KR970007261B1 - High speed back-up processing method for ticom ñ - Google Patents

Abstract

A high speed backup treatment method for high speed medium computer is provided. The method comprises a first step of requesting a high speed backup process in a central processing unit(3); a second step of receiving the request of the backup process and classifying a backup process treatment configuration into a backup process between the same input/output controllers, a backup process between the same banks and a backup process between different banks in the same input/output controller, and a backup process between different input/output controller, and which respectively performs the backup processes.

Description

Fast Backup Processing Method for High Speed Medium Computers

1 is a block circuit diagram showing a configuration of a high speed medium computer system applied to the present invention.

2 is a block circuit diagram showing in detail the input / output processor of FIG.

3 is a flowchart illustrating a process of generating a backup process in the input / output processor of FIG.

Figure 4 shows a processing operation according to the backup mode according to the present invention, (a) is a flow chart of the backup processing operation between the internal device input and output controller (b) is a flow chart of the backup processing operation between the internal device bank. (c) is a flow chart of a backup processing operation between internal devices of the input / output processor.

5 is a flowchart illustrating a backup processing operation in a raw input / output processor according to the present invention.

6 is a flowchart of a backup processing operation in a target input / output processor according to the present invention.

* Explanation of symbols for main parts of the drawings

1: System Bus 2: Shared Memory (MEM)

3: Central Processing Unit (MPU) 4: System Controller (SCM)

5: I / O processor (IOP) 6,6a to 6d: I / O controller (SCSI)

7: Data storage device 7a: Character input and output device (console)

8: operating system 9: system monitor

10: I / O kernel 11: Flag wheel

12: Janggeum Wheel 13: Size Wheel

14: buffer area l5: system bus interface

16 Dynamic Memory Access (DMA) Controller 17 Input / Output Processor (CPU)

18: Local RAM (LRAM) 19: ROM

20,20 ': Bank 22: Message handler

24: command process 27: backup process

The present invention is to realize a high speed data copy by the weight of the input / output processor firmware regardless of the position of the device in copying data between devices for data storage in a high speed medium-sized computer (particularly Tycom 3). For example, when the operating system of the central processing unit or the monitor of the system controller requests a backup to the I / O processor according to the command format in the message transfer protocol, the raw I / O processor can determine the backup type according to the location of the device and process the backup. If the target device is located in the other I / O processor, the message is delivered to the target I / O processor. If the target device is located in the two I / O processors, the message is sent to the target I / O processor. High-speed, medium-sized computers that can It relates to a backup process bangeop.

In general, a high-speed medium-size computer (eg, Chicom 3) is a tightly coupled multiple processing system composed of a shared memory (2) and a plurality of central processing units (3) based on a system bus (1) as shown in FIG. (8) to maximize the processing power of the central processing unit (3) and the input / output processor (5) and the system controller (4) is equipped with an input / output controller (6) and a character input / output device (7a) for block input and output, respectively. An I / O processor is an intelligent processor that controls the matching and message processing method and the processing method and order of I / O request units.

In a typical computer system, data storage devices are used between devices for storing large data copying machines for the purpose of storing or storing data. These data are inputted from the raw device even though the data are transferred between devices without change. Since data transfers to space and user space and goes back to the target device, there is a problem that kernel memory allocation and retrieval is repeated, and inefficient process involving unnecessary data movement between memories.

The present invention has been made to solve the above problems and an object of the present invention is irrelevant to the position of the storage device when copying data between storage devices mounted on several input / output processors in a high speed medium computer system such as Tycom 3. By providing high speed data copying only by the function inside the input / output processor, it provides a high speed backup processing method of a high speed medium sized computer which can shorten the data movement path and improve the performance of the whole system.

In order to achieve the above object, a feature of the high-speed backup processing method of a high-speed medium computer according to the present invention is that a plurality of central processing units, a plurality of input / output processors, a system controller, and a shared memory are connected by one system bus to each processor. A high-speed medium computer system such as Tycom 3 that communicates by a message transmission method in a system, comprising: a first request for high-speed backup by message transmission to an I / O processor from a system monitor executed by an operating system running on a central processing unit or a system controller; And the same I / O controller, the same bank, between different banks, or between different I / O processors according to the location of the data storage device (the location of the raw device and the target device) in the I / O processor in response to the backup request of the first step. Backup operation by classifying backup processing type and determining each backup processing procedure A second step of performing. When the backup operation performed in the second step is a backup between the I / O processors, the target I / O processor receives a message, specifies a data transfer method using a synchronization method and a raw I / O processor in the shared memory, and outputs these data to the target device. The process of releasing all resources allocated at the end of the backup operation of the third step and receiving a message from the target I / O processor during backup between the I / O processors. And the fourth step of returning a message to the processor requesting the backup.

Hereinafter, a preferred embodiment of the high speed backup processing method of a high speed medium computer according to the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block circuit diagram showing a schematic configuration of a general high speed medium computer system to which the present invention is applied, and includes a plurality of central processing units 3 and a plurality of buffers equipped with a symmetric operating system 8 to maximize processing power. A shared memory (2) having built-in 11 to 14 so that various processes can share a memory, two system controllers (4) having a console (7a) and a system monitor (9) as character input / output devices; Four input / output processors (5) incorporating an input / output controller (6) and an input / output kernel (10), which are block input / output buses for inputting and outputting block data from the data storage device (7), are connected to one system bus (1). Showing the construction.

The system controller 4 or the central processing unit 3 can request a backup to the input / output processor 5 by a message transmission method, and the input / output processor 5 configures messages and packets regardless of the position of the apparatus. .

Only when the source device and the target device exist in different I / O processors, the I / O processor receives the shared memory 2, writes it to the address roll packet, passes it to the shared memory 2 through the two annular buffers, Data transfer is performed and the structure of the buffer consists of a flag wheel 11, a talk wheel 12, a size wheel 13 and a buffer area 14.

FIG. 2 is a detailed block circuit diagram of the input / output processor of FIG. 1. The input / output processor 17 connects to the system bus 1 through the system bus interface 15 using Intel's Pentium, and the two data banks ( 20), 20 'are connected to four input / output controllers 6a, 6b, 6c, and 6d, respectively, and a dynamic memory access controller (hereinafter referred to as DNA) controller 16 is connected to each other. A block program between the host area and the data bank is enabled, and a firmware program including a backup function of the input / output processing processor 17 is mounted in the ROM 19, and the input / output processing processor 17 performs a local RAM 18 during the backup function. ) As the stack and data area.

FIG. 3 is a flowchart illustrating a process of creating a backup process in the input / output processor of FIG. 2, and requesting a backup by a message transmission method from a system controller or a central processing unit. When a backup is requested (21), the input / output processing processor 17 drives the message handler 22, and at this time wakes up the instruction process 24 waiting for a message (23).

When the command process 24 analyzes the message and confirms that it is a backup request, it generates a backup process 27 for performing the backup function according to the fork function 26 which creates a new process, and the backup process 27 generates a packet. Analyzes the backup processing in the same device of the same I / O controller, the backup processing in the other I / O controllers connected to the same bank, the backup processing between the I / O controllers connected to the other banks in the same I / O processor, or the backup between other I / O processors. Perform the corresponding backup type during the process.

4 is a flowchart of the processing operation according to the backup mode according to the present invention.

(A) of FIG. 4 is an operation flow chart showing a backup processing method between the internal autonomous controllers of the same I / O controller. A twelfth step (S12) of copying data into the bank allocated in the eleventh step by the copy command of the second step and a thirteenth step (S13) of recovering the bank after repeating the twelfth step by the backup request amount.

(B) of FIG. 4 is a flowchart illustrating a backup processing method between other I / O controllers connected to the same bank. Twenty-second step S22, twenty-third step S23 for writing to the target input / output controller again, and twenty-fourth step S24 for repeating the twenty-second step and the twenty-third step by the backup request amount and then recovering the bank; Is made of.

FIG. 4 (c) is an operation flowchart showing a backup processing method between input / output controllers connected to different banks in the same input / output processor. First, step 31 (S31) in which a source bank and a target bank are allocated, and from the source input / output controller A thirty-second step (S32) of reading data from the original bank allocated in the thirty-first step; a thirty-third step (S33) of performing a memory copy from the source bank of the thirty-second source bank to the target bank; Step 34 (S33) of writing data of the target bank copied in step 33 to the target I / O controller; and step 35 of recovering the original bank and the target bank after repeating step 34 by the backup request amount ( S35).

5 is a flowchart illustrating a backup processing operation performed by a raw input / output processor in a backup processing method between other input / output processors according to the present invention. First, a source bank is allocated to construct and transmit a message requesting a backup to a target input / output processor. Steps 41 (S43-S44), 42 steps (S45-S47) of checking each end of the two buffers indicated in the shared memory and checking whether the backup request amount is completed or not, and backing up the findings of the 42nd step If the end is confirmed, mark the flag as end and release the tag to recover the original bank (steps S53 to S54). If the result of the investigation of the step 42 is in progress, check the flag and input the local buffer. A 44th step (S49) of checking whether or not it can be used for the purpose; When the buffer is available as a result of the 44th step, the data is read from the raw I / O processor into the bank and the data is transferred to the buffer of the main memory by DMA again. Step 46 (S50-S52), step 47 (S55) for repeating the steps 42 to 46 for the next buffer, and then repeating as much as the backup capacity, and backup processing operation after performing the step 47 When this is completed, the flag is marked with the end, and the 48th steps (S56 to S57) of recovering and ending the original bank are performed.

FIG. 6 is a flowchart of a backup processing operation performed in a target input / output processor in the form of backup between different input / output processors according to the present invention. First, step 51 (S58) in which a target bank is allocated, and two buffers indicated in the shared memory. Step 52 (S59) and (S62) for deciding whether to output each of the buffers by checking the flags of the buffer and presently outputting the data; Step 53 (S61) and. If the end of the flag of the buffer of step 52 is marked as the end of the step (54) (S66) and performs the termination procedure to recover the target bank to send the message to the source I / O processor, and the result of step 52 Step 55 (S63 to S65) of releasing the memory by sending a DMA to the target bank after the DMA is transferred to the target I / O processor and marking the output completion on the flag. Step 56 (S67), which repeats the process until the end is displayed in the flag, and when the backup processing is finished in the target I / O processor after performing step 56, the target bank is recovered and the message is raw. The seventy-seventh steps S68 to S69 are transmitted to the I / O processor.

Referring to the configuration as described above will be described the operation of the fast backup processing method of the high-speed medium-size computer according to the present invention.

First, when a backup is requested to the I / O processor through an interrupt through the system bus (1) by the message transmission method in the system controller or the central processing unit (21), a backup process is generated in the I / O processor as shown in FIG. The input / output processing processor 17 drives the message handler 22, and at this time wakes up the instruction process 24 waiting for a message (23).

When the command process 24 analyzes the message and confirms that it is a backup request, the command process 24 generates a backup process 27 for performing the backup function of the input / output controller according to the fork function 26 which creates a new process, and the backup process 27 ) Analyzes the packet and classifies the corresponding backup processing method to perform the backup operation.

First, in case of backup processing operation between devices installed in the same I / O controller, this is possible by the unique function of the I / O controller. After receiving a certain capacity from the bank (S11), the data is repeatedly copied as much as the backup request by the copy command of the I / O controller. After the operation (S12), the bank is recovered (S13).

Second, in case of backup processing operation between other I / O controllers connected to the same bank, a certain capacity is allocated from the corresponding bank (S21), read from the source I / O controller (S22), and write back to the target I / O controller (S23) as much as the backup request amount. After that, the bank is recovered (S24).

Third, in the case of backup processing operations between I / O controllers connected to different banks in the same I / O processor, first, the source bank and the target bank are allocated (S31), and data is read from the source bank allocated from the source I / O controller (S32). The message is copied to the target bank (S33), the data of the copied target bank is written to the target I / O controller (S34), and the original bank and the target bank are recovered (S35) after the backup request amount is repeated.

Fourth, in case of backup processing operation between different I / O processors, the raw I / O processor receives the original bank (S43), constructs and sends a message requesting backup to the target I / O processor (S44), and each of the two buffers indicated in the shared message. After the payment is made (S46) to check whether the backup request to the end (S47).

When the backup is confirmed as the end of the investigation (Y branch in step S47), the flag is marked as the end (S53), the unlock is released (S54), and the original bank is recovered (57) and terminated. Check (S49) flag in step S47 to check whether this buffer is available for input (N49 in S49). The data is transferred to the buffer of the memory (S50), the input flag is marked in the buffer flag (S51), the unlock is unlocked (S52), the above process is repeated for the next buffer, and the backup capacity is repeated again (S55).

When the backup is terminated by repeating the above process, the flag is marked as end (S56), and the original bank is recovered (S57) to end.

In case of a back-up operation between different I / O processors, when the target I / O processor receives a backup request from a raw I / O processor, the target bank is first allocated (S58), and then each of the two buffers indicated in the shared memory (S60). Then, the current flag of this buffer is examined (S56) to determine whether to output.

If there is no data in the buffer as a result of the flag check (0 branch of S62), the unlock is repeated (S61), and if the end of the flag of the buffer is marked (2 branch of S62), the unlock is released (S66). (S68) and a termination procedure for transmitting a message to the raw I / O processor (S69). If there is data to be output as a result of the flag check (l quarter of S62), the DMA is transferred to the target bank and then output to the target I / O processor. (S63), mark the completion of output on the flag (S64), release the lock (S65), repeat for the next buffer, repeat again until the flag is displayed (S67), and the backup process ends in the target I / O processor. When the target bank is retrieved (S68) and the message is transmitted to the raw I / O processor (S69), the message receiving handler of the raw I / O processor requests the backup. The whole process of backup is ended By conveying the message to the device.

As described above, according to the high speed backup processing method of a high speed medium-size computer according to the present invention, data transfer between data storage devices mounted on various input / output processors in a computer system such as Tycom 3 is performed regardless of the position of the device. Since the path can be shortened, high-quality data transmission is possible by reducing the incidence of bad data due to multi-level data transmission. Also, by reducing the number of interrupts on the system bus, the data transmission on the data bus is reduced, thereby increasing the available bandwidth. There is usefulness to improve performance.

Claims (6)

In a high speed medium computer system such as Tycom 3, in which a plurality of central processing units, a plurality of input / output processors, a system controller, and a shared memory are connected by a system bus, and communicate with each other by a message transmission method, the central processing unit According to the first step of requesting a high-speed backup by sending a message to the I / O processor from the system monitor performed by the operating system controller performed by the operating system, and receiving the backup request of the first step according to the position of the data storage device in the I / O processor The backup processing types are divided into backup processing between the same I / O controller, backup processing between the same bank, backup processing between different banks in the same I / O processor, and backup processing between different I / O processors. Step 2 to carry out the backup process When the backup process performed in the second step is a backup process between other I / O processors, the target I / O processor receives a message and specifies a data transfer method using a raw I / O processor and a synchronization technique in the shared memory and targets the data. The third step is performed until the process of outputting to the device is terminated, and all resources allocated when the backup operation of the third step is completed are released, and a message is returned from the target I / O processor during backup between the I / O processors and processed. And a fourth step of returning a message from the processor requesting the backup, including a process. The method of claim 1, wherein the backup processing between the same I / O controllers of the second step comprises: an eleventh step of allocating a predetermined capacity in a corresponding bank; and a data allocated to the bank allocated in the eleventh step by a copy command of the I / O controller. And a thirteenth step of copying the data, and a thirteenth step of recovering a bank after repeating the twelfth step by the backup request amount. The method of claim 1, wherein the backup processing between the other I / O controllers connected to the same bank in the second step comprises: a second step of allocating a predetermined capacity in the corresponding bank; and a twenty-second step of performing a read in the original I / O controller; And the twenty-third step of writing to the target input / output controller again; and the twenty-fourth step of repeating the twenty-second and twenty-second steps as the backup request amount and recovering the banks. Way. The method of claim 1, wherein the backup processing between the input / output controllers connected to the other banks in the same input / output processor of the second step comprises: step 31 of allocating a source bank and a target bank; A target step 32 of reading data from the allocated source bank; a step 33 of performing a memory copy from the source bank of step 32 to the target bank; and data of the target bank copied in step 33; And a thirty-fourth step of writing to the input / output controller, and a thirty-fifth step of recovering the original bank and the target bank after repeating the thirty-fourth step by the backup request amount. The method of claim 1, wherein the backup processing in the raw input / output processor of the third step comprises: step 41 of constructing and transmitting a message requesting a backup to the target input / output processor by receiving the original bank; Step 42 of checking each of the two buffers and checking whether or not the backup request amount is terminated, and if it is confirmed that the backup is completed, the flag is marked as end, the unlock is released, and the original bank is recovered. Step 43, and step 44, if a check result backup operation is in progress, checking a flag to check whether the buffer is currently available for input; and data in the result buffer of step 44 If there is no information, the 45th step of releasing the check is repeated and the data is transferred from the raw input / output processor to the bank when the buffering result of the 44th step is available. The process from step 42 to step 46 for reading the data and transferring the data to the buffer of the main memory by DMA again, indicating that there is input data in the buffer flag and releasing the lock, and the steps 42 to 46 for the next buffer. A fast medium medium comprising the step 47 of repeating and repeating as much as the backup capacity, and the step 48 of marking the flag as an end and recovering and retrieving the original bank when the backup processing operation is terminated after performing the step 47. How to handle fast backups of your computer. The method of claim 1, wherein the backup processing of the target input / output processor of the third step comprises: step 51 of allocating a target bank, locking each of the two buffers indicated in the shared memory, and examining a flag of the current buffer. Step 52 of determining whether or not to output the data; and if there is no data in the result buffer of step 52, releasing the check and repeating the check; Step 54, which releases the lock table bank and releases the message to the raw I / O processor; if there is data to be output as a result of the 52, the DMA is transferred to the target bank and then output to the target I / O processor. Marking that the output is completed on the lag to release the lock, and repeating the steps 52 to 55 for the next buffer to the flag. And a fifty-sixth step of repeating until the end is displayed; and fifty-seventh step of recovering the target bank and transmitting a message to the raw I / O processor when the backup I / O processor is terminated after the step 56 is performed. Fast backup processing method for a medium-sized computer.