The timer device 24 is started so as to turn on the timer completion flag 25 at the predetermined time S' (step 512).
is calculated as follows:

S' = S - (the current time obtained from the clock 21) ... (5) Furthermore, it becomes necessary to update the data path idle time 40 (step 513). This is because the M
This is because the data transfer path 14 is further reserved by MT 15 in the T group. The data transfer time of this MT 15 is
This can be calculated by dividing the MT block length 43 corresponding to the MT transfer rate 41. This value is called u. In reality, the data transfer path 14 will be newly occupied for a time period that includes overhead time V in addition to this. Therefore, the data transfer path idle time 40 is updated as follows:

(Data transfer path free time 40) ← (Data transfer path free time 40) + u + V ……
(6) After this, return to step 500.

Finally, the case where an I/O request is received from channel 12 (step 503) will be described.

First, it is checked whether the queue 22 is empty (step 514). If it is not empty, a jump is made to step 518, and if it is empty, a wake-up request is issued to this MT (step 515). After the wake-up request is issued, the data transfer path idle time 40 is set (step 516). The value of the data transfer path idle time is determined by the following formula:

(data transfer path idle time 40)←(current time obtained from the clock 21)+(MT startup time 42)+u+V (7) Note that u and V are the same as those defined previously.

After the data transfer path free time 40 set is completed,
The identifier (MTiD) 30 of the MT 15 is set in the queue 22 (step 517). After this, the process returns to step 500.

If the queue 22 is not empty, a search is made from the top of the queue 22 to find an MT 15 whose startup flag 31 is OFF (step 518). If such an MT 15 is found, a jump is made to step 522. If such an MT 15 is not found, a startup process is immediately performed for this MT 15 (step 519), or the timer device 24 is started to perform startup process after a fixed time (step 520).
This is determined based on the time of the data transfer path free time 40. In other words, even if the start-up process is started immediately, if the start-up completion time is later than the data transfer path free time 40, the start-up process must be started immediately. If this is not the case, the start-up process time must be delayed so that the start-up process is completed within the data transfer path free time 40.
Specifically, this determination in 9 is carried out as follows.

S=(data transfer path idle time 40)-(current time obtained from clock 21)-
(MT start-up time 42) ... (8) If S becomes 0 or less, the start-up process will start immediately, so jump to step 515. If S is greater than 0, a start-up request can be issued S hours from the present. For this reason, a request is issued to the timer device 24 to turn on the timer completion flag 25 after t hours (step 520). Next, this M
The identifier of T15 is added to the end of the queue 22 as MTiD30 (step 521).
1 is reset to OFF. Then, the process returns to step 500.

MT1 whose startup flag 31 is already OFF is in the queue 22.
If there is an MT 15, the timer device 24 has already been started for this MT 15, so the identifier of the MT 15 which has accepted the I/O request is added to the end of the queue 22 as the MTiD 30 (step 522). At this time, the start-up flag 31 is turned OFF. After this, the process returns to step 500.

In this way, in this embodiment, a dedicated path is provided for issuing a start-up request based on the fact that the data transfer time of the input/output device has a certain distribution, the amount of data to be transferred is known, and the data transfer completion time is predictable.
Since it is possible to control the start-up time so that other input/output devices requiring input/output processing will complete their start-up processing at exactly that time, efficient data transfer can be achieved.

[Effects of the Invention]

As described above, according to the present invention, for an input/output device that performs start-up processing offline prior to data transfer processing, it is possible to efficiently use the data transfer path between the control device, thereby enabling efficient data transfer.

[Brief description of the drawings]

FIG. 1 is a process flow chart of a control device showing one embodiment of the present invention, FIG. 2 is an overall block diagram of an input/output control system to which the present invention is applied, FIG. 3 is a detailed block diagram of the control device of FIG. 2, FIG. 4 is a block diagram of the queue of FIG. 3, FIG. 5 is a detailed block diagram of the MT management information section of FIG. 3, FIG. 6 is an overall block diagram of an input/output control system showing another embodiment of the present invention, FIG. 7 is a detailed block diagram of the control device of FIG. 6, FIG. 8 is a block diagram of the queue of FIG. 7, FIGS.
1 is a process flow chart of the control device shown in FIG. 10: CPU, 11: main memory device, 12: channel, 13, 13A: control device, 14: data transfer path, 15: input/output device (MT), 16: timer, 17: control information path, 20: processor, 21: clock, 22: queue, 23: MT management information section, 24: timer device, 25: timer completion flag, 27: data transfer flag, 2
8: Data transfer device.

Claims (4)

[Claims] [Claim 1] A computer system comprising: a plurality of input/output devices each of which starts a startup process offline and requests a data transfer at a known data transfer rate; a central processing unit; a channel device connected to said central processing unit and selected by said central processing unit; an input/output control device including calculation means connected to said channel device, which controls data transfer between said channel device and the input/output devices, and which issues a request to start startup process to any one of the input/output devices in response to a request from said channel device, and which calculates a transfer end time for said input/output devices based on a known startup start time, a known data transfer rate, and a known amount of data transferred;
an input/output control system having a data transfer path connected between the input/output device and an input/output control device in order to transfer only data from the input/output device, and comprising means in the input/output control device for responding to a request from a channel device so that start-up processing of an input/output device other than the input/output device is completed at the data transfer completion time of the input/output device currently transferring data. [Claim 2] In the input/output control system described in claim 1, each of the multiple input/output devices has a timer, and the input/output control device includes means for setting each of the timers to an arbitrary unit time and causing the input/output device to start start-up processing after the unit time has elapsed. 3. In the input/output control system according to claim 1 or 2, said input/output control device comprises means for calculating a time when one input/output device will end its data transfer time, and calculating a time when another input/output device can start start-up processing at that time based on a known start-up start time, and means for issuing a start-up start request to the other input/output device via the data transfer path when the data transfer path is not transferring data,
An input/output control system according to the present invention, characterized in that, when the input/output device receives the issued start request, it sets a timer of another input/output device to the calculated start time. 4. In the input/output control system according to claim 1, said input/output control device comprises: a calculation means for calculating a transfer end time of said input/output device based on a known start time of startup, a known data transfer rate, and a data transfer amount designated by a central processing unit; and a calculation means for calculating a transfer end time of said input/output device based on a known start time of startup, a known data transfer rate, and a data transfer amount designated by a central processing unit, and a calculation means for calculating a transfer end time of said input/output device based on a known start time of startup,
and means for issuing a startup request to another input/output device via said data transfer path.