JPH06105446B2 - Data transfer control method - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates to a data transfer control system, and in particular, a startup process prior to a data transfer process is performed off-line with a control device, and each startup process time is constant. The present invention relates to a data transfer control method of a plurality of input / output devices having a distribution.

[Background of the Invention]

I / O devices that perform startup processing offline with the control device and have a uniform distribution of startup processing time
The "input / output control system" (Japanese Patent Application No. 60-39868) proposed by the present inventors describes a start-up control system for efficiently using a data transfer path. In addition, pre-reading / collective writing can be considered as a method of using the data transfer path more effectively.
Since no consideration was given to bulk writing, the efficiency of use of the data transfer path was not always good.

[Object of the Invention]

An object of the present invention is to improve such a point that has not been considered in the past, and to provide a continuous data transfer path between an input / output device and a control device that performs a start-up process prior to the data transfer process offline with respect to the control device. The present invention provides a data transfer control method that enables efficient data transfer and realizes a buffer size that is used efficiently and that has a buffer size according to the amount of data for prefetching and bulk writing.

[Outline of Invention]

In order to achieve the above object, the data transfer control method of the present invention continuously performs data transfer of one input / output device and data transfer of another input device, and a third input is performed between these two data transfers. The start-up start time of the output device is calculated, and the calculated time is given to the timer provided in the input / output device. By performing such control, data transfer is sequentially performed and the utilization rate of the data transfer path is improved. Also,
After data transfer, read-ahead / collective writing control is performed from the time when the input / output device is physically operating, and the utilization rate of the data transfer path is improved.

Example of Invention

An embodiment of the present invention will be described in detail below with reference to the drawings.

In the present embodiment, prior to the data transfer processing, the startup processing that puts the MT in a readable / writable state is performed offline with respect to the control device, and the input / output device whose startup processing time has a constant distribution is targeted. And Here, a cartridge type magnetic tape device (hereinafter referred to as a cartridge type MT device) is used. Since the vacuum column, which is indispensable for high-speed startup of a magnetic tape (hereinafter referred to as MT), is removed from the cartridge MT device, the startup process needs to be performed off-line with the control device. Further, the startup processing time has a constant distribution, which satisfies the above condition. Hereinafter, the cartridge type MT device will be simply referred to as an MT device.

FIG. 1 is a hardware configuration diagram of an I / O system having the MT device to which the present invention is applied.

This I / O system includes a CPU 1 that controls various devices, a main storage device 2 that stores system data, a channel 3 that is connected to the CPU 1 and the main storage device 2, a control device 4 that controls information of lower devices, and a prefetch. .Buffer 5 for temporarily storing batch data, data transfer path 6 used for transferring various data, MT device group 7 for storing data, and each
It consists of a timer 8 connected to the MT device. Cartridge type MT equipment takes several times as long as the startup processing time to position the tape after data transfer. In order to prevent a decrease in the utilization rate of the data transfer path 6 due to this, it is necessary to pre-read / collect data. Therefore, the buffer 5 is connected to the control device 4 in order to temporarily store the data. Also, in order to prevent a decrease in the data utilization rate due to the startup process, a control method is adopted so that the data transfer of the next MT device can be started immediately after the data transfer of one MT device is completed. For that purpose, it is necessary to issue a start-up processing request so that when the previous MT apparatus finishes the data transfer, the start-up processing of the next MT apparatus is completed. However, when data is being transferred through the data transfer path 6, it is not possible to issue a request to stand up to another MT device. Therefore, when the data transfer path 6 is vacant, a request is issued in advance to start the start-up process after a unit of time. In order to perform such control, each
A timer 8 is provided in the MT device.

FIG. 2 is a detailed configuration diagram of the control device of FIG.

The control device 4 includes a processor (a) 10 that controls data transfer between the channel 3 and the buffer 5, a processor (b) 11 that controls data transfer between the buffer 5 and the MT device group 7,
Further, it is composed of an MT management information section 12, a buffer management information section 13, and a clock 14 which have information for controlling both processors. As mentioned above, the input / output device such as the cartridge type MT device cannot start up at high speed.
Even after the data transfer is completed, the start-up preparation (stopping and rewinding the tape after the data transfer) requires several times as long as the start-up processing time. Therefore, if the data transfer unit is small, the probability that all MT devices are in the startup process or ready for startup is high, and the utilization rate of the data transfer path 6 is reduced. Therefore, a processor (b) 11 that performs pre-reading and batch writing between the MT device group 7 and the buffer 5 and a processor (a) 10 that controls data transfer between the channel 3 and the buffer 5 are provided. Operate in parallel to improve system throughput.

FIG. 3 is a diagram showing the contents of the MT management information unit 12, and FIG. 4 is a diagram showing the contents of the buffer management information unit 13.

First, the MT management information unit 12 provided for each MT device will be described. The MTID 20 indicates which MT device is the information, and the MT status 21 indicates which of the MT device is in the non-active state, the mount state in which the MT device is loaded with the tape wound on the reel, the running state, and the rewind state. Indicates the state of. In the read / write state 22, the MT device is in the read state but is in the write state. The data amount 23 indicates the amount of data regarding the MT device existing in the buffer 5, and the buffer address 24 indicates the address of the leading data. Further, a startup flag 25 indicating whether the MT device is in the startup process, an operating flag 26 indicating whether the MT device is in operation, a pre-reading / bulk writing completion indicating that the pre-reading / bulk writing is completed Flag 2
7, a wait flag 28 for turning on when a processing request from the channel 3 is given, and a temporary flag 29.

Next, the buffer management information section 13 will be described. This is because the number of active MTs 40 that gives the number of active MT devices that change dynamically, the free buffer amount 41 indicating the amount of free buffers, the free buffer pointer 42 indicating the start address of the free buffers, and the data transfer path 6 are It is composed of a data transferable time 43 indicating a free time.

5 and 6 are time charts showing specific examples of the data transfer control system of the present invention.

Next, a data transfer control method using the control device 4 described above will be described. Cartridge type MT equipment does not have a vacuum column, so high-speed startup processing cannot be performed. Therefore, this startup processing is performed off-line with the control device 4. However, if the same control as that of such a disk device is performed, the data transfer path is used by another MT device after the completion of the start-up process (control device 4
Operation when it cannot be connected to) and the data transfer path 6
Is not used effectively. MT device is different from disk
Startup processing (corresponding to disk seek and search)
Time has a constant distribution. Therefore, if the start-up control is performed so that the data transfer of another MT device is started at the data transfer end time of one MT device (known if the data transfer amount is known before the data transfer starts), the data transfer path 6 can be used effectively. Specific control will be described with reference to FIG. In this figure, two time axes are shown. The upper axis shows which MT device occupies the data transfer path 6, and the lower axis shows which MT device the start-up process is performed at what time. The time on both the upper and lower axes is the same time. When it is known that the data transfer of one MT device (Ma) is performed from time t ₁ to t _3, it is desirable that the data transfer of another MT device (Mb) is performed from t ₃ . For that purpose, the start-up process of Mb must start from t ₂ . However, at time t ₂ , Ma data is being transferred, and the startup processing request cannot be issued through the data transfer path 6. To solve this problem, the timer 8 is provided. That gives the time t ₂ -t ₁ in the timer 8 connected to the Mb in the figure the time t ₁ (break of data transfer and data transfer). The timer 8 starts from t ₁ and t ₂ −
Issue a request to start up Mb after t ₁ . In this way, at the break between the data transfer of one MT device and the data transfer of another MT device,
The data transfer path 6 can be effectively used by adopting the control method of giving the start time to the timer 8 of the third MT device (Mc). Further, in order to perform such control, it is necessary to always know the time when the data transfer path 6 becomes free. This is indicated by the data transferable time 43 in the buffer management information section 13.

In addition, since the cartridge type MT device does not have a vacuum column, not only the start-up process but also the process after the end of data transfer requires several times as long as the start-up process time. This process means stopping and rewinding the tape, during which
Data transfer of this MT device becomes impossible. The decrease in the utilization rate of the data transfer path 6 due to this is prevented by the data transfer control method by prefetching / collective writing. Before describing this control method, designation of symbols is first performed.

s: Start-up processing time n: Number of active MTs t: MT data transfer rate FIG. 6 shows control when there are three MT devices (Ma, Mb, Mc). This figure also has four time axes as in FIG. The top axis shows the usage status of the data transfer path 6,
The other three axes represent the operation of each MT device. When the data transfer of the MT device from time t ₁ to t ₃ (Mx) has been performed, the data transfer Ma According to launch control described above
It starts from t ₃ . After the Ma data transfer ends at t ₅ ,
Inter 4s (until time t ₉₎ can not transfer data Ma. That is, after the currently executed data transfer is completed, the tape running is stopped (s), and the tape running before the stop is returned to the reverse so that the tape can be accessed after the last transferred data (s). , If the tape running is stopped (s) and the tape is started up from this state (s), time s is required at each stage, so it takes 4s until the current data transfer is completed and the next transfer becomes possible. Takes. Therefore, during this period, data transfer (pre-reading / reading by other MT devices (Mb, Mc))
If collectively written, the data transfer path 6 can be effectively used. Based on the above, the read-ahead / collective writing amount Vn is determined by the following equation.

Vn = 4 · s · t / (n−1) (n ≠ 1) (1) Processor a10 and processor b10 that perform the above-mentioned control
The processing flow chart of the above is shown in FIG. 7 to FIG. First, the 7th
The process flow diagram of the processor a10 will be described. The processor a10 starts its operation when there is an MT device which has been turned on from the prefetching / collective writing completion flag 27 (step 50) or when there is a processing request from the channel 3 (step 51). When there is an MT device in which the prefetch / collective writing completion flag 27 is ON, the processing shown in FIG. 8 is performed (step 7).
3). When there is a processing request from the channel 3, first,
It is checked whether or not the MT status 21 of the requested MT device is in the non-active state (step 52). If the MT status 21 is in the non-active state, the MT status 21 is set to the mount state (step 53). After that, the processing request is analyzed (step 54) to check whether it is a rewind request, a read request, or a write request.

If it is a rewind request, it is checked whether the read / write state 22 is a read state (step 55). In the read state, the pre-read data in the buffer 5 is no longer necessary, so the buffer amount occupied by this data is added to the empty buffer amount 41 (step 56). Next, the MT status 21 is set to the rewind state (step 57), and the number of active MT devices is reduced by one (step 58). At the same time, a new Vn is calculated according to equation (1) (step 59). After that, the processing shown in FIG. 9 is performed (step
60). The processing shown in FIG. 9 is for processing a write processing request that is awaited without a free buffer. If it is in the write state at step 55, the MT status 21 is set to the rewind state (step 61).

Next, the processing when the result of the processing request analysis is a read request in step 54 will be described. First, it is checked whether the read / write state 22 is off (step 62), and if it is off, the read state is set (step 63). Next, it is checked whether or not the requested data is in the buffer 5 (step 64), and if not, the wait flag 28 is turned on (step 65). If so, transfer the data (step
66), the free buffer amount 41 and the data amount 23 are updated (step 67). Thereafter, the processing shown in FIG. 9 is performed (step 68).

Furthermore, if the processing request is a write request in step 54, first check the read / write state (step 69),
If it is off, put it in the light state (step 70).
Next, it is checked whether or not there is an empty buffer in the buffer 5 (step 71), and if there is no empty, the wait flag 28
Turn on (step 65). If there is a space, the data is transferred (step 72) and the free buffer amount 41 and the data amount 23 are updated (step 73).

FIG. 8 shows a processing flow when there is an MT device in which the prefetch / collective writing completion flag 27 is ON in step 50 of FIG.

First, the pre-read / collective writing completion flag 27 is turned off (step 100), and it is checked whether the read / write state 22 is a read state (step 101). If it is in the write state, the process jumps to step 105. In the read state, it is further checked whether the wait flag 28 is on (step
102), the processing ends when it is off. When it is on, the data is transferred (step 103) and the free buffer amount 41 and the data amount 23 are updated (step 104).

FIG. 9 is a flow chart for processing a write request which is awaited when there is no free buffer when the buffer 5 is considered to be free. First, the temporary flag of all MT devices
Turn on 29 (step 150). Then the temporary flag 29
If the MT device is turned on (step 151), it is checked whether or not there is a free buffer (step 152). If there is not, the processing is terminated, and if there is, it is checked whether the read / write state 22 is a write state (step 153). If in the lead state, jump to step 158. In the light state,
The weight flag 28 is checked (step 154). If it is off, the process jumps to step 158, and if it is on, the weight flag 28 is turned off (step 155). Subsequently, data transfer is performed (step 156), and the free buffer amount 41 and the data amount 23
Is updated (step 157). Next, the temporary flags 29 are turned off (step 158), and it is checked whether all the temporary flags 29 are off (step 159). If all are off, the process is terminated, and if any MT device is on, the process returns to step 151.

10 to 12 are flowcharts of the processor b11.

Processor b11 repeats the processing shown in FIG.
First, check whether there is any completion notification from the MT device (step 200), and then check whether the MT status 21 has a mount or rewind condition. After the above processing is completed, read-ahead / bulk writing is performed according to the amount of data in the buffer 5. When any completion notice is sent from the MT device in step 200, the process shown in FIG. 12 is performed (step 201). If not, it is checked whether or not there is an MT device whose MT status 21 is mounted (step 202). If there is, a mount request is issued to that MT device (step 203), and it is further checked whether there is any MT device in the rewind state (step 20).
Four). If there is an MT device in the rewind state, the read / write state 22 is checked (step 205). If it is in the read state, a rewind request is issued (step 205).
206). When in the light state, the MT in buffer 5
All the data of the device are determined as the collective writing amount (transfer amount) (step 207), and the start-up flag 25 is turned on (step 208). The next data transferable time is calculated from the data transfer amount determined in step 207, the data transferable time 43 in the buffer management information unit 13 is updated, and a start request is issued to the MT (step 209). Next, the MT device having the largest data amount v23 (Vn / n-v in the case of read processing) is selected (step 210), and it is checked whether or not the startup flag 25 of the MT device is off (step 211). If it is on, it is checked whether or not there is an MT device with a large amount of data v (Vn / n-v in the case of read processing) (step 212), and if there is, an MT device is selected (step 213), Return to step 211. If not, return to step 200. Step two
If the start-up flag 25 is off at 11, the processing shown in FIG. 11 is performed (step 214). Then, the process returns to step 200.

FIG. 11 shows a method of issuing a prefetch / bulk writing request.

First, it is checked whether or not it is in the read state (step 250). In the read state, it is checked whether or not the empty buffer amount Ve in the buffer 5 is larger than the prefetch amount Vn at that time (step 25).
1). If it is small, the process is terminated, and if it is large, the data transfer amount is set to Vn (step 252). In the light state,
It is checked whether or not the data amount v is larger than the collective writing amount Vn (step 253). When the data amount v is small, the processing is ended, and when it is large, the data transfer amount is set to Vn (step 254). After determining the data transfer amount in steps 252 and 254, the start-up start time is calculated (step 255). The start-up start time is calculated as follows. The data transfer available time 43 is checked from the buffer management information unit 13, and the startup processing time is subtracted from this time. This time is given to the timer 8 and the start-up flag 25 is turned on (step 256). Next, the time when this data transfer ends is calculated from the data transfer amount determined in steps 252 and 254, and the data transfer available time 43
And issue a startup request to MT (step 25
7).

FIG. 12 shows the processing when some completion notice is sent from the MT device in step 200 of FIG. There are three completion notifications from the MT device: mount completion, startup processing completion, and rewind processing completion. The completion notification is analyzed (step 300), and if it is the mount completion notification, the MT status 211 is set to the execution state (step 301). Increase the number of active MT 40 by one,
(Step 302), the read-ahead / collective writing amount Vn, which is a function of the number n of active MTs, is updated (Step 303). In the case of the start-up completion notification, the start-up flag 25 is turned off (step 304) and the data is transferred (step 30).
Five). The pre-reading / collective writing completion flag 27 is turned on (step 306), and the free buffer amount 41 and the data amount 23 are updated (step 307). Next, it is checked whether or not the MT status 21 is in the rewind state (step 308). If not, the processing is ended, and if it is in the rewind state, a rewind request is issued (step 309). Decrease the number of active MTs by one (step 310) and update the read-ahead / bulk writing amount Vn (step 311). In the case of the rewind completion notification, the MT status 21 is set to the non-active state (step 312).

In this way, according to the present embodiment, in the data transfer path between the input / output device and the control device, which performs the start-up process prior to the data transfer process offline with respect to the control device, the buffer size according to the data amount is continuously transferred. Data transfer becomes possible.

〔The invention's effect〕

As described above, according to the present invention, continuous data transfer is possible at the minimum buffer size, so that improvement in system throughput and optimization of the buffer size can be expected.

[Brief description of drawings]

FIG. 1 is a block diagram of an I / O system applied to the present invention, and FIG.
FIG. 3 is a detailed block diagram of the inside of the control device of FIG. 1, and FIG.
FIG. 4 is a diagram showing the contents of the MT management information part in the control device, FIG. 4 is a diagram showing the contents of the buffer management information part in the control device of FIG. 2, and FIGS. 5 and 6 are the data transfer control method of the present invention. FIG. 7 to FIG. 12 are time charts showing specific examples of the above, and FIG. 4 ... control device, 5 ... buffer, 8 ... timer.

 ─────────────────────────────────────────────────── --- Continuation of the front page (72) Inventor Takashi Doi 2880 Kozu, Odawara-shi, Kanagawa Hitachi Ltd. Odawara factory (72) Inventor Toshifumi Nishimura 2880 Kozu, Odawara, Kanagawa Hitachi Odawara Plant Within

Claims (2)

[Claims] 1. A control device having a function of performing data transfer between a channel and a buffer, and between a buffer and an input / output device in parallel, a buffer, and a startup processing time each having a constant distribution, and the startup processing In the input / output system consisting of a plurality of input / output devices that are off-line with the control device, the start-up process of another input / output device is completed when the data transfer of any input / output device is completed, and the data transfer is completed. In order to start the data transfer, a means for setting the start processing start time to another I / O device at the break of data transfer of one I / O device is provided, and data is transferred so that one of the I / O devices is in the process of data transfer. A data transfer control method characterized by determining the amount of pre-reading and bulk writing. 2. The read-ahead / collective write amount is determined from the number of active input / output devices, the startup processing time of the input / output devices, and the data transfer rate of the input / output devices. The data transfer control method according to claim 1.