WO2001050283A1

WO2001050283A1 - Information processing apparatus and method of information distribution

Info

Publication number: WO2001050283A1
Application number: PCT/JP2000/000040
Authority: WO
Inventors: Yuichiro Morita; Hidemitsu Naya; Takanori Yokoyama; Hideki Osonoi; Takao Yanai
Original assignee: Hitachi, Ltd.
Priority date: 2000-01-07
Filing date: 2000-01-07
Publication date: 2001-07-12
Also published as: JPWO2001050283A1

Abstract

A system reduces the process load in information distribution from a particular processor to a plurality of processors. One or more storage devices are connected with a plurality of processors through a network. The processor of the information distributor stores information in the storage device. The storage device detects the information and notifies the processors of the destinations. The processors of the destinations read distributed information from the memory in accordance with the notification from the storage device.

Description

Specification

Information processing system and information distribution method

The present invention relates to information distribution in a distributed system including a plurality of processing devices, and particularly to a method for distributing the same information from one processing device to a plurality of processing devices. Background art

As a new infrastructure for distributed systems, SAN (Storage Area Network) is beginning to be realized (Nikkei Electronics PP.57-63 1998.5 48 No715). The SAN is a network configuration in which a plurality of processing units and a plurality of disk units are connected. By connecting a disk unit that has been individually connected to each processing unit to the plurality of processing units, a plurality of processing units can be connected. Will be able to physically share the disk device. A typical medium for achieving SAN is Fiber Channel. Fiber Channel realizes high-speed serial transmission using optical fiber or copper wire, and is capable of one-to-one connection, single-connection using Fabric, and loop-connection using FC-AL (Fibre Channel Arbitrated Loop). Also, the SCI-3 protocol, which is the main protocol for peripheral devices such as disk devices, can be implemented on top of the Fiber Channel protocol (Prior Art 1).

In Japanese Patent Application Laid-Open No. 6-348657, a shared disk shared between two host computers is provided, and data is exchanged through writing and reading to and from the shared disk device. A technique for shortening the time required for transferring a large amount of data is disclosed (Prior Art 2). However, the above prior art 1 does not describe in detail the exchange of data between the processing devices. Further, the prior art 2 described above has a groove formed by a shared disk, and does not specifically describe a mode for realizing a data exchange between three or more host computers. Disclosure of the invention

An object of the present invention is to provide a method and system for efficiently realizing information distribution between a plurality of processing devices. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a diagram showing a system configuration according to the present invention.

FIG. 2 is a diagram showing a configuration of a processing apparatus according to the present invention.

FIG. 3 is a diagram showing a data write operation to a storage device by the processing device.

FIG. 4 is a diagram showing a command descriptor block of a data write.

FIG. 5 is a diagram showing a data read operation to the storage device by the processing device.

FIG. 6 is a diagram showing a command / descriptor / block of a data read.

FIG. 7 is a diagram showing a configuration of a storage device.

FIG. 8 is a diagram showing a data write Z-read operation by the storage device. <FIG. 9 is a diagram showing an operation of message transmission by the storage device. FIG. 10 is a diagram showing a command 'descriptor block' of the SEND command. FIG. 11 is a diagram showing an Asynchronous Event Notification message.

FIG. 12 is a diagram showing a storage format of a table.

FIG. 13 is a diagram showing a flow of the overnight delivery via the storage device. FIG. 14 is a diagram showing a configuration of the switch 3.

FIG. 15 is a diagram showing a basic example of the SCS I-3 protocol. FIG. 16 is a diagram showing an outline of the overnight transfer by Fiber Channel. FIG. 17 is a diagram showing a second configuration of the storage device.

FIG. 18 is a diagram showing a second storage format of the table.

FIG. 19 is a diagram showing a second flow of the overnight delivery via the storage device.

FIG. 20 is a diagram showing a second configuration of Fabric.

FIG. 21 is a diagram showing a third storage format of the table.

FIG. 22 is a diagram showing a third flow of data distribution via the storage device.

FIG. 23 is a diagram showing a third configuration of Fabric. BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, examples of the present invention will be described.

1 a, lb,..., 1 n are processing units for executing desired processing, 2 a, 2 b,..., 2 n are storage units for storing data and programs, and 3 are a plurality of processing units And a switch for connecting a plurality of storage devices. Each processing device can access an arbitrary storage device via the switch 3, and a plurality of processing devices can physically share one or more storage devices. In this system, when the same data is distributed from one processing device to one or more other processing devices, the source processing device switches instead of transmitting the distribution data to each of the destination processing devices. The distribution data is written to one of the storage devices via the switch 3, and the storage device in which the distribution data has been written notifies the destination processing device via switch 3 of the writing of the distribution data, and receives this notification. Each of the processing devices that read the data reads the distribution data from the storage device of the notification source. In addition, the same method is used when the same program is distributed from one processing device to one or more other processing devices.

FIG. 2 is a diagram showing a configuration of a processing apparatus 1a according to the present invention. The other processing devices have the same configuration.

11 1 is a microprocessor unit (MPU), 12 is main memory, 15 is a system bus, 13 is a bridge connecting MPU 11 with main memory 12 and system bus 15, and 14 is an MPU 11 is an interrupt controller (INTC) for controlling the interrupt to 1, and 16 is an adapter for connecting the processing unit 1 a to the switch 3.

The configuration of the adapter 16 will be described with reference to FIG. 2. 16 1 is a receiving circuit (RX) that receives a serial signal from the switch 3 to the processing unit 1a, and 16 2 is a receiving circuit (RX) from the processing unit 1a. A transmission circuit (TX) that transmits a serial signal to the TX, 163 is a serial / para conversion circuit that converts the serial signal received by the RX 161 into a parallel signal, and 164 is a parallel output to switch 3. A para-serial conversion circuit for converting signals to serial signals and passing them to に 16 2, 16 5 is an FC protocol processing circuit that processes Fiber Channel protocols, and 16 6 is a SCSI 13 protocol 167 is a buffer memory that temporarily stores transmission data and received data, and 168 is a main memory 20 and a buffer memory. A DMA controller (DMAC) that controls dynamic memory access (DMA) transfer to and from the memory 167. In this embodiment, the switch 3 is configured with Fiber Channel, which is the main network medium of SAN, and the communication protocol between the processing device and the storage device is SCSI-3, which is the main protocol of the peripheral device. Using the RX 161, TX 162, serial Z-para conversion circuit 163, para-Z serial conversion circuit 164, FC protocol processing circuit 165, SCSI protocol processing circuit 166 ANSI X3.230 (Fibre Channel Enhanced Physical and Signaling Interface), ANSI X3.269 (Fibre Channel Protocol), ANSI X3T10 / 995D (SCSI-3 Primary Commands), ANSI X3.270 (Fibre Channel Enhanced Physical and Signaling Interface) It is realized based on SCSI-3 Architecture Model).

A basic example of the SCSI-3 protocol will be described with reference to FIG. Assuming that an access source is an initiator and an access destination is an overnight get, first, in a command phase (1500), the initiator sends a command 'descriptor' block (CD B) to the target. The CDB specifies the contents of a command (write, read, etc.), the logical block area to be accessed on the hard disk, etc., and its format is defined by the SCS I-3 specification.

Next, if the specified command is a write access, write data is transmitted from the initiator to the target in a data out phase (1501), and if the specified command is a read access, Then, the data is sent from the target to the initiator in the data infuse (1502).

The target then indicates in the status phase (1503) that the command was executed successfully or that an error occurred during execution. The status is sent to the initiator, and finally the end of command execution is notified to the initiator (1504).

Next, an outline of the overnight transmission using Fiber Channel will be described with reference to FIG.

Fiber Channel basically transfers data in frames of the format shown in 1603. The frame 1603 is composed of an SOF (1604) indicating the start of the frame, a header (1665), a data field for storing transfer data (166), and an error check. CRC for

(166) and E〇F (166) indicating the end of the frame.

The header (1605) is composed of the frame type (1609), various control units (1610), the header count (1611), and the source ID.

(1 6 1 2), destination ID (1 6 1 3), and sequence count

(16 14) and a frame length (16 15) for setting the frame length.

The sequence (1602) is composed of one or more frames, and one sequence is one phase of SCSI-3 (command phase, data out phase, data phase, status phase, etc.) One of them).

The exchange (1600) is composed of one or more sequences, and realizes access to complete peripheral devices such as data write data in the SCS I-3.

The details of the overnight transmission using Fiber Channel are described in ANSI X3.230 (Fibre Channel Enhanced Physical and Signaling Interface). The data write operation to the storage device by the processing device 1a will be described with reference to FIG. 3 and FIG.

First, the MPU 11 passes a command descriptor block (CDB) set as shown in FIG. 4 to the SCSI protocol processing circuit 166 to instruct transmission to the storage device at the write destination (3 0 1 0). The CD B in Fig. 4 is defined by the SCSI-3 specification. The operation code indicates the type of command such as write or read (OAh for a write), and L UN indicates the logic of the target. Logical block address indicates the start logical block number of the logical block group to be written, and Transfer length indicates the length of the logical block group to be written or read. And control indicate various control bits.

Next, the SCSI protocol processing circuit 166 passes the CD B and the transmission destination ID to the FC protocol processing circuit 165 to instruct transmission, and the FC protocol processing circuit 166 stores the CD B in the data portion. Is formed and output to the switch 3 via the para-Z-to-serial conversion circuit 16 4 (step 302).

Next, the MPU 11 sets the DMAC 168 to transmit the write data stored in the main memory 12 so that the DMAC supplies the write data to the adapter 16 via the DMA (3). 0 3 0).

Next, the MPU 11 instructs the SCSI protocol processing circuit 166 to transmit the write data to the storage device to which the data is to be written, and the SCSI protocol processing circuit 166 activates the DMA to activate the main memory 1 While extracting the data from buffer 2 via buffer memory 1667, pass the data and the transmission destination ID to FC protocol processing circuit 1665 to instruct data transmission, and perform FC protocol processing. The circuit 165 forms a frame in which the write data is stored in the data section and The signal is output to the switch 3 via the LZ serial conversion circuit 1 64 and the TX 62 (304 0).

Next, wait for the status to be returned from the write destination storage device.

(3500). If the status is returned, check its contents to see if the command execution was completed successfully (3060).

Finally, it waits for a command execution completion message to be sent from the write destination storage device (3700), and completes the command upon receipt of the message.

Next, a data read operation from the storage device by the processing device 1a will be described with reference to FIGS. 5 and 6. FIG.

First, the MPU 11 transfers the command, disk rib, and evening block (CDB) set as shown in FIG. 6 to the SCI protocol processing circuit 1666 (51010). # 08 in FIG. 6 is basically the same as CDB in FIG. 4, except that the Operation code is set to 08h indicating a read command.

Next, the MPU 11 sets the DMA C 168 to store the read data on the main memory 20 so that the read data is taken out of the adapter 16 by DMA (520 0).

Next, the MPU 11 instructs the SCSI protocol processing circuit 166 to transmit the CDB to the storage device of the read destination, and the SCSI protocol processing circuit 166 passes the CDB and the transmission destination ID to the FC protocol processing circuit 165. Transmission is instructed, and the FC protocol processing circuit 165 forms a frame in which the CDB is stored in the data section, and outputs the frame to the switch 3 via the para-Z serial conversion circuit 164 and the TX 162 (503030) ).

Next, the storage device of the read destination is transferred to the processing device 1a via the switch 3. When one byte is transmitted, the read data is sent to the SCSI protocol processing circuit 166 via the RX 161, serial Z-parameter conversion circuit 163, and FC protocol processing circuit 165. The protocol processing circuit 166 sends the received read data to the buffer memory 166, activates the DMAC 168, and stores the read data in the main memory by DMA (5040). Next, it waits for the status to be returned from the storage device at the read destination (5500). When the status is returned, check the contents to confirm whether the command execution was completed normally (5506).

Finally, it waits for a command execution completion message to be sent from the storage device to which the data is read (5700), and upon receipt of the message, completes the command.

FIG. 14 is a diagram showing a configuration of the switch 3.

Switch 3 of the present embodiment shows a Fabric that forms a Fiber Channel Fabric topology.

3 1 a, 31,..., 31 i, 31 j,..., 31 χ, and 31 y are F ports that are connected one-to-one with the processing device or the storage device, respectively. 32 is a Connection Sub-Fabric for realizing a line connection between arbitrary F ports, 33 is a Connectionless Sub-Fabric for realizing a non-line connection between any F ports, and 34 is a Fabric control unit.

The F port 3a transmits the serial signal received by the RX311, the RX311 that receives the serial signal from the processing unit or the storage device, the TX312 that transmits the serial signal to the processing unit or the storage device, and the It has a receiving buffer (BUF) 313 for temporarily saving and transmitting it to the Connection Sub-fabric 32 or Connectionless Sub-fabric 33. Other F ports have the same configuration. Connection Sub-Fabric 32, when a connection between two F ports is established, occupies a route between the two and executes frame transfer, and multiple frames are transferred at high speed between the two F ports. it can.

Connectionless Sub-Fabric 33 does not fix the connection between two F ports, but executes transfer between arbitrary F ports in frame units, and realizes multiplex transfer between multiple F ports.

The connection sub-fabric 32 and the connection sub-fabric 33 are selected by the processing device of the distribution source and the distribution destination. In other words, the processing device of the distribution destination sends the connection sub-fabric 32 and the connection sub-fabric 33 to the control unit 1610 of the header 1605 of the frame 1603 shown in FIG. Specify or use the shift. Switch 3 sends the designation of the distribution source to the processing device of the distribution destination, and the processing device of the distribution destination determines whether or not the connection sub-Fabric specified by the distribution source can be used, and if so, switches to that effect. Return to 3. When the distribution destination and the distribution source are matched, switch 3 distributes the information using the connection sub-Fabric of the matched one.

FIG. 7 is a diagram showing a configuration of a storage device 2a according to the present invention. The other storage devices have the same configuration.

Reference numeral 21 denotes an interface for connecting the storage device 2a to the switch 3, and reference numeral 22 denotes a magnetic disk.

The structure of the interface 21 will be described with reference to FIG. 7. The receiving circuit (RX) 211 receives a serial signal from the switch 3 to the storage device 2a. A transmission circuit (TX) that transmits a serial signal to 3, 2 13 is a serial-to-parallel conversion circuit for converting the serial signal received by RX 211 into a parallel signal, and 2 14 is an output to switch 3 A signal for converting a parallel signal to a serial signal and passing it to ΤX2 1 2 A re-conversion circuit, 215 is an FC protocol processing circuit that processes the Fiber Channel protocol, 216 is a SCSI protocol processing circuit that processes the SCSI-3 protocol, and 217 is a magnetic disk 22 Buffer memory that temporarily stores write data to the disk and read data from the magnetic disk 22.MPU 21 controls access to the magnetic disk 22 and communication by switch 3. Reference numeral 9 denotes a non-volatile memory (ROM) for storing a program for controlling access to the magnetic disk 22 and communication by the switch 3, and reference numeral 220 denotes a logical block area for storing distribution data, a distribution source processor ID, and a distribution destination. This table stores multiple combinations of processor IDs. In this embodiment, as explained in FIG. 2, the communication protocol between the processing device and the storage device is SCSI-3, switch 3 is Fiber Channel, and RX211, TX212, serial are used. The Z-parameter conversion circuit 2 13, the para-Z serial conversion circuit 2 14, the protocol processing circuit 2 15, and the command processing circuit 2 16 are the ANSI X3.230 (Fibre Channel Enhanced Physical and Signaling Interface), ANSI X3.269 (Fibre Channel Protocol), ANSI X3T10 / 995D (SCSI-3 Primary Commands), and ANSI X3.270 (SCSI-3 Architecture Model).

The overnight write Z read operation by the storage device 2a will be described with reference to FIG.

First, the command 'descriptor block (CDB) sent from the processing unit via switch 3 is sent to the SCSI protocol via RX 211, serial / parallel conversion circuit 21 3 and FC protocol processing circuit 2 15 The command is input to the protocol processing circuit 216 (81010), and the command to be executed is determined based on the command code of the CDB (8200). If the command code is a write, CDB The write data to be received thereafter is written to the logical block area of the magnetic disk 22 determined by the Logical block address t Transfer 1 length of the magnetic disk 22 via the buffer memory 217 (8030), and the co-referencing and code Is read, the data in the logical block area of the magnetic disk 22 determined by the Logical Block Lock Address and Transfer Length of the CDB is read out via the buffer memory 217 and the FC protocol processor 2 15. Transmit to the processing device of the access source via the para-Z serial conversion circuit 2 14, TX 2 1 2, and switch 3 (8040).

After receiving the write data overnight or transmitting the read data overnight, it transmits a status to the processing device of the access source indicating whether or not the command execution has been completed normally (8505).

Next, the completion of the command is notified to the access source processing device (8060). Finally, a message transmission for realizing data distribution is executed (8070). In the case of distributing data from one processing device to one or more other processing devices, in this embodiment, the distribution source processing device writes the distribution data to the storage device, and the distribution destination processing device stores the distribution data respectively. By reading from the device, overnight delivery is realized. Send message above

In this data distribution, (8070) notifies the processing device of the distribution destination that the distribution data has been written to the storage device, or notifies the processing device of the distribution source that the distribution data has been read. To use.

The storage device determines whether the data written from the processing device is distribution data, based on the logical block area. The storage device stores in a table a combination of a specific logical block area for data distribution, the ID of the processing unit of the distribution source that distributes data via the logical block area, and the ID of the processing unit of the distribution destination. ₂ 2 0 stores, Akuse of CDB sent from processor By comparing the destination logical block area with the table 220, it can be determined whether or not the access is data distribution, and if the access is data distribution, the distribution source and distribution destination. As shown in Fig. 12, the storage format of the table 220 is as follows: the logical block number and the file, the ID of the processing unit of the distribution source, the number of distribution destination nodes, and the ID of each processing unit of the distribution destination are An entry shall be specified, and invalid digits in the binary notation of the logical block number shall be specified from the least significant digit to any digit in the field to indicate the logical block area for data distribution. For example, if the lower 8 digits are invalidated by a filter, an area of 2 ⁸ = 2 56 logical blocks will be specified.

The message transmission operation will be described with reference to FIGS. 9, 10, and 11. FIG.

This message transmission is specified in the specification of SCS I-3

This is realized by the Asynchronous Event Notification (AEN) mechanism. AEN is a function for notifying an event that has occurred asynchronously in a processing device or a storage device from the target side to the overnight initiator. AEN transfers the message shown in Fig. 11 by the CDB of the SEND command shown in Fig. 10 (however, the AEN bit is set to 1). In the information of the message, the head number and the block length of the logical block area in which the distribution data is written by the processing device of the distribution source are set (9101). The start number and block length of this logical block area are the same as the logical block number and block of the CDB of the data light received from the processing device of the distribution source.

Next, a CDB for the AEN is set (9101), and the CDB and the message are transmitted to one of the processing devices of the distribution destination with reference to the table 220 (90030). , 9 040).

Next, wait for the status reply from the message recipient (9500), When the status is received, it checks whether or not the message transmission has been completed normally (9006), and waits for a command end notification from the message transmission destination (9007). When the command end is received, the process from sending the CDB (9303) to waiting for the command end (9007) is repeated until the message is sent to all the destination processing devices.

FIG. 13 is a diagram showing a flow of data distribution via the storage device. First, the distribution source processing device transmits a CDB indicating the write of the distribution data to the storage device (1300), and then transmits the distribution data (1301). After executing the write command, the storage device returns a status and a command end notification to the source processing device (1302, 1303), and sends a CDB indicating AEN to the destination processing device. It transmits (1304), and then transmits a message indicating the logical block storing the distribution data (1305). After receiving the message, the delivery destination processing device returns a status and a command end notification to the storage device (1306, 1307), and then refers to the message and transmits the delivery data to the storage device. A CDB indicating the command is transmitted (13008).

The storage device executes the data read command and sends the delivery data to the destination processing device (1309), and then returns a status and a command end notification to the source processing device (13310,1). 3 1 1), sends a message 08 indicating “8” to the distribution source processing device (1 3 1 2), and then transmits a message indicating the logical block storing the distribution data (1 3 1) 3)

After receiving the message, the distribution source processing device returns a status and a command end notification to the storage device (1314, 1315), and then refers to the message to determine the corresponding data distribution. Judgment is completed, and it is detected that distribution data can be updated. In the above embodiment, when the same data is distributed from a certain processing device to one or more other processing devices, the distribution data is written to any of the storage devices via the network, and the distribution data is written. The received storage device notifies the processing device of the distribution destination via the network that the writing of the distribution data has been completed, and the processing devices that have received this notification read the distribution data from the storage device of the notification source. The source processing device does not need to confirm the preparation of the destination processing device or notify the completion of writing of the distribution data to the destination processing device.This eliminates the need for processes such as data processing overhead. .

In the above embodiment, a mechanism is provided for notifying the storage device of writing or reading of the distribution data. However, as another second embodiment, the processing device notifies the writing or reading of the distribution data so that the storage device is notified. A system that eliminates the need for a special mechanism of the device will be described. The system configuration of this embodiment is the same as that of FIG. 1, and the processing device and network of this embodiment are the same as those of FIGS. 2 and 14, respectively. On the other hand, the storage device of this embodiment is basically the same as that of FIG. 7, but there is no need to notify the writing or reading of the distribution data. Therefore, as shown in FIG. The configuration excludes 0.

The processing unit of the distribution source stores the combination of the logical block area of the storage device to which various types of distribution data are written and the ID of the processing device of the distribution destination in the main memory 12 and writes the distribution data to the storage device. This is referred to when notifying the destination processing device of data writing. As shown in Fig. 18, the storage format of the combination of the logical block area of the storage device to which various types of distribution data is written and the processing device ID of the distribution destination is a distribution data ID that identifies the distribution data. And the logical block number and logical block indicating the beginning of the logical block area of the storage device to which the distribution data is to be written, the number of distribution destination nodes, and the ID of each processing device at the distribution destination Is one entry.

FIG. 19 is a diagram showing a front end of the overnight delivery via the storage device in the second embodiment.

First, the distribution source processing device transmits a CDB indicating the write of the distribution data to the storage device (1900), and then transmits the distribution data (1901). After executing the data write command, the storage device returns a status and a command end notification to the distribution source processing device (1902, 1903). Next, the distribution source processing device transmits a CDB indicating the AEN to the distribution destination processing device (1904), and then transmits a message indicating the logical block storing the distribution data (1905). .

After receiving the message, the distribution destination processing device returns a status and a command end notification to the distribution source processing device (1906, 1907), and then refers to the message and transmits the distribution data to the storage device. A CDB indicating a read is transmitted (1908).

The storage device executes the data read command and transmits the distribution data to the distribution destination processing device (1909), and then returns a status and a command end notification to the distribution source processing device (1910, 199). 1 9 1 1).

Next, the distribution destination processing device transmits a CDB indicating the AEN to the distribution source processing device (1912), and then transmits a message indicating the logical block storing the distribution data (1913). ).

After receiving the message, the distribution source processing device returns a status and a command end notification to the distribution destination processing device (1914, 1915), and then completes the data distribution by referring to the message. And that the distribution data can be updated is detected.

In the above embodiment, the processing device notifies the writing and reading of the distribution data. By doing so, there is no need to provide a storage device with a mechanism for notifying the writing and reading of the distribution data, and the distribution can be realized by the existing storage device.

In the above embodiment, the storage device or the processing device notifies the writing or reading of the distribution data to or from the storage device. However, as another third embodiment, Fabric, which is the core of the network, transmits the data to the storage device. A system for notifying the writing and reading of the distribution data to eliminate the need for a special mechanism of the storage device and to reduce the overhead of the data distribution processing in the processing device will be described. The system configuration of this embodiment is the same as that of FIG. 1, and the processing device and the storage device of this embodiment are the same as those of FIGS. 2 and 17, respectively. On the other hand, the Fabric of the present embodiment is basically the same as FIG. 14. As shown in FIG. 20, a table 35 to be referred when notifying the writing and reading of the distribution data to the storage device is added. The configuration is as follows.

Fabric 3 in FIG. 20 stores a combination of a storage device ID and a logical block area to which distribution data is to be written, a processing device ID of a distribution source, and a processing device ID of a distribution destination in a table 35. Refer to when relaying distribution data writing from the original processing device to the storage device. As shown in FIG. 21, the storage format of the table 35 is a format in which a storage device ID for writing distribution data is added to the storage format of FIG.

FIG. 22 is a diagram showing a flow of data distribution via a storage device in the second embodiment.

First, the distribution source processing device transmits a CDB indicating the write of the distribution data to the storage device (2200), and then transmits the distribution data (2201). After executing the data write command, the storage device returns a status and a command end notification to the distribution source processing device (2202, 2203). Fabric compares the destination storage device ID and logical block of CDB2200 with Table 35, sends a CDB indicating AEN to the corresponding destination processing device (2204), and then sends the delivery data. Then, a message indicating the storage device ID and the logical block in which is stored is transmitted (225).

After receiving the message, the delivery destination processing device returns a status and a command end notification to Fabric (220, 2207), and then refers to the message and sends it to the storage device. A CDB indicating a read is transmitted (2208).

The storage device executes the data read command, transmits the distribution data to the destination processing device (2209), and returns a status and a command end notification to the distribution processing device (2210, 2). 2 1 1).

Fabric checks the destination storage device ID and logical block of CDB 228 against Table 35, sends a CDB indicating AEN to the corresponding source processing device (2 2 1 2), and then distributes A message indicating the logical block storing the data is transmitted (2 2 1 3).

After receiving the message, the distribution source processing device returns a status and a command end notification to Fabric (2 214, 221 5), and then completes the corresponding data distribution by referring to the message. And that the distribution data can be updated is detected.

In the above embodiment, the Fabric notifies the storage device of the writing or reading of the distribution data, thereby eliminating the need for a special mechanism of the storage device and preventing the overhead of the data distribution processing in the processing device. Can be reduced.

In each of the above embodiments, a plurality of processing devices at the delivery destination read out the delivery data written in one storage device. However, in order to distribute the access to the storage device, the delivery destination Dividing the processing units into multiple groups, The distribution source processing device can write the same distribution data to a plurality of storage devices, and each distribution destination processing device can read the distribution data from a different storage device for each group.

In addition, there may be a form in which a cache for temporarily storing distribution data read from the storage device is provided on the network to speed up the reading of the distribution data from the storage device.

As shown in FIG. 23, each F port 31a, 31b,..., 31i, 31j,. When the distribution data is read from a storage device with a plurality of processing devices, the data read by the first processing device from the storage device is cached at the F port to which the storage device is connected. By storing the data in the cache 314 and reading the distribution data from the cache 314 instead of reading the distribution data from the storage device, the other processing devices can read the distribution data at high speed.

According to the present invention, when the same data is distributed from a certain processing device to one or more other processing devices, the distribution data is written to any storage device via a network, and the distribution data is written. The inserted storage device notifies the processing device of the distribution destination via the network of the writing of the distribution data via the network, and the processing devices that have received this notification read the distribution data from the storage device of the notification source. The processing device does not need to confirm the preparation of the destination processing device or notify the completion of writing of the distribution data to the destination processing device.This eliminates the overhead of data distribution processing. There is. In another form, the processing device notifies the writing and reading of the distribution data, so that there is no need to provide a mechanism for notifying the writing and reading of the distribution data to the storage device. Realize distribution There is an effect that can be.

In another form, the Fabric notifies the storage device of the writing and reading of the distribution data to the storage device, thereby eliminating the need for a special mechanism of the storage device and preventing the data distribution processing in the processing device from being overtaken. This has the effect of reducing the cost.

Further, the distribution destination processing device is divided into a plurality of groups, the distribution source processing device writes the same distribution data to the plurality of storage devices, and each distribution destination processing device transmits distribution data from a different storage device for each group. By reading, there is an effect that access to the storage device can be distributed.

Further, a cache for temporarily storing the distribution data read from the storage device is provided in the network, and the processing device which first accesses the storage device stores the distribution data read from the storage device in the cache, and the other processing devices store the distribution data. By reading the distribution data from the cache instead of reading the distribution data from the cache, there is an effect that the reading of the distribution data can be speeded up. Industrial applicability

It can be used in the fields of industrial equipment, electric power, railways, automobiles, etc. to which a distributed system consisting of multiple processing units can be applied.

Claims

Range of request

1. Multiple processing units,

A storage unit for storing the information, a table for storing the storage area of the writing destination of the distributed information, the address of the processing device of the transmission source, and the address of the processing device of the distribution destination in association with each other; In response to the writing of the information sent from the processing device, the storage device in which the distribution information is written is notified to the processing device of the distribution destination with reference to the table, at least one storage device, An information processing system comprising: a processing device; and connection means for connecting the storage device.

2. In the information processing system of claim 1,

The information processing system, wherein the storage device refers to the table in response to reading of the stored information by the distribution destination processing device, and notifies the distribution source processing device that the information has been read.

3. The storage device,

It has a table that stores the type of information to be distributed, the address of its own processing device, and the address of the processing device of the distribution destination, and writes information to the storage device and refers to the table to determine the distribution destination. A plurality of processing devices for notifying a processing device of a type of information and a storage area of the storage device in which the information is written;

An information processing system comprising: a connection unit that connects the storage device and the processing device.

4. In the information processing system according to claim 3,

An information processing system, wherein, in response to reading of the information stored in the storage device, the processing device refers to the table and notifies the distribution source processing device that the information has been read.

5. Multiple processing units,

At least one storage device,

The storage device is connected to the processing device, and the storage device and the storage area to which the information to be distributed is written are associated with the address of the processing device of the distribution source and the address of the processing device of the distribution destination. When information is written from the processing device to the storage device, the processing device at the distribution destination is determined with reference to the table, and the distribution information is written to each of the processing devices. An information processing system comprising: the storage device; and connection means for notifying the storage area.

6. The information processing system according to claim 5,

An information processing system, wherein the connection unit refers to the table in response to reading of the stored information by the processing device of the distribution destination and notifies the processing device of the distribution source that the information has been read.

7. In the information processing system according to claim 5 or 6,

The connection means has a cache means for temporarily storing information read from the storage device by the processing device, and stores information read by one of the processing devices to which information is distributed in the cache means. An information processing system, wherein the other processing device at the distribution destination reads information from the cache unit.

8. In a system including a plurality of processing devices, at least one storage device, and a plurality of the processing devices and a connection unit that connects the storage devices, information is distributed from one processing device to another processing device. The information processing device of the information distribution source writes the information to the storage device, the storage device notifies the information distribution destination processing device that the information has been written, and the information distribution destination processing device Each of the storage devices is based on a notification from the storage device. Information distribution method for reading information from the device.

9. A method for distributing information from one processing device to at least one other processing device in a system having a plurality of processing devices and at least one storage device,

The distribution source processing device writes the information to the storage device and notifies the information distribution destination processing device that the information has been written to the storage device, and the information distribution destination processing device receives the information from the processing device. Reading information from the storage device based on the notification of the information.

10. In a system including a plurality of processing devices, at least one storage device, and a plurality of the processing devices and a connection unit that connects the storage devices, one system includes at least one other processing device. Method of distributing information to

The distribution source processing device writes the information to the storage device via the connection unit, and the connection unit notifies the distribution destination processing device that the information has been written, and the distribution destination processing device An information distribution method, wherein information is read from the storage device based on a notification from the connection unit.

11. A method for distributing the same information from one processing device to another plurality of processing devices in a system having a plurality of processing devices and a plurality of storage devices,

An information distribution method, wherein the distribution source processing device writes the same information to the plurality of storage devices, and each of the information distribution destination processing devices reads information from any one of the plurality of storage devices. .