JPH0731643B2 - I/O control unit retry control scheme - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a retry control method for an input/output control device in a computer system, and in particular to a retry control method for an input/output control device suitable for the case where a fault related to the operation of an input/output control device having a sorting function or the like occurs when the input/output control device is operating across multiple channel programs.

PRIOR ART

In computer systems, in order to deal with momentary failures, software retries are generally used to recover from the failure when one occurs. In this case, the input/output control unit (IOC) usually retries from the beginning of the channel program where the failure occurred.

[Problem that the invention aims to solve]

In a typical IOC, the channel program is the basic unit of processing, so retrying on a channel program basis does not cause any logical contradictions. However, when the amount of data to be processed becomes large, such as in the case of recent IOCs with sorting functions, it becomes impossible to transfer all the data in one channel program, and it becomes necessary to transfer the data across multiple channel programs. For example, when sorting tens of thousands of pieces of data in an IOC,
If one attempts to transfer all the data using one channel program, it would be necessary to occupy a large amount of memory at once, which is unrealistic. For this reason, a more realistic method is usually to prepare a buffer area of an appropriate size in memory and transfer the data there. In this case, if the amount of data is large, it will not fit into the buffer area at once, so the channel program must be issued multiple times to carry out a series of processes. In such a case, if a failure occurs and a retry is attempted in the middle of the channel program, the retry will not be performed correctly. This will be explained below with a concrete example.

Here, as shown in FIG. 2, the IOC channel commands are defined as follows:

TOR...Transfers IOC orders (control data) that specify operations to the IOC. WRT...Transfers data (input data) that the IOC should process to the IOC.
(WRITE operation), RD... reads data processed by the IOC (output data) from the IOC (READ operation). An IOC order describes an order type and other information that specifies the operation the IOC should perform on input/output data (herein referred to as "processing" to distinguish it from READ/WRITE operations).

Figure 3 shows a sequence of channel programs for performing a series of processes. Here, a channel program begins with a TOR command to transfer an IOC order to the IOC, followed by multiple WRT commands. Also, a WRT command transfers one buffer (a data byte that stores a collection of multiple records is called a buffer) to the IOC. CC stands for command chain.

Now, the channel program sequence shown in Figure 3 is being executed.
Consider the case where an error occurs related to the operation of C, and a retry is made in the middle of the channel program. The conventional retry processing procedure for this case is shown in Figure 7. Figure 7 shows an example where an error occurs in the middle of the execution of the second channel program, and a retry is made, and the retry is made from the beginning of the second channel program.

In Fig. 7, up until the time of the failure, the IOC has already executed the original processing instructed by the TOR command for the buffer passed up to the third WRT command of the second channel program. For example, in the case of sorting, the data up to the time of the failure, i.e., all of the first channel program and the data transferred by the third WRT command of the second channel program, have already been sorted within the IOC. For this reason, if retrying is performed directly from the beginning of the second channel program, the same data will be duplicated and included in the processing result, and the correct processing result will not be obtained. On the other hand, when the internal processing executed up to that point is discarded, the processed data held by the IOC contains a mixture of data already transferred by the first channel program and data transferred by the second channel program, so the processing m executed by the first channel program will be discarded.
must be discarded, and when retried, processing must be started from the beginning of the first channel program.

In addition, since the IOC cannot obtain information that identifies the relationship between the channel program received from the upper device after an abnormal termination and the channel program that had been processed up until the time of the failure, it has had to deal with the situation by always discarding the results before the time of the failure.

SUMMARY OF THE PRESENT EMBODIMENT An object of the present invention is to provide a retry control method for solving the above problems on the IOC side.

[Means and actions for solving the problems]

The present invention adds identification information (e.g., a channel program serial number, a serial number for each order type) to a channel program, which allows the issuance order to be identified, and provides the input/output control device with a means for storing the identification information of the executed channel program and the number of processed operations (e.g., the number of processed buffers).

The input/output control device compares the identification information of the newly issued channel program with the identification information of the previously executed channel program to see if they are the same, and if they are the same, it recognizes that it is a retry process, and simply transfers the processed buffers in that channel program, skipping them without performing any other operation, and starts processing from the unprocessed buffers thereafter. This allows the retry process to be performed efficiently and correctly even when a series of processes is executed across multiple channel programs. Moreover, the device only needs to reissue the retry channel program with its identification information.

[Example]

An embodiment of the present invention will now be described with reference to the drawings.

Fig. 1 is a block diagram of one embodiment of the present invention. In Fig. 1, a main unit (CPU) 10 manages channel programs, starts an input/output control unit (IOC) 20, and sends channel commands to the IOC 20. The IOC 20 is made up of a working memory 30 which is a storage location for input data from the CPU 10 or output data to the CPU 10, a channel interface control unit 40 which controls transfer between the CPU 10, a storage unit 50 which stores processing status for retry control, a main unit (such as a sorting processor) 60 which performs operations specified by IOC orders for input data from the CPU 10, and a main control unit 70 which controls the entire IOC. Generally, the data storage location of the working memory 30 is divided into an input buffer area and an output buffer area, and the CP
Input data from U10 is stored in the input buffer area, and the processed data from the main unit 60, i.e., the output data to the CPU 10, is stored in the output buffer area to prevent data from being destroyed. The storage unit 50 may also be configured to be the working memory 30.

Here, the channel program is as shown in Figure 3.
The basic operation of the C20 is as follows: First, the CPU10 issues a TOR command to the I
Next, after receiving one buffer transfer from the CPU 10 by the WRT command, the main unit 60 is started and the specified operation is performed according to the IOC order for each buffer, and this process is repeated until all buffers are completed. In this way, processing for each buffer is completed.

Next, the retry control operation according to the present invention will be described.

4 shows the processing procedure of the first embodiment of the retry control method of the present invention. The CPU 10 assigns a channel program sequence number (CPNO) that can identify the issuance order of the channel programs to the IOC order transferred by the TOR command, and
0. In IOC20, the following is notified for each channel program:
The CPNO is stored in the memory unit 50, and further, the number of processed buffers (processed buffer number, PBNO) in the channel program is counted and stored in the memory unit 50. When a fault occurs, the CPU 10 retries the channel program from the beginning where the fault occurred, and
The IOC order transferred by the TOR command is notified to the IOC 20 again with the channel program sequence number (CPNO) being processed when the failure occurred. The IOC 20 checks whether the CPNO of the channel program sent this time is the same as the CPNO of the channel program executed immediately before that stored in the storage unit 50. If they are the same, it recognizes that it is a retry process and transfers the processed buffer (stored PBNO) in that channel program to the IOC 20.
For the buffer that is not yet processed, the transfer process is performed, but the original process is not performed. The original process is performed from the next unprocessed buffer. Figure 4 shows the case where an error occurs in the second channel program, as in the example in Figure 7, and a retry is performed.

Here, the operation of FIG. 1 will be explained in detail with reference to the example of FIG.

The channel program serial numbers are assigned consecutively starting from 1, and the storage area of the memory unit 50 for the channel program serial numbers is initially set to "0." In the case of the first channel program,
Since NO = 0, it is recognized that new CPNO ≠ old CPNO and this is not a retry. The CPU 10 starts the IOC 20 with one channel program. The channel interface control unit 40 first receives an IOC order transferred from the CPU 10 with a TOR command, and writes the IOC order into the working memory 30. Next, the memory unit 50 retrieves the CPNO (old CPN) of the immediately previous execution.
Next, the channel program sequence number (new CPNO) is read from the IOC order in the working memory 30 and stored in the memory unit
Write to 50 and store CPNO.

Next, the new CPNO is compared with the old CPNO. At this time, if the new CPNO ≠
In the case of the old CPNO, the number of processed buffers PBNO is set to 0, and the following processing is performed.

Transfer one buffer using the WRT command.
The data is stored in working memory 30 (input buffer area). Main unit 60 is started. Main unit 60 refers to the data in working memory 30 and performs the specified operation. When the operation of main unit 60 is completed, PBNO is counted up (PBNO = PBNO + 1). PBNO is stored in memory unit 50.

The above steps 1 to 4 are repeated until all WRT commands (all buffers) are completed.

Moreover, if the new CPNO is equal to the old CPNO, it is recognized as a retry process, the number of processed buffers (PBNO) is read from the memory unit 50, and the main unit 60 simply performs a transfer process to the working memory 30 (input buffer area) for each buffer by each WRT command for the number of PBNO, without starting up the main unit 60 (this is called SKIP processing).

After the SKIP process for the PBNO is completed, that is, after the retry process in which the original process immediately before the occurrence of the fault is omitted is completed, the following process is performed.

Transfer one buffer using the WRT command.
Store the data in the working memory 30 (input buffer area). Start the main unit 60. The main unit 60 refers to the data in the working memory 30 and performs the specified operation. When the operation of the main unit 60 is completed, count up the PBNO. Store the PBNO in the memory unit 50. The above processes to are repeated until all WRT commands (all buffers) are completed. Note that new CPNO = old CPNO
If the immediately preceding PBNO is 0, the SKIP process is not performed, and the above steps 1 to 4 are repeated until all WRT commands are completed.

5 shows the processing procedure of the second embodiment of the retry control method of the present invention. The CPU 10 assigns an OSY sequence number (ORNO) that can identify the issuing order for each order type (OSY) to the IOC order transferred by the TOR command, and notifies the IOC of this number.
The IOC stores the OSY and ORNO for each channel program, and further, counts how many of the buffers have been processed in the channel program, and stores the PBNO. When a retry process is performed after a fault occurs, the IOC 20 counts the OSY and ORNO for the current channel program.
If NO is the same as the OSY and ORNO of the stored channel program executed immediately before, it is recognized as a retry process, and the processed buffer in the channel program is simply transferred, and no other action is taken, and the process is started from the unprocessed buffer thereafter. Figure 5 also shows a case where an error occurs in the second channel program, causing a retry.

Fig. 6 shows the processing procedure of the third embodiment of the retry control method of the present invention. The example in Fig. 6 shows a case where a reset is performed due to a fault, and then a retry is performed. In this case, if the entire IOC is reset, the internal state executed up to that point is discarded, and a retry cannot be performed. Therefore, all the execution (internal) state required for the original processing is saved, and the other parts are reset. The processing thereafter is the same as that in Fig. 4 or 5.

[Effects of the Invention]

As described above, according to the present invention, identification information that allows the issuance order of channel programs issued by a host device to be recognized is displayed, and the input/output control device stores the identification information of the channel program in addition to the number of processed operations in the channel program being executed, thereby enabling the input/output control device, which performs a series of processes using a plurality of channel programs, to correctly perform retry processing, and
This has the advantage that the retry process can be completed in a short time, and also has the advantage that the host device only needs to display the identification information and reissue the retry channel program, and there is no need to reconfigure the channel program.

[Brief description of the drawings]

Fig. 1 is a block diagram of an embodiment of an input/output control device according to the present invention, Fig. 2 is a diagram showing an example of a channel command, Fig. 3 is a diagram showing an example of a multiple channel program performing a series of processes, Figs. 4 to 6 are diagrams showing the processing procedures of each embodiment of the retry control method of the present invention, and Fig. 7 is a diagram showing the processing procedures of a conventional retry control method. 10: main system device (CPU), 20: input/output control device (IOC), 30: working memory, 40: channel interface control unit, 50: storage unit, 60: main device, 70: main control unit.

Claims (1)

[Claims] [Claim 1] An input/output control device which performs processing related to data content such as rearrangement and modification in accordance with multiple channel programs issued by a host device, for data to be transferred by the channel programs, wherein identification information enabling the issuance order to be recognized is displayed in the channel programs, and the input/output control device is provided with a means for storing the identification information of the executed channel program and the number of operations processed within the channel programs, wherein the input/output control device compares the identification information of the newly issued channel program with the stored identification information of the channel program executed immediately prior, and if they match, recognizes it as a retry process, and in the retry process, terminates the operation without performing the original processing for data to be transferred by operations corresponding to the number of operations processed in the channel program, and then executes processing including the original processing from the subsequent unprocessed operations. A retry control method for an input/output control device, characterized in that