JPH09305525A - Input/output processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable an input/output operation without dropping performance even when the connection distance of a host processor and an input/output device becomes from 100km to 200km. SOLUTION: A center channel (RIOPCCH) 11 of a host 1 has plural sets of buffers, holds all the communication control data concerning an input/output instruction from a program and all the communication data sent from this host 1 to a storage subsystem (IO) 5 in these buffers and collectively sends all these communication control data and all the communication data to a remote channel (RIOPRCH) 31 of a remote input/output processor(RIOP) 3. The RIOPRCH 31 holds both the received data and based on the communication control data, the communication data are transmitted through an input/output channel (CNC) 34 to the IO 5. Besides, based on the communication control data, the communication data are read from the IO 5 through the CNC 34 and held and these communication data are transmitted to the RIOPCCH 11. The RIOPCCH 11 writes all the received communication data in the area of HMS 12 designated from the program.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an input / output processing system which is particularly effective in a data communication system in which the connection distance between a host processor and a storage subsystem is long.

[0002]

2. Description of the Related Art In an increasingly sophisticated information processing society, the preparation of a computer system for a large earthquake has become an important issue. One of the challenges is data backup. For that purpose, a data backup center is constructed, and the center is connected to the main computer center 10
It is to separate from 0 to 200 km or more. Moreover, in order to cope with the advanced information processing society, it is necessary to recover from a disaster as soon as possible, and in order to do so, it is necessary to back up data in real time in parallel with normal business. This is because the closer the backup data is to the level of actual use, the more likely it will be to return to the state before the occurrence of the major earthquake. Furthermore, it is desirable that the backup does not affect the performance of normal business as much as possible.

As a data communication for long distance connection, there is a method of connecting a host processor and a storage device via a communication network such as a public line. However, due to the slow communication speed of the communication network, the overhead of the program, etc., the time required for the data transfer becomes large, and the backup of the data cannot be endured. As another method, SA22-72
02-01 "ESCON I / O interface"
In the input / output channel and the storage device of the host processor supporting the above, the distance between the channel and the storage device is 10
It is possible to extend the length to 0 to 200 km. But,
In the above interface, the longer the distance,
The time involved in backing up the data increases. This is because the simplest input / output command requires at least three round trips between the input / output channel and the storage device. If the transmission delay time of the optical cable used for the interface is delayed by 5 ns per 1 m, it is 200K.
For connection distance of m, (200 × 1000) × 5 ×
(3 × 2) = 6000000 [ns] = 6 [ms] also takes time. Data backup is usually performed by a plurality of I / O commands, and considering that a plurality of I / O commands are combined into one I / O command, backup by the interface extension method takes time. So much, I can't stand a very real-time backup.

[0004]

SUMMARY OF THE INVENTION An object of the present invention is to solve the above-mentioned problems, and the performance is not deteriorated so much even if the connection distance between the host processor and the storage subsystem becomes long. -To enable data communication and realize real-time backup of data.

[0005]

In order to solve the above problems, the present invention includes a host processor and a storage subsystem, and a remote input / output processing device connected between the host processor and the storage subsystem. An input / output processing system in a data communication system comprising: the host processor, for one input / output instruction instructed by a program, all communication control data relating to the instruction.
Data and all communication data sent from the host processor to the storage subsystem to the remote input / output processing device at once, and read from the storage subsystem based on the communication control data Means for receiving all communication data received from the remote input / output processing device and writing the data in an area designated by a program, wherein the remote input / output processing device receives input data transmitted from the host processor. Means for receiving and holding all the communication control data and all communication data associated with an output command, means for transmitting all communication data to the storage subsystem based on the communication control data, and the communication Means is provided for receiving communication data from the storage subsystem based on control data and transmitting the communication data to the host processor.

[0006]

Embodiments of the present invention will be described with reference to FIGS. FIG. 1 shows a schematic configuration of the present invention. A host processor (hereinafter, host) 1 capable of operating programs such as an operating system and an application program has a basic processing unit (hereinafter, BPU) 14 that executes instructions from the program, and a channel path selection process. Channel processing device (hereinafter, CH
P) 13 and remote input / output processing device (hereinafter RIOP) 3
It includes a center channel (hereinafter, RIOPCCH) 11 for remote input / output processing device and a main storage (hereinafter, HMS) 12 of a channel path connected to. A remote input / output processing unit remote channel (hereinafter referred to as RIOPRCH) 31 connected to the RIOPCCH 11 in the host 1 and an input / output device (to the RIOP 3 connected with a connection distance of 100 to 200 km from the host 1) Below, I
O) 5 input / output channel (hereinafter, CNC) 3 connected to
4, a channel processing device for remote input / output processing device (hereinafter, RCH) for controlling the RIO PRCH 31 and the CNC 34, etc.
P) 33 and a remote input / output device main storage (hereinafter, RMS) 32.

The host 1 and the RIOP 3 are connected by an optical fiber 2 having a transfer rate of 1 Gbps.
And IO5 are connected by an optical fiber 4 of 200 Mbps. NIKKEI COMPUTER 1994.6.13 is used as an optical fiber using 1 Gbps.
There is one described on page 155 of
SA22-7202-01 "ESC
There is an optical fiber adopted in "ON I / O interface". In addition, the CNC 34 described in this embodiment is
The protocol between the interface and IO5 is assumed to be the interface protocol. Furthermore, it is assumed that RIOP3 and IO5 are connected at a short distance (100 to 200 m).

BPU14, CHP13, R in the host 1
Each IOPCCH 11 can access the HMS 12. The BPU 14 and CHP 13 are the HMS 12
Communication is possible using the BPU-CHP communication area inside, and CHP13 and RIOPCCH11 are HMS12.
Communication is possible using the CHP-CH communication area inside. FIG. 6 shows the position of each communication area in the HMS 32. Also, BPU14, CHP13, RI
The OPCCH 11 obtains the address of the communication area associated with each when the power is turned on.

RIO PRCH31 and RC in RIO P3
The HP 33 and the CNC 34 can be accessed by the RMS 32, respectively. Also RCHP33 and RIOPRCH
31 is RCHP-RIO PRCH in RMS 32
Communication is possible using the communication area, and RCH33 and CN
C34 can communicate using the RCHP-CNC communication area in RMS32. FIG. 7 shows the position of each communication area in the RMS 32. Also, RCH
The P33, RIO PRCH 31, and CNC 34 obtain the addresses of the communication areas associated with them when the power is turned on.

Next, RIOPCCH11 and RIOPRCH
31 will be described with reference to FIG. RIOPCCH
Reference numeral 11 has a buffer. This buffer has a C section 111 used for sending a communication control data group relating to input / output operations to the RIO PRCH 31, and all data sent to the IO5.
The WD section 112 used to send the data to the RIOPRCH 31, and all the data received from the IO5 to the RIOPR.
RD section 113 used for receiving from CH31
And the R section 11 used when reporting the end of IO5
Consists of four. The C and R parts have a capacity of 1024 bytes, and the WD and RD parts have a capacity of 4096 bytes. On the other hand, RIOP
The RCH 31 also holds a similar buffer group, the C unit 311 that receives the communication control data, and the RIOPCCH.
RD unit 312 that receives data from the R11 and RIOPC
WD section 3 used for transmitting data to CH11
13 and an R section 314 used to send the end status to RIOPCCH 11. In addition, RIOPCCH11
And RIOPRCH31 have C part / WD part / R respectively
I have two sets of D section / R section. Each set is recognized as buffer numbers 0 and 1, for example.

RIOPCCH11 and RIOPCH31
Shall communicate by frame 6. Frame 6
Is shown in FIG. The frame 6 is composed of a control word (4 bytes), a variable length data section, and a CRC section for checking the validity of the frame. The type of the frame is indicated in bits 3-7 of the control word. The types are RIOPCCH11 to RIOPRC.
Transfer of C part to H31, RIOPCCH11 to RIO
Transfer of data to PRCH31 (WD112 to RD3
12), transfer of data from RIOPCH31 to RIOPCCH11 (WD313 to RD113), RIO
Transfer of R part from PRCH31 to RIOPCCH11,
There is. Also, the A bit of bit 0 of the control word is C
It is used at the time of transfer of the R and R parts. When transferring the C section, the data is transferred from RIOPCCH11 to RIOPRCH31.
Data is transferred, this bit is set to 1, otherwise it is set to 0.
Becomes Also RIOPCH1 to RIOPCCH1
On the other hand, when there is data transfer, the A bit of the control word of the frame at the time of transfer of the R section becomes 1, and when there is no data, it becomes 0. Bit 1 of the control word of the frame is not used and 0 is always stored.

The E bit of bit 2 of the control word of the frame becomes 1 in the case of the final frame related to the transfer of the buffer. The E bit is always 1 in the transfer of the C part and the R part. However, in the data transfer relating to RD and WD, there are cases where one frame transfer is not enough. In other words, if the amount of transferred data specified by the program exceeds 4096 bytes, there will be no choice but to use a plurality of frames. Therefore, in the case of a frame related to data transfer, the E bit may be 0. Frame control work
The buffer number is placed in bits 8 to 15 of the command. Since only two buffers are set in this embodiment, the buffer numbers are only 0 and 1. Bits 16 to 31 of the control word of the frame store the number of bytes of data transfer sent in the frame. The maximum transfer count in one frame is 4096. The data that is actually sent after the control word of the frame (C, RD, WD, R part data).
Is included. The length is the length indicated by the count section of the control word of the frame. C behind the data
RC data is added.

Each buffer of the RIOPCCH 11 can be accessed by the HMS 12 in the host 1, and the RIOPCCH 11 can be accessed by the RIOPCCH 11.
Each buffer of PRCH31 is RMS32 in RIOP3
Can be accessed. As shown in FIG. 7, the RMS corresponds to the C section 311, the RD section 312, the WD section 313, and the R section 314 of the buffer in the RIOPRCH 31.
Each storage area exists in 32. The capacity of each area is shown in FIG. Therefore, the total amount of data specified by the program must not exceed 128K bytes in total for write and read. Also, the CCW limit must not exceed 128 (1024/8).

An example of the case where an input / output instruction is issued from a program is shown below. As commands and communication control data, SA22-7201-00 "Enterprize S"
systems Architeture / 390 Pr
The one described in "inciples of Operation" is used. Start as an input / output command
Subchannel (SSCH) command is used, and communication control data is Channel Command Command Wor.
d (CCW) is used. The CCW format is shown in FIG. The program creates CCW in the program area of HMS12 before issuing the input / output command,
Issue an I / O command (SSCH command) to the hardware. When the instruction is issued, the BPU 14 performs a process related to the instruction, and informs the CHP 13 of information regarding the instruction.

Upon receiving the information, the CHP 13 selects an appropriate channel path based on the information. This case
In this case, RIOPCCH 11 is selected and information such as the CCW storage address is transmitted. When the RIOPCCH 11 receives the information, the RIOPCCH 11 has two buffers (number 0 and number 1).
Check the free status of the buffer. If none of the buffers is empty, the RIOPCCH 11 notifies the CHP 13 that the execution of the process is pending. Upon receiving the notification, the CHP 13 will retry the RIOPCCH 11 after a certain period of time. If either buffer is empty, the input / output instruction is executed by the empty buffer.

The RIOPCCH 11 first fetches all CCWs related to the input / output instruction into the C section 111 of the buffer.
Then, the CCW is RIO according to the frame structure of FIG.
It transmits to the C section 311 of the buffer of PRCH31. In addition, RIOPCCH11 is based on the CCW,
All the data to be sent to IO5 is fetched into the WD portion of the buffer every 4096 bytes and sequentially transmitted to the RD portion 313 of the buffer in the RIO PRCH 31. afterwards,
RIOPCCH11 is the data from RIOPCH31.
Data and the transfer of the R section are awaited. At this time, if another buffer is empty, a new input / output command can be accepted.

RIOPRCH31 is RIOPCCH1
When the data of the C section 111 is received from 1, the data is written in the C section storage area in the RMS 32 based on the buffer number in the frame. In addition, the control work of the frame of the C section
If the A bit in the command is 1, then RIOPCCH1
Data sent from 1 (WD112 to RD312)
Wait) The completion of the transfer is waited until the E bit of the control word of the frame becomes 1. The data is RMS3.
It is stored in the RD section storage area corresponding to the buffer in FIG. When the RIOPRCH 31 recognizes the above completion, it sends a report to the RCHP 33. In the report, the buffer number in the received control word of the frame is notified to the RCHP 33.

When the RCHP 33 receives the report, it converts the value of the data address part of the CCW according to the RMS 32. This is the CCW sent to HMS1 in host 1.
This is because the data address created in 2 needs to be converted for RMS32. RCH
When the conversion is completed, P33 selects CNC34 as the channel connected to IO5, and RIOPCCH1
The operation is performed on the basis of the CCW of the C section sent from 1.

At that time, RCHP33 is RIOPRCH
The buffer number notified from 31 is notified to the CNC 34. The CNC 34 holds this buffer number until the following operations are completed.

When the CNC 34 receives the notification, the RIOPC
CCW (RCH of C section 311 sent from CH11
The following operation is performed based on the buffer number 0 or 1 (C section / RD section / WD section / R section is selected) according to the buffer number taught from P33. The processing is SA22-720
2-01 "ESCON I / O interface"
The data sent from the CNC 34 to the IO 5 is the W sent from the RIOPCCH 11.
Data of the D section 112 (RD as the RIO PRCH 31
(Data of section 312) is used, and the data received from IO5 is RM.
It is stored in the WD section storage area in S32. Also CNC
The end report (device status byte, etc.) between 34 and IO5 is stored in the R section storage area in the RMS 32 by the CNC 34. If no data is received from IO5, as can be seen in the R section,
The received data count value is set to zero. CNC34
If there is data received from IO5, sets the count value of all the data in the R section, and if CNC3
If there is no data received by No. 4, the count value is set to zero, and RCHP indicates that the processing is completed.
Notify 33. At this time, the CNC 34 is
RCH the buffer number related to the operation performed by 34
Notify P33.

Upon receiving the above notification from the CNC 34, the RCHP 33 further notifies the RIO PRCH 31 of the notification (also notifies the related buffer number). Upon receiving the notification, the RIO PRCH 31 checks by the data of the R section 314 whether the data stored in the RMS 32 is valid. If there is no valid data, the RIOPRCH31 does not send the data to the RIOPCCH11, but sets the A bit of the control word = 0 and the data of the R section 314 (the CNC 34 sets the R section storage area of the RMS32). (The data of the R section 314 is taken into the R section 314 in the RIOPCH 31)) and the R section 114 of the RIOPCCH 11
Transfer to. If there is valid data, RIOPRCH31 sends it to WD in RIOPRCH31.
Taken into the part 313, RD of RIOPCCH11
Then, the data of the R section 314 is sent with the A bit of the control word = 1. This completes the operation of RIOP3 relating to the input / output instruction.

The RIOPCCH 11 has an R section 114 and an RD.
When the section 113 receives the data of the R section 314 and the data of the WD section 314, the reception data is stored in the HMS based on the CCW set by the program stored in the C section 111. However, if the A bit of the control word of the frame relating to the transfer of the received R portion is 0, it is not stored. Then, the input / output instruction issued by the program is completed according to the value of the R portion.

The flow of these operations will be summarized with reference to FIG. FIG. 5 shows an example of a command chain of write-1 / read / write-2. The write here means a command accompanied by data transfer from the RIOPCCH 11 to the IO5, and the read means the IO5 to the RI.
A command accompanied by data transfer to the OPCCH 11 is shown. The above operation is one input / output command 1 (SSCH
In (1)), the program instructs the hardware by using a plurality of CCWs (see FIG. 2). That is, each of write-1 / read / write-2 corresponds to one CCW, and the command chain is indicated by the CC bit in the FLAG in the CCW.
In FIG. 5, data associated with write-1 is represented as (X), data associated with read is represented as (Y), and data associated with write-2 is represented as (Z).

SSCH (1) issued from the program
Is received by RIOPCCH11, RIOPCCH
11 transfers the C section including the CCW information to the RIO PRCH 31 (including the CCW regarding the write-1 / read / write-2). At that time, RIOPCCH11
Since it is necessary to perform data transfer from I to IO5, the C section is transferred with A bit = 1. And RIO
The PCCH 11 transfers the transfer of the WD part after the transfer of the C part by R
Perform on IOPRCH31. The data of (X) and (Z) are included in the WD portion. In this example, since all the data transfer is completed by one transfer of the WD section, the A bit becomes 0 and the E bit becomes 1 at the time of the transfer of the WD section.

RIO PRCH31 is a part of these C, WD
When the copy is completed, it notifies the RCHP 33 of this.
Upon receiving the notification, the RCHP 33 converts the value of the data address in the CCW included in the C section according to the RMS 32. This is because the data address of the sent CCW is an address for the HMS 12 in the host 1 and needs to be converted for the RMS 32. When the conversion is completed, the RCHP 33 notifies the CNC 34. When the CNC 34 receives the notification, the CNC 3
4 performs input / output operation with IO5 based on the C section sent from RIOPCCH11.

First, the command regarding the light-1
Is sent to IO5, and then the data (X) related to write-1 is sent. When the transfer of the data ends, the information (STS) indicating the end regarding the write-1 is sent from the IO5. When this is sent, read processing of the next command of write-1 is performed. IO of lead command
5, data corresponding to the read (Y) is sent from the IO 5 to the CNC 34. When this is completed, information (STS) indicating the end related to the read is sent from the IO5 to the CNC 34. Then, upon receiving the STS, the CNC 34 sends a Comman for the LIGHT-2
Send d to IO5. After that, the data (Z) regarding the write-2 is sent to IO5. When the IO 5 receives it, the information (STS) indicating the end is sent to the CNC 34. The operation of these CNC34 and IO5 is SA22-7.
202-01 "ESCONI / O Interfac
e ”is followed.

The CNC 34 sends the data (X) and (Z) transferred from the RIOPCCH 11 and stored in the RMS 32 to the IO 5, and stores the data (Y) received from the IO 5 in the RMS 33. I
When the operation with O5 is completed, the CNC 34 will return to RCHP3.
3 is notified of completion. When the RCHP 32 receives the completion notification, the RCHP 32 notifies the RIOPRCH 31. The RIO PRCH 31 that received the notification is the CNC 34
Receives the data (Y) received from IO5 by RIOPCCH1
1 as a WD section (WD of RIO PRCH31
To transfer the department information). After that, the R part related to SSCH (1) is added to RIOPCCH11 by RIOPRCH31.
To send. When the RIOPCCH 11 receives the R part, based on the CCW for SSCH (1), the RIOP
The data (Y) sent from RCH31 is sent to HMS12.
Store in. After that, an input / output interrupt (I / O interrupt) is issued to the program, and a completion report regarding SSCH (1) is issued.

RIOPRCH11 is RIOPRCH31
C part related to SSCH (1) to RIO
The second input / output command (SSCH) from the program until the R part is sent back from RCH31 to RIOPRCH11.
Even if (2)) is issued, the operation of this embodiment is possible, and the transfer of the C section regarding SSCH (2) is performed on the BUF # 1 side (the exchange regarding SSCH (1) is BUF # 0.
Is being executed).

Since the round trip of the frame between the host 1 and the RIOP 3 is only one round trip, the transmission delay is minimized even if the connection distance between the host 1 and the RIOP 3 becomes long. For example, in the case of 5 ns / m, if the connection distance is 200 km, a transmission delay of 2 ms occurs in a round trip. Further, according to the present invention, since two input / output instructions can be simultaneously executed by a set of two buffers, the processing can be multiplexed and the apparent transmission delay is halved. Although not mentioned in the present embodiment, the apparent delay time can be further reduced by further expanding the number of buffer planes. It will be possible if a large-capacity memory can be mounted on one chip by integrating the memory in the future. If this is expanded to about 100 planes, the present invention can withstand full-time backup of data.

[0030]

The transmission delay can be minimized even if the connection distance between the host processor and the input / output device becomes long, and the input / output operation can be performed with almost no deterioration in performance. Further, it is possible to back up data in real time due to a disaster or the like.

[Brief description of drawings]

FIG. 1 is a diagram showing a schematic configuration of a system of an embodiment.

FIG. 2 is a diagram showing the structure of a channel command word (CCW).

FIG. 3 is a diagram showing structures of RIOPCCH and RIOPRCH.

FIG. 4 is a diagram showing a structure of a frame.

FIG. 5 is a diagram for explaining an example of a processing sequence of the system.

FIG. 6 is a view of the main storage (HMS) in the host
It is a figure which shows a mat.

FIG. 7: Main storage for remote input / output device (RMS)
It is a figure which shows the format of.

[Explanation of symbols]

1 Host Processor (Host) 11 Remote I / O Processor Center Channel (RIOP
CCH) 111 C section buffer 112 WD section buffer 113 RD section buffer 114 R section buffer 12 Main memory (HMS) 13 Channel processing unit (CHP) 14 Basic processing unit (BPU) 2 1 Gbps optical fiber 3 Remote input / output processing unit ( RIOP 31 Remote channel for remote I / O processor (RIO)
PRCH) 311 C section buffer 312 RD section buffer 313 WD section buffer 314 R section buffer 32 Main memory (RMS) 33 Channel processing device (RCHP) 34 Channel device (CNC) 4 200 Mbps optical fiber 5 Input / output device (IO)

Claims (1)

[Claims] 1. An input / output processing system in a data communication system, comprising: a host processor and a storage subsystem; and a remote input / output processing device connected between the host processor and the storage subsystem. In response to one input / output command instructed by a program, the host processor stores all communication control data relating to the command and all communication data sent from the host processor to the storage subsystem. Means for collectively transmitting the data to the remote input / output processing device, and receiving from the remote input / output processing device all communication data read from the storage subsystem based on the communication control data. Is written in an area designated by a program, the remote input / output processing device is provided with an input / output command transmitted from the host processor. Means for receiving and holding all the communication control data and all communication data, means for transmitting all the communication data to the storage subsystem based on the communication control data, and the communication control data An input / output processing system comprising means for receiving communication data from the storage subsystem and transmitting the communication data to the host processor based on the above.