JP2822582B2 - Input/Output Processing Control Method - Google Patents

Description

[Detailed Description of the Invention] [Table of Contents] Overview Industrial Application Field Prior Art and Problems to be Solved by the Invention Means for Solving the Problems Actions Examples (Figs. 1 and 2) Effects of the Invention [Overview] A hardware (input/output processor (IOP)) under the control of a central processing unit (CPU) determines the path status of input/output processing and
The state of the I/O device is managed as a management information block (SCH), and when the central processing unit (CPU) executes an I/O command, the I/O control parameters specified by the I/O command are converted and set in the management information block (SCH), and the I/O processing request queue (SCH queue) and route selection are performed by the above-mentioned hardware using an I/O control method, and at least the logical above-mentioned queue (SCH queue) is selected for a group of I/O devices using the same route in the hardware.
and an input/output processing control method in a computer system which performs an attempt to start an input/output processing request having the highest priority in each queue (SCH queue), the method aims to solve, on the input/output processor (IOP) side, a problem in which a specific input/output processing request is made to wait for a long period of time due to synchronization between the processing speed on the central processing unit (CPU)/input/output processor (IOP) side and the busy release timing on the input/output control device (IOC) side, and when a busy condition is detected as a result of an attempt on a certain input/output processing request, which makes it impossible to start the input/output processing and keeps the input/output processing request in the queue (SCH queue), in the case of an input/output device busy condition, the input/output processing request is moved to a position in the queue (SCH queue) where it has a lower priority for start than input/output processing requests of other input/output devices which use the same path, and in the case of a path ...
The system is configured to manage the priority of the input/output processing request in the queue (SCH queue) so that it continues to be held at a high priority position, or to move the input/output processing request to a lower priority position in the queue than the input/output processing requests of other input/output devices if the elapsed time from the issuance of the input/output processing request does not exceed a certain value when an input/output processing start attempt is made, and to maintain the priority position in the queue at a high priority position if the elapsed time exceeds a certain value.

[Industrial application field]

The present invention relates to an input/output processing control method in a computer system in which an input/output processor (IOP) under a central processing unit (CPU) manages a queue of input/output processing requests and selects an input/output path that can be used for each input/output processing request. In particular, the present invention relates to a method for controlling input/output processing in a computer system in which an input/output processor (IOP) under a central processing unit (CPU) manages a queue of input/output processing requests and selects an input/output path that can be used for each input/output processing request.
The present invention relates to an input/output processing control method that solves the problem of a specific input/output processing request being kept waiting for a long period of time due to synchronization with the busy release timing of the other side.

Due to the diversification of data processing by recent computer systems, the number of input/output devices (I/O devices) connected to the computer systems has increased.
The number of I/O commands issued by the central processing unit (CPU) has also increased, and the format of I/O commands has changed from the traditional start I/O (SIO) to the
The central processing unit (CPU) selects the channel device (C) indicated by the operand of the start input/output (SIO) instruction.
H) and the operating status of the input/output device (I/O)
In the method of issuing the start input/output (SIO) command and having the channel device (CH) read the channel control word (CCW),
Due to the increasing overhead in the central processing unit (CPU), the central processing unit (CPU) now transmits a start subchannel (SSCH) command, which uses the number of a subchannel provided for an input/output device (I/O) as an operand, to the input/output processor (IOP), and transfers the selection of a path to the input/output device (I/O), which was previously performed by the central processing unit (CPU), to the input/output processor (IOP). The central processing unit (CPU) then transfers the selection of a path to the input/output device (I/O), which was previously performed by the central processing unit (CPU), to the input/output processor (IOP).
As soon as a response is received indicating that the start subchannel (SSCH) command from the START SUBCHANNEL (SSCH) command has been normally transmitted, the START SUBCHANNEL (SSCH) command is executed and the next command is executed.

Conventionally, when the processing speed of a central processing unit (CPU) is synchronized with the busy release timing of an input/output control unit (IOC), there has been a problem in that a specific input/output processing request is kept waiting for a long period of time.

In this case, in the conventional input processing control method described above, the problem was dealt with by shifting the timing of issuing the relevant input/output command in the input/output processing routine in the operating system (OS), etc.; however, as described above, in an input/output processing control method in which an input/output processor (IOP) is introduced to reduce the overhead related to input/output processing in the central processing unit (CPU) and the input/output processor (IOP) selects paths related to input/output processing, the operating system (OS) cannot recognize the occurrence of the above-mentioned synchronous state, so it has become necessary to deal with the problem on the input/output processor (IOP) side.

[Prior art and problems to be solved by the invention]

FIG. 3 is a diagram for explaining a conventional input/output processing control system, in which (a) shows an example of a system configuration, (b) shows an example of a subchannel format, (c) shows an example of a subchannel queue control block (SQCB) format, (d1) shows an example of subchannel queue processing, and (d2) shows an example of a subchannel queue control block (SQCB) format.
2) shows an example of input/output processor (IOP) startup logic,
(e) shows a schematic diagram of the synchronization problem.

In FIG. 1(a), an input/output processor (IOP) 2 includes:
In order to manage the status of all the I/O devices 51-53 and to manage I/O processing requests/processing based on I/O commands (SSCH) issued by the central processing unit (CPU) 1, management information blocks called subchannels (SCH) 201-203 which are created one-to-one for the I/O devices 51-53 are stored in an internal memory (not shown).

An example of a conventional format of the subchannels (SCH) 201 to 203 is shown in FIG.

The subchannels (SCH) 201 to 203 hold information for managing the status of the corresponding I/O devices 51 to 53, and also hold I/O processing requests made by software executed by the central processing unit (CPU) 1, and
Until the initiation of the I/O processing request 53 is completed, the request is put into a queue (subchannel queue) 20, which will be described later, and awaits processing (IOP processing) by the input/output processor (IOP) 2 regarding the initiation of the I/O processing request.

The above software runs on the central processing unit (CPU) 1. The software manages input/output requests (I/O requests) made by upper-level application programs in the form of queues for each of the input/output devices 51-53.

That is, when the input/output processor (IOP) 2 notifies the software by an input/output interrupt (I/O interrupt) that an input/output operation of a certain input/output device 51-53 has been completed, the software checks whether or not the next input/output request from that input/output device 51-53 is present in the queue, and if so, transmits the input/output processing request to the input/output processor (IOP) 2 by an input/output command (the SSCH command mentioned above).

At this time, as described above, the input/output control parameters indicated by the operands of the input/output command are converted into a format that is easy for the input/output processor (IOP) 2 to process, and set in the control information blocks (SCH) 201-203.

In FIG. 1A, 111, 112, ... shown in the central processing unit (CPU) 1 are input/output request groups corresponding to the input/output device 51, and form a queue.
Reference numerals 122, . . . , 131, 132, . . . denote input/output request groups corresponding to the input/output devices 52, 53. These input/output requests are sequentially transmitted to an input/output processor (IOP) 2 by the software.

Next, referring to FIG. 2B, the sub-channel (SCH) 201
The schematic configuration of 203 will be described.

As described above, the sub-channels (SCHs) 201 to 203 are
Corresponding to each of the above-mentioned input/output devices 51 to 53, they are provided in the internal memory of the input/output processor (IOP) 2, and by an input/output command (SSCH) from the software, the input/output control parameters indicated by the operands of the input/output command are converted and set in the subchannels (SCH) 201 to 203 at the positions indicated by the SCH No. of the input/output command.

In Fig. 1(b), the "I/O device status descriptor" is used to manage the status of the I/O devices 51-53, and describes information such as whether an I/O request exists within the subchannels (SCHs) 201-203, whether the I/O process is currently being executed, and if so, whether data transfer is being actively performed, or whether preparation/cleanup for the I/O process is being performed only on the I/O device 51-53 side without using the interface with the channels 100-101. If the most significant bit (shown with diagonal lines) in this "I/O device status descriptor" is '1', this indicates that the device is present in the queue (subchannel queue) 20, which will be described later.

In addition, the subchannels (SCH) 201 to 203 are provided with a "queue forward pointer" and a "queue backward pointer",
This is a queue pointer for managing the queue 20, and the forward and backward subchannels (SC
H) Identifiers 201 to 203.

The subchannels (SCH) 201 to 203 are at the top of the queue.
Or, if it is located at the end, the forward pointer or the backward pointer is set to the subchannel (SCH) 2
It is configured to contain the identifier 01 to 203 itself.

The "number of connection channels" indicates how many paths are available to reach the input/output devices 51 to 53. For example, 8
The binary value of the bits is described.

In the example shown in FIG. 1B, there are a maximum of eight paths. Normally, the I/O devices 51 to 53 are addressed by a single unit address (UA), so selecting one path means:
The purpose is to select one of the channels 100 .

In other words, when there are N paths for a certain I/O device 51-53, it means that the device is connected to N different channels 100-100.
The identifiers of 〜 are written in "Connection Channel 0" to "Connection Channel 7".

Input/Output Processor (IOP) 2 is a subchannel (SCH)
When attempting to start a subchannel (SCH) 201-
When the channel number 203 is acquired from the queue (subchannel queue) 20, one of the channels described in the subchannels (SCH) 201 to 203 that is currently available is selected, and the information described in the subchannels (SCH) 201 to 203 is transmitted to the selected channel 100 to 203. The channel number 203 is then used to transmit the information to the channel 100 to 203 that received the transmitted information.
uses the unit address (UA) in the information to perform bus arbitration for the input/output interface and start the corresponding input/output device 51 ~.

Generally, the above-mentioned queue (subchannel queue) of input/output processing requests created in the input/output processor (IOP) 2
20 is not just one.

If there is only one channel, even if a channel 100~ unrelated to the subchannel at the head of the queue is available, it cannot be linked to the queue processing, which is inefficient. On the other hand, if there are too many channels, management becomes complicated, and since the queues are selected by a round robin method or the like, the average waiting time may actually increase significantly.

For this reason, in many current computer systems, I/O devices 51-53 are shared between I/O control devices 10. In other words, the paths to I/O devices 51-53 rarely cross I/O control devices 10. Taking advantage of this fact, a queue (subchannel queue) 20 is created corresponding to the I/O control device 10.

Subchannels (SCH) 201 to 203 shown in FIG.
The "queue identification information (SQCB number)" in the table may be considered as a serial number within the system of the input/output control device 10 to which the input/output devices 51 to 53 are connected.

By setting a numerical value which uniquely specifies this input/output control device 10 within the system in the "queue identification information (SQCB number)" field of the subchannels (SCH) 201-203, for example at the time of power-on reset, it becomes possible to easily match the subchannels (SCH) 201-203 with the queue (subchannel queue) 20 when performing queue processing for the subchannels (SCH) 201-203.

The queue (subchannel queue) 20 is managed by a subchannel queue management block (SQCB) 21 which is stored in the internal memory of the input/output processor (IOP) 2 in correspondence with the input/output control devices 10-.

(C) The figure shows the subchannel queue control block (SQC
B) 21 format examples are shown.

In this figure, the "I/O control unit status descriptor" indicates the status of the I/O control unit 10 corresponding to this subchannel queue control block (SQCB) 21 in the I/O processor (IO
This is an information area for the I/O control device 10 to manage. It contains information about the physical protocols supported by the I/O control device 10,
Support function information, and further, whether or not the input/output control device 10 is currently in use, are described.

Next, the "number of subchannels in queue" describes the number of subchannels (SCH) 201 to 203 enqueued in the queue (subchannel queue) 20 managed by the subchannel queue management block (SQCB) 21.

The next "Head Subchannel Identifier" and "Tail Subchannel Identifier" fields contain identifiers that describe the head and tail subchannels (SCH) 201 to 203 of the queue 20.

The "forward SQCB pointer" and "backward SQCB pointer" refer to the queue 20 provided in correspondence with the input/output control device 10.
The length of the subchannel queue control block (SQCB) (the number of queued subchannels) is not '0'.
The contents of the SQCB list structure are entered as identifiers of the subchannel queue control blocks (SQCB) 21 located before and after the SQCB list structure.

This is because the structure of the subchannel queue is two-dimensional in a round-robin manner {(a)
(see input/output processor (IOP) 2 in the figure), and subchannels (SCHs) 201 to 202 which hold valid input/output processing requests.
Only the subchannel queue control blocks (SQCBs) 21 to which one or more subchannels 3 are connected are list-linked and scanned by the input/output processor (IOP) 2, whereby the first subchannel (SCH) 201-203 of the subchannel queue 20 managed by one of the subchannel queue control blocks (SQCBs) 21 is extracted and started.

Next, the central processing unit (CPU) 1 executes the I/O command (SSCH).
When the command is issued, the input/output processor (IOP) 2 selects the subchannel (S
CH) 201 to 203 are the above-mentioned queue (subchannel queue)
The process of enqueuing a subchannel (SCH) 201 to 203 from the queue and attempting to start it will be described with reference to (d1) and (d2).

First, in FIG. (d1), the central processing unit (CPU) 1
However, when the aforementioned I/O command (SSCH) is issued, the subchannel number (SCH No.) indicated by the I/O command (SSCH) is
Based on the above, the input/output control parameters indicated by the operands of the input/output command (SSCH) are converted into a format that is easy for the input/output processor (IOP) 2 to use, and set for the subchannels (SCH) 201 to 203 in question, and the previously set "S
The "QCB number" is recognized, and the request is enqueued in the subchannel queue 20 indicated by the "SQCB number."

And the subchannel queue control block (SQCB)
If the subchannel queue control block (SQCB) 21 is connected to the SQCB list structure, it will eventually be scanned by the input/output processor (IOP) 2. However, if there are no subchannels (SCH) 201-203 in the subchannel queue control block (SQCB) 21 and the subchannel is not connected to the SQCB list structure until that point, the subchannel queue control block (SQCB) 21 is incorporated at the end of the SQCB list structure and the subchannel enqueue process is terminated.

Next, as shown in FIG. (d2), the input/output processor (IOP) 2 scans the list structure in which the valid subchannel queue control blocks (SQCBs) 21 are linked,
The subchannels (SCH) 201 to 203 are retrieved, and the subchannel queue control block (SQCB) 21 pointed to by the SQCB pointer (not shown) is retrieved.
Fetch the subchannel queue management block (SQ
The number of queue subchannels in CB) 21 is '0'.
If so, it means that there is no I/O device 51~ under the control of the relevant I/O control device 10 and making an I/O processing request, so it dequeues from the list combination and fetches the subchannel queue control block (SQCB) 21 pointed to by the next SQCB pointer. {See processing steps 300-302, 310} The start/end of the SQCB list combination is pointed to by the subchannel queue control block-list control block (SQLCB) 22 in the internal memory of the input/output processor (IOP) 2, and when the SQCB pointer is moved to the next subchannel queue control block (SQCB) 21 in the above processing, if the backward pointer of the subchannel queue control block (SQCB) 21 is the SQCB itself, the SQLCB 22
The head identifier of the is fetched as a new pointer.

The above SQLCB22, SQCB21, and subchannel (SCH) 201 to
The two-dimensional structural diagram of the input/output processor (IOP) 2 in FIG.

In the above-mentioned subchannel queue management block (SQCB) 21, if the "number of queued subchannels" is not "0," this means that there is a subchannel (SCH) 201-203 in the subchannel queue 20 that is requesting input/output processing, so the subchannel (SCH) 201-203 at the top of the subchannel queue 20 managed by the subchannel queue management block (SQCB) 21 is fetched.

The above-mentioned channels 100 to 103 and the corresponding input/output devices 51 to 53 are described in the subchannels (SCH) 201 to 203.
If the device is busy or there is no available path, i.e., channel 100, then the subchannel (SCH)
201 to 203 are moved to the end of the subchannel queue.
{See process steps 303 to 305} In the above, if there is a device-free and available path, i.e., channel 100~, that channel 100~
The sub-channels (SCHs) 201 to 203 are transmitted and activated to the sub-channels (SCHs) 201 to 203.

If the startup fails (i.e., subchannel (SCH) 2
Even if the information in 01 is sufficient to start the device,
51-52 may still be busy or may be being accessed by another I/O control device 10), the subchannel (SCH) 201-203 is moved to the end of the subchannel queue 20, but if the start is successful, the subchannel (SCH) 201-203 is dequeued from the subchannel queue 20. {See process steps 304, 306-310} As described above, in the conventional I/O processing control system, since the subchannel queue 20 is managed by the round robin method, it is a natural consequence that the first-in, first-out (FIFO) is not satisfied overall, but the FIFO within the subchannel queue 20 under the subchannel queue management block (SQCB) 21 is not achieved either.

That is, in the IOP start logic flow shown in FIG. (d2) above, step 30 marked (with a dashed line mark) in the lower right corner
In 5,308, the subchannel queue management block (S
In the subchannel queue 20 under the QCB 21, the first subchannel (SCH) 201 ~ is moved to the end, and the first subchannel (SCH) 201 ~ is not necessarily executed first.

For this reason, the following events may occur:

In FIG. 3(a), for the input/output devices 51 to 53,
When multiple I/O requests 111-, 121-, 131- exist in a main memory device (not shown) of a central processing unit (CPU) 1, an I/O request (e.g., 111 shown in the figure) for an I/O device 51 is submitted to the I/O device 51 via a subchannel (SCH) 201 and activated.

When the I/O control unit 10 becomes busy due to this I/O processing, and an I/O request (e.g., 121) for the I/O device 52 is fetched by the I/O processor (IOP) 2 via the subchannel (SCH) 202, the subchannel (SCH) 202 is moved to the end of the subchannel queue 20 under the subchannel control block (SQCB) 21 corresponding to the I/O control unit 10, since the I/O control unit 10 is in a busy condition. (See step 305 in the flow of the above (d2) figure.) After this, the processing of the request 111 for the I/O device 51 is completed, and the software executed by the central processing unit (CPU) 1 immediately issues the next I/O request 112 for the I/O device 51.
The I/O request 11 is issued by an I/O command (SSCH).
The information of 2 is stored in the subchannel (SCH) 201 and enqueued at the end of the subchannel queue 20.

In general, the busy condition of the I/O control device 10 may not be released for a while even after the I/O processing is completed due to conditions on the I/O control device 10 side. In this case, if the subchannel (SCH) 202 corresponding to the I/O request 121 to the I/O device 52 again meets the busy condition, the subchannel (SCH) 202 in which the I/O request 121 is stored will be released.
is moved to the end again, so that the priority of the subchannel (SCH) 201 in which the next I/O request 112 for the newly input I/O device 51 is stored becomes higher.
The original priority relationships will be repeated.

If the central processing unit (CPU) 1/input/output processor (IO
If the processing speed on the P)2 side and the busy release timing on the I/O control device 10 side are synchronized in the above-mentioned relationship, the start of the subchannel (SCH) 201 will be successful again, and the processing of the I/O device 51 corresponding to the I/O request 112 will start. This relationship is shown diagrammatically in Figure 3(e). Here, the numbers in parentheses indicate the I/O request numbers.

When this synchronization phenomenon occurs, the subchannel queue
Subchannel (SCH) 201 at 20 and subchannel (SC
Therefore, the priority relationship with the subchannel (SCH) 202 remains the same (i.e., the subchannel (SCH) 201 has a higher priority), and the I/O device 52 corresponding to the I/O request 121 cannot operate for a long period of time.

The above example occurred due to a busy condition in the I/O control device 10. However, in a case where three or more I/O devices 51-53 are connected, as shown in FIG. 3(a), it is easy to imagine that a busy condition could also occur in channels 100-101.

This phenomenon occurs when the input/output processor (IOP) 2 becomes more sophisticated and is controlled by a microprogram, and the time it takes for one subchannel (SCH) 201 to be processed and the next subchannel 202 to be fetched becomes longer, or when the input/output processor (IOP) 2 becomes slower.
P) When the processing speed of the central processing unit (CPU) 1 side increases compared to the processing speed of the central processing unit (CPU) 2, a new I/O request to the same I/O device 51 that the central processing unit (CPU) 1 has is generated.
111 and 112 are transmitted to the input/output processor (IOP) 2 and input to the subchannel (SCH) 201, and the occurrence of the above-mentioned synchronization phenomenon also increases.

In the generation of computer systems in which input/output control was managed in an operating system (OS) executed by a central processing unit (CPU) 1, the above-mentioned synchronization problem was dealt with by, for example, shifting the timing of issuing an input/output command for an input/output processing request when an input/output processing request that was not being processed in the operating system (OS) in the synchronized state was detected.
If most of the input/output processing is performed by the input/output processor (IOP) 2 in order to reduce the overhead in the CPU 1, then since the input/output processing is an event that is invisible to the operating system (OS), measures must be taken on the input/output processor (IOP) 2 side.

In view of the above-mentioned conventional drawbacks, the present invention provides a computer system in which the management of a queue of input/output processing requests and the selection of an available input/output path for each input/output processing request are performed by hardware, for example, an input/output processor (IOP), as follows:
The problem of certain input/output processing requests being kept waiting for a long time due to the synchronization of the processing speed of the central processing unit (CPU)/input/output processor (IOP) and the timing of the input/output control unit (IOC) to release the busy state was solved by the input/output processor (IO
The object of the present invention is to provide an input/output processing control method that can solve this problem on the P side.

[Means for solving the problems]

The above problems are solved by an input/output processing control system configured as follows.

(1) The hardware under the control of the central processing unit (CPU) manages the path status of the I/O processing and the status of the I/O device as a control information block (SCH), and the central processing unit (CP
In a computer system using an I/O processing control method executed by the above hardware, when an I/O control unit (CPU) executes an I/O command, the I/O control parameters designated by the I/O command are converted, set in the control information block (SCH), and submitted, the management of a queue (SCH queue) of I/O processing requests and the selection of a path for the I/O processing are performed, and the hardware configures at least logical SCH queues for a group of I/O devices using the same path, and attempts to start an I/O processing request having the highest priority in each queue (SCH queue), when, as a result of the attempt on a certain I/O processing request, a busy condition is detected in which it is impossible to start the input/output processing and the I/O processing request remains in the queue (SCH queue), in the case of a busy condition of the I/O device, the I/O processing request is moved to a position in the queue (SCH queue) where it has a lower priority for start than I/O processing requests of other I/O devices using the same path, If the busy condition is a path busy condition, the I/O processing request is managed so as to continue to be held in a higher activation priority position in the SCH queue than I/O processing requests of other I/O devices using the same path.

(2) In the computer system, the hardware includes a means (A) for setting the time when the input/output processing request was issued by the central processing unit (CPU) in the management information block (SCH) for managing the input/output processing request, a means (B) for comparing the time information set by the means (A) in the management information block (SCH) with the current time to calculate the elapsed time since the input/output processing request was issued, and a means (C) for comparing the value calculated by the means (B) with a previously given value, and if the elapsed time does not exceed a certain value during an input/output processing start attempt, the input/output processing request is prioritized over input/output processing requests of other input/output devices using the same path.
The I/O request is moved to a position in the SCH queue where its priority is lowered, and if the elapsed time exceeds a certain fixed value, the I/O request is managed so as to be maintained in a higher priority position in the SCH queue than I/O processing requests of other I/O devices using the same path.

[Action]

That is, according to the present invention, in a computer system in which the queue (subchannel queue) management of input/output processing requests and the process of selecting an available input/output path for each input/output processing request are executed by hardware, for example, an input/output processor (IOP), the processing speed of the central processing unit (CPU)/input/output processor (IOP) side and the input/output control device (IO
C) synchronizes with the busy release timing on the I/O processor (IOP) side to solve the problem of a specific I/O processing request being kept waiting for a long time. This is due to the fact that the synchronization problem is caused by the subchannel queue not being fully controlled in the first-in, first-out (FIFO) manner.
For the I/O processing request condition, if the I/O device is busy, it is necessary to allow other I/O devices to use the route. Therefore, as in the conventional method, the subchannel is moved to the end of the subchannel queue under the subchannel queue control block (SQCB). However, if the route of the channel/I/O control device is busy,
In most cases, all I/O devices under the subchannel queue control block (SQCB) use the same route as the route currently permitted to be used, so there is little point in moving the subchannel (SCH) to the end of the subchannel queue. Therefore, in order to get as close as possible to first-in, first-out (FIFO) control, the subchannel (SCH) is left at the top of the subchannel queue.

This allows the priority between subchannels (SCH) in the subchannel queue to be partially fixed without increasing the hardware, for example, by only modifying a part of the microprogram in the input/output processor (IOP).
This has the effect of reducing the occurrence of events in which a specific subchannel (SCH) is not processed for a long period of time.

In addition, as a means of making the control in the subchannel queue as close as possible to first-in, first-out (FIFO) control,
is kept at the head of the subchannel queue under the subchannel queue control block (SQCB).

Therefore, by adding some hardware, such as a timer for monitoring time in the input/output processor (IOP), each subchannel (SC) in the subchannel queue can be
H) is partially fixed, and it is necessary to reduce the occurrence of events where a specific subchannel (SCH) is not processed for a long time. Also, by adopting this method, the residence time in the subchannel queue is averaged, so the disturbance to the system performance caused by the competition between jobs is reduced, and the system performance is improved.

[Example]

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a diagram showing one embodiment of the present invention, and FIG. 2 is a diagram showing another embodiment of the present invention.
2) shows the change processing step, and (b) shows the "current time entry area" added to the subchannel (SCH) 201~, which is the subchannel queue 2 in the input/output processor (IOP) 2.
In the process of enqueuing and dequeuing the subchannels (SCH) 201-0, the first-in, first-out (FIF)
In order to carry out the present invention, a means for controlling the temperature of the liquid crystal display (O) is required. Note that the same reference numerals denote the same objects throughout the drawings.

The following is an example of the computer system configuration shown in FIG.
The input/output processing control method of the present invention will be explained with reference to Figs. 1 and 2 while referring to the processing flows of (1) and (d2).

Even when the present invention is implemented, the basic operation of input/output processing, that is, management of the queue (subchannel queue) 20 of input/output processing requests and selection of an available input/output path for each input/output processing request, is performed in the input/output processor (IOP) 2 as follows:
Since there is nothing particularly different, we will omit the details here and instead explain the means to avoid the problem of a specific input/output processing request not being processed for a long time, i.e., the problem of it getting stuck in the subchannel queue 20, caused by the synchronization of the processing speed on the central processing unit (CPU)/input/output processor (IOP) side and the busy release timing on the input/output control device side.

First, one aspect of the present invention is shown in FIG.
This replaces a part of the microprogram of the input/output processor (IOP) startup logic explained in FIG. (d2), i.e., steps 305 and 308 marked (with a diagonal line mark) in the lower right corner "move the subchannel to the end of the queue", with the flow shown in FIG. 1.

That is, when the I/O processing condition of the fetched subchannel (SCH) 201 ~ is an I/O device busy condition, it is necessary to allow other I/O devices to use the path, so as in the conventional method, the subchannel (SCH) 201 ~ is moved to the end of the subchannel queue 20 under the subchannel queue control block (SQCB) 21. However, when a busy condition occurs regarding a path of a channel/I/O control device, in most cases, all of the I/O devices 51 ~ 53 under the subchannel queue control block (SQCB) 21 use the same path as the one currently permitted to be used. Therefore, focusing on the fact that there is little point in moving the subchannel (SCH) 201 ~ to the end of the subchannel queue 20, a first-in, first-out (FIFO) method is used.
In order to get as close as possible to the control, the subchannel (SC
H) 201 to be left at the top of the subchannel queue 20.

In this way, the control in the sub-channel queue 20 can be made as close as possible to first-in, first-out (FIFO) control, thereby reducing the occurrence of the above-mentioned synchronization problem.

Next, another embodiment will be described with reference to FIG.

First, in the subchannel enqueue processing flow shown in Figure 3 (d1), the step marked (with a diagonal line mark) in the lower left corner, "Enqueue the subchannel to the subchannel queue specified by the subchannel queue control block (SQCB)", is replaced with the flow shown in Figure 2 (a1).

That is, a timer (not shown) is provided in the input/output processor (IOP) 2, and the input/output command (SSCH) is
U) 1, the subchannel queue 2
When the timer is enqueued to 0, the value of the timer at that time is recorded in the "current time entry area" (see FIG. 2(b)) added to the subchannel (SCH) 201 ~.

Then, steps 305 and 308 (steps marked with a diagonal line at the bottom right) of "moving to the end of the queue of the subchannel" explained in FIG. 3(d2) are replaced with the flow shown in FIG. 2(a2).

That is, when attempting to start the subchannel (SCH) 201 ~, the enqueue time recorded in the "current time entry area" of the subchannel (SCH) 201 ~ is compared with the current time, and if the residence time of the subchannel (SCH) 201 ~ in the subchannel queue 20 exceeds a certain value, the subchannel is not moved to the end of the subchannel queue 20 under the subchannel queue control block (SQCB) 21,
Make sure to keep it at the top.

If the residence time is less than a certain value, the subchannel (SCH) 201 is queued as in the conventional method.
Move it to the end of 20.

In this way, the control in the sub-channel queue 20 can be made as close as possible to first-in, first-out (FIFO) control, thereby reducing the occurrence of the above-mentioned synchronization problem.

In this way, the present invention provides a method for enqueuing and dequeuing in a subchannel queue based on a round robin method, and does not achieve first-in, first-out (FIFO) control.
Due to the correlation between the processing speed of the input/output processor (IOP) and the timing of the I/O control device to release the busy state, a specific subchannel (SCH) may be busy for a long period of time.
Focusing on the fact that the subchannel may be held up in the subchannel queue, if the busy condition is an I/O path busy condition, the subchannel (SCH) is prevented from being moved to the end, or the time that the subchannel remains in the subchannel queue is monitored, and if the stay time exceeds a certain time, the subchannel (SCH) is prevented from being moved to the end, thereby approaching first-in, first-out (FIFO) control as much as possible.

[Effects of the Invention]

As explained above in detail, the input/output processing control method of the present invention is a method in which the hardware (input/output processor (IOP)) under the control of the central processing unit (CPU) records the path status of the input/output processing and the status of the input/output device in the management information block (SCH).
The central processing unit (CPU) manages the I/O instructions (SSC
H), converts the I/O control parameters indicated by the I/O command and stores the management information block (SCH)
The queue of I/O processing requests (SCH
In input/output processing in a computer system in which at least the logical SCH queues are configured in the hardware for managing input/output processing requests and selecting paths for the input/output processing, and in which an attempt is made to start an input/output processing request having the highest priority in each SCH queue, when a busy condition is detected as a result of an attempt on a certain input/output processing request that makes it impossible to start the input/output processing and that causes the input/output processing request to remain in the SCH queue, if the busy condition is a busy condition of the input/output device, the input/output processing request is moved to a position in the SCH queue where it has a lower priority for starting than input/output processing requests of other input/output devices using the same path, and if the busy condition is a path busy condition, the input/output processing request is managed so as to continue to be held in a position in the SCH queue where it has a higher priority for starting. Alternatively, when an input/output processing start attempt is made, if the elapsed time since the input/output processing request was issued does not exceed a certain fixed value, the input/output processing request is moved to a position in the queue where its priority is lower than that of input/output processing requests from other input/output devices, and if the certain fixed time has elapsed, its priority is maintained at a higher position in the queue. Therefore, in the former case, without increasing the amount of hardware, for example, by merely changing a portion of the microprogram inside the input/output processor (IOP), the priority between each subchannel (SCH) of the subchannel queue can be partially fixed, thereby reducing the occurrence of an event in which a specific subchannel (SCH) is not processed for a long time.

In the latter case, by adding some hardware such as a timer for monitoring time in the input/output processor (IOP), the priority between the subchannels (SCH) in the subchannel queue is partially fixed, which has the effect of reducing the occurrence of an event in which a specific subchannel (SCH) is not processed for a long time. Also, by adopting this method, the retention time in the subchannel queue is averaged, so that the disturbance to the system performance caused by the competition between jobs is reduced, and the system performance is improved.

[Brief explanation of the drawings] Fig. 1 shows one embodiment of the present invention, Fig. 2 shows another embodiment of the present invention, and Fig. 3 explains a conventional input/output processing control method. In the drawings, 1 is a central processing unit (CPU), 111, 112, -, 121, 122, - are input/output requests, 2 is an input/output processor (IOP), 20 is a queue (SCH queue) or subchannel queue, 21 is a subchannel queue control block (SQCB), 22 is a subchannel queue control block-list control block (SQLCB), 201-203 are subchannels (SCH), 10 is an input/output control unit, 51, -53 are input/output devices, 101, 101 are channels, and 300-310 are processing steps.

Claims (2)

(57) [Claims] Claim 1: The hardware (2) under the control of the central processing unit (CPU) (1) stores the path status of the input/output processing and the status of the input/output device (51-) in a management information block (SCH) (201
When the central processing unit (CPU) (1) executes an input/output command, it converts the input/output control parameters indicated by the input/output command and stores them in the management information block (SCH).
In a computer system using an input/output processing control method in which the above hardware (2) manages a queue (SCH queue) (20) of input/output processing requests set and submitted to a queue (201-) and selects a path for the input/output processing, and in the hardware (2), at least logical queues (SCH queues) (20) are configured for a group of input/output devices (51-) using the same path, and an attempt is made to start an input/output processing request having the highest priority among the queues (SCH queues) (20), when a busy condition is detected as a result of an attempt on a certain input/output processing request, which causes the input/output processing request to remain in the queue (SCH queue) (20), in the case of a busy condition of an input/output device (51-), the input/output processing request is sent to another input/output device (51-) using the same path,
In the case where the busy condition is a path busy condition, the I/O processing request is moved to a position in the queue (SCH queue) (20) where the priority of activation is lower than that of the I/O processing request of another I/O device (51) using the same path.
The input/output processing control method according to the present invention is characterized in that the input/output processing request (I/O request) is managed so that it continues to have a higher priority in the queue (SCH queue) (20) than the I/O processing requests (I/O requests (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, Claim 2: The hardware (2) under the control of the central processing unit (CPU) (1) stores the path status of the input/output processing and the status of the input/output device (51-) in a management information block (SCH) (201
When the central processing unit (CPU) (1) executes an input/output command, it converts the input/output control parameters indicated by the input/output command and stores them in the control information block (SC
In a computer system using an input/output processing control method in which the hardware (2) manages a queue (SCH queue) (20) of input/output processing requests set and entered in a central processing unit (CPU) (201-), and selects a path for the input/output processing, the hardware (2) configures at least a logical queue (SCH queue) (20) for a group of input/output devices (51-) using the same path, and attempts to start an input/output processing request having the highest priority among the queues (SCH queue) (20), the time when the input/output processing request was issued by a central processing unit (CPU) (1) is recorded in the management information block (SCH (201-)) for managing the input/output processing request,
a means (A) for setting a time information set by the means (A) in the management information block (SCH) (201-) to a current time, and a means (B) for calculating the elapsed time since the I/O processing request was issued, and a means (C) for comparing the value calculated by the means (B) with a value given in advance, and if the elapsed time does not exceed a certain value when an I/O processing start attempt is made, the I/O processing request is moved to a position in the queue (SCH queue) (20) where it has a lower priority than I/O processing requests of other I/O devices (51-) that use the same path, and if the elapsed time exceeds a certain value, the I/O request is moved to a position in the queue (SCH queue) (20) where it has a lower priority than I/O processing requests of other I/O devices (5) that use the same path.
0) to maintain a high priority.