JP2008140415A - Data duplication system, data transmission/reception method, and program for duplicating data in storage - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To complete quickly data transmission/reception, even if data are discarded during transmission. <P>SOLUTION: In a storage 20 which is a transmission source, at least one piece of redundant data for error correction are generated from transmitted original data. The storage 20 transmits the original data and the redundant data by different transmission unit. The storage 20 is equipped with a redundancy part for duplicating data transmitted to another storage 21, and a communication part for transmitting the original data and the redundant data individually to the storage 21. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a data replication system, a data transmission / reception method, and a program for replicating data in a storage.

  In order to maintain the functions of a computer system even when a disaster or the like occurs, a computer system provided with a normal (active) system and a standby system has been realized. For example, EMC Corporation implements a system that performs mirroring using normal storage and standby storage. Information on this system is disclosed at the URL "http://www.emc2.co.jp/local/ja/JP/products/product_pdfs/srdf/srdf.pdf".

  Japanese Patent Laid-Open No. 2000-305856 describes a system for duplicating data between a main center (normal system) and a remote center (standby system).

  In general, in a normal system, a normal storage and a normal host computer (hereinafter referred to as a host) using the storage are connected. The same applies to the standby system. The normal storage and the standby storage are connected via a communication network such as a dedicated line or the Internet.

  In the system disclosed in “http://www.emc2.co.jp/local/en/JP/products/product_pdfs/srdf/srdf.pdf” or the system described in Japanese Patent Laid-Open No. 2000-305856, Data is sent directly from the normal system to the standby system. However, data may be transmitted from the normal system to the relay apparatus, and data may be transmitted from the relay apparatus to the standby system. In general, the relay device transmits data received from the normal system to the standby system, and when receiving a notification of completion of reception from the standby system, the normal system transfers the data to the standby system. Notifying completion. When the normal system transmits data to the relay device, the normal system starts a next process after receiving a notification from the relay device that the data transfer is completed.

  In general, when data is transmitted / received via a communication network, the data is divided and transmitted / received for each data transmission unit. This data transmission unit is different for each communication protocol. Here, a case where a packet is used as a data transmission unit will be described as an example. A data transmission unit such as a packet includes not only data to be transmitted but also an error detection code for detecting an error (data corruption) in the transmission process. Examples of error detection codes include checksum data and CRC (Cyclic Redundancy Check) data. When the device receiving the data detects an error in the data using the error detection code, the device discards the packet.

  Even if no error has occurred in the data, if the communication network is congested (a high communication load), the packet is discarded on the communication network. If the packet is discarded and no response is received from the transmission destination, the transmission source transmits the packet again. If packet discarding is started when congestion occurs in the communication network, the communication load may be further increased or the utilization rate of the communication network may be decreased due to retransmission by the transmission source. In order to avoid such a problem, recently, a method has been adopted in which devices constituting a communication network arbitrarily discard packets when the communication load exceeds a predetermined threshold. This method is called RED (Random Early Detection).

  There has also been proposed a data transmission method in which a device that receives data can correct an error in the data. For example, in Japanese Patent Application Laid-Open No. 57-138237, a device that transmits data to be transmitted and a parity bit created from the data separately and receives the data to be transmitted and the parity bit performs error correction. The data transmission system to be performed is described.

  In the following, mirroring and backup are distinguished as follows. “Mirroring” is defined as writing the same data to two or more storages including the storage when a write command (write command) is output from the host to a certain storage. Replication is also treated as a kind of mirroring. In addition, “backup” is defined as copying the contents of a certain storage to another storage at an arbitrary timing without being triggered by a write command for the storage.

  In addition, in a computer system, a technique has been realized in which a running program is interrupted at an arbitrary time point and restarted later. For example, in a super computer manufactured by NEC Corporation sold under the name SX series, the execution state of a process (for example, the state of a memory or a register) at an arbitrary time is saved, and the program is suspended / resumed. I am doing so. This function is sometimes called a checkpoint / restart function.

  When data is transmitted from a normal system to a standby system via a relay device, there are the following problems. The relay apparatus transmits the data received from the normal system to the standby system, and when the reception completion notification is received from the standby system, notifies the normal system that the transfer has been completed. The normal system does not proceed to the next processing unless receiving a notification that the transfer is completed from the relay device. Therefore, there is a problem that it takes time until the next system can start after the normal system transmits data to the relay apparatus. In particular, when a relay device or a standby system is provided as a disaster countermeasure, the relay device is arranged at a remote location of the normal system, and the standby system is arranged at a far place. As a result, it takes time to send and receive data and notifications between the normal system and the relay device, and time to send and receive data and notifications between the relay device and the standby system. The start of the process will be delayed.

  For example, the round trip time (round trip time from sending data to getting a response) in communication with a point 100 km away is on the order of ms (milliseconds). The host advances each process on the order of μs (microseconds). Therefore, the time until the response to the transmission is obtained causes a processing delay.

  In the data transmission process, if a packet or other data transmission unit is discarded, the normal storage serving as a transmission source retransmits the data. Then, it takes more time to complete data transmission, and the start of the next process is further delayed.

  In addition, when data backup processing is performed from a normal storage to a standby storage, in addition to a long data transmission distance, the amount of data to be transmitted increases, which further increases the transmission time. .

  Further, when an abnormality occurs in a normal system, the standby system cannot always start processing immediately. For example, when starting a certain process X, the writing of data A and data B to the storage must be completed. When a normal host outputs a write command for writing data A to the storage and performs mirroring, the data A is reflected in the standby storage. Then, it is assumed that an abnormality occurs in the normal system. In this case, since the data B is not written in the standby storage, the process X cannot be started immediately. When the processing is started in the standby system, it is necessary to write the data B to the standby storage and start the processing X, or to delete the data A and start again from the processing before the processing X. For this reason, it takes time to resume the processing in the standby system.

  In addition, there is a case where a function that can resume the processing if the data recording state is reached is not realized by the application program. For example, in the above example, there is a case where a function that can resume the process from the process X if the host writes A and B in the storage may not be realized. Hereinafter, a function capable of resuming processing when the storage is in a predetermined data recording state is referred to as a “resuming function”. Even when the resume function is not realized by the application program, it is preferable that the process can be quickly resumed in the standby system if an abnormality occurs in the normal system.

  Therefore, an object of the present invention is to allow data transmission / reception to be completed quickly even if data is discarded in the transmission process.

  In the data transmission / reception method for transmitting data to a transmission destination for receiving data from a transmission source for transmitting data according to the present invention, the transmission source generates at least one redundant data for error correction from the transmitted original data. The original data and the redundant data are transmitted in separate data transmission units.

  At the destination, before completing the reception of all data groups that are a set of original data and redundant data, an error is received when a part of the data group that can be partially subjected to error correction processing is received. It is preferable to execute a correction process. According to such a method, even if a part of data is discarded in the transmission process, the transmission source does not need to transmit the data again.

  For example, the transmission source divides the original data into divided data, and creates redundant data that can restore the original data even if one or more of the divided data is lost.

  For example, parity data or ECC (Error Correcting Code) may be used as redundant data.

  Further, duplicate data of transmission data may be used as redundant data.

  The original data and the redundant data may be sent to different communication networks. According to such a method, even if a failure or the like occurs in one communication network, it is possible to proceed with processing using data received from the other communication network.

  The data replication system according to the present invention is a data replication system in which data in the first storage is mirrored or backed up to the second storage via a communication network. The first storage is data for controlling data transfer. Transfer processing means, and redundancy means for creating at least one error correction data from the original data to be transmitted. The data transfer processing means includes the original data and the redundant data created by the redundancy means. It transmits by a separate data transmission unit, It is characterized by the above-mentioned.

  The second storage includes data restoration means for performing error correction processing using redundant data received from the first storage, and storage processing means for storing the data restored by the data restoration means in a storage medium. The means received a part of the data group that can partially execute the error correction processing on the original data before completing the reception of all the data group that is a set of the original data and the redundant data from the first storage. It is preferable to execute error correction processing in stages. According to such a configuration, even if a part of data is discarded in the transmission process, the transmission source does not need to transmit the data again.

  For example, the redundancy means in the first storage divides the original data into divided data, and creates redundant data that can restore the original data even if one or more of the divided data are lost.

  The redundancy means may use parity data or ECC as redundant data.

  The redundancy unit may create duplicate data of the original data as redundant data.

  The data transfer processing means may send the original data and the redundant data to different communication networks. According to such a configuration, even if a failure or the like occurs in one communication network, it is possible to proceed with processing using data received from the other communication network.

  Further, a program for duplicating data in a storage according to the present invention is provided in the first storage in a data duplication system that mirrors or backs up data in the first storage to the second storage via a communication network. The computer is configured to execute a process of creating at least one redundant data for error correction from the transmitted original data and a process of transmitting the original data and the redundant data in different data transmission units.

  According to the present invention, at least one redundant data for error correction is created from the transmitted original data, and the original data and the redundant data are transmitted in separate data transmission units. Therefore, in the standby system, the original data can be restored from a part of the set of the original data and the redundant data, and even if a part of the data is discarded in the transmission process, there is no need to transmit again. As a result, the data transfer can be completed quickly.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a first embodiment of a data replication system according to the present invention. In the data replication system shown in FIG. 1, a storage 11 is locally connected to a host (host computer) 10 that uses the storage 11. The storage 11 is connected to the relay device 15 via a communication network (hereinafter referred to as a network) 13 such as the Internet or a dedicated line. Further, the relay device 15 is connected to the storage 12 via the network 14.

  The storages 11 and 12 are, for example, a single magnetic disk device, an optical disk device, or a magneto-optical disk device. As the storages 11 and 12, a disk array device that is a single magnetic disk device, an optical disk device, or a set of magneto-optical disk devices can be used. The host 10 and the storage 11 are connected by SCSI, Fiber channel, Ethernet (registered trademark), or the like.

  FIG. 2 is a block diagram showing a configuration example of the storage 11 shown in FIG. The configuration of the storage 12 is also as shown in FIG. As shown in FIG. 2, the storage 11 includes a storage controller 100 and a storage medium 101 that is a storage body. The storage controller 100 includes a communication unit 102 that communicates with the host 10 and other storages or relay devices, a processing sequencer 103 that manages each processing sequence, an IO scheduler 104 that controls the order of processing instructions for the storage medium 101, and an IO scheduler. A medium processing unit 105 that controls the operation of the storage medium 101 in accordance with a processing command issued by the 104, and a buffer memory 106 that temporarily stores data from the host 10 or the like to the storage medium 101 and data from the storage medium 101 to the host 10 or the like. Including. The processing sequencer 103 is realized by a CPU that operates according to a program, for example.

  FIG. 2 shows a case where only one combination of the medium control unit 105 and the storage medium 101 is included. The storages 11 and 12 may include a plurality of recording media 101, and a medium control unit 105 may be provided corresponding to each storage medium. However, there is one IO scheduler 104 included in each storage 11, 12. When there are a plurality of storage media 101, the processing sequencer 103 designates a storage medium to be processed, and the IO scheduler 104 causes the media control unit 105 corresponding to the designated storage medium to perform processing.

  FIG. 3 is a block diagram illustrating a configuration example of the relay device 15 illustrated in FIG. 1. The relay device 15 includes a communication unit 150 that communicates with the storages 11 and 12, a relay processing unit 151 that performs sequence management of relay processing, and a buffer memory 152 that temporarily stores data received from the storages 11 and 12. The relay processing unit 151 is realized by a CPU that operates according to a program, for example.

  As shown in FIGS. 1 and 2, the storage 11 is provided with a communication unit 102, and the storage 11 itself transfers data mainly. That is, the host 10 does not read data from the storage 11 and transmits data to the relay device 15, but the storage 11 directly transmits data to the relay device 15. The storage 12 also receives data directly from the relay device 15. However, the host 10 instructs the timing at which the storage 11 transmits data to the relay device 15.

  In addition, the relay device 15 is installed outside the range where the influence of the disaster is expected to spread when a disaster such as an earthquake occurs at the installation location of the storages 11 and 12. That is, even if the storages 11 and 12 become inoperable due to a disaster, a position (distance from the storages 11 and 12) where the operation can be continued at that position is calculated in advance, and the relay device 15 is calculated at the calculated position. Is installed. For example, when the storage 11 is affected by an earthquake with a presumed seismic intensity (for example, seismic intensity 6-7), the relay device 15 is installed at a position where the relay device 15 has a seismic intensity that does not cause damage or the like. The Further, the relay device 15 is configured so that the data transfer time between the storage 11 or the storage 12 and the relay device 15 is shorter than the data transfer time when data is directly transferred between the storage 11 and the storage 12. Installed. Therefore, the data replication system functions as a disaster resistant data management system.

  Next, the operation of the data replication system will be described with reference to the flowcharts of FIGS. Here, as an example of a backup process, a storage 11 as a data transfer source transfers data to a storage 12 as a data transfer destination. FIG. 4 is a flowchart showing the operation of the processing sequencer 103 in the storage controller 100, and FIG. 5 is a flowchart showing the operation of the relay processing unit 151 in the relay device 15.

  The host 10 outputs a copy command to the communication unit 102 in the storage controller 100 of the storage 11 when executing the backup process. The copy command is a command for instructing backup. The copy command includes information for specifying a range of data to be copied, that is, a data range for a backup process, and information for specifying a copy destination storage. Upon receiving the copy command, the communication unit 102 passes the copy command to the processing sequencer 103 and instructs the processing sequencer 103 to start the copying process.

  When the processing sequencer 103 receives the copy command, as shown in FIG. 4, first, the processing sequencer 103 sets the first block in the range of data to be copied as a block to be transferred (step S100). Next, the communication unit 102 is instructed to transmit a write command to the relay device corresponding to the copy destination storage specified by the copy command (step S101). The write command includes information indicating the block size, that is, the data amount. Also, a request for reading data to be transferred is registered in the IO scheduler 104. In response to the read request, the IO scheduler 104 instructs the medium control unit 105 to read data to be transferred. The medium control unit 105 causes the data to be transferred to be output from the storage medium 101 to the buffer memory 106 according to the read instruction (step S102). When all data to be transferred is output from the storage medium 101 to the buffer memory 106, the medium control unit 105 outputs a read completion notification to the processing sequencer 103.

  Then, the processing sequencer 103 waits for both a reception completion message from the relay device 15 input via the communication unit 102 and a read completion notification from the medium control unit 105 (step S103). When both are input, the communication unit 102 is instructed to transfer the data stored in the buffer memory 106 to the relay device 15. In response to the instruction, the communication unit 102 transmits the data stored in the buffer memory 106 to the relay device 15 (step S104). Then, the processing sequencer 103 waits for a reception completion message from the relay device 15 input via the communication unit 102 (step S105).

  The communication unit 102 transmits all the data stored in the buffer memory 106 to the relay device 15 and outputs a reception completion message to the processing sequencer 103 when receiving a reception completion message from the relay device 15. When the processing sequencer 103 inputs a reception completion message, the processing sequencer 103 checks whether or not the transfer of all data to be copied to the relay device 15 has been completed (step S106). If not completed, the next block in the range of data to be copied is set as the block to be transferred (step S108), and the process returns to step S101.

  If the transfer of all data to be copied has been completed, the processing sequencer 103 instructs the communication unit 102 to output a completion notification to the host 10 (step S107), and ends the processing.

  Note that the data amount of one block is an amount set in the system in advance. Further, the processing sequencer 103 may change the data amount of one block according to the capacities of the buffer memory 106 of the transfer destination storage 12 and the buffer memory 152 of the relay device 15.

  The processing sequencer 103 registers with the IO scheduler 104 as a type of processing (read / write), an ID for identifying the processing, information indicating an area in the storage medium 101 to be processed, There is information indicating an area of the target buffer memory 106. Note that the processing sequencer 103 may instruct a plurality of read processing instructions and a plurality of write processing instructions, and the ID for identifying the process is used to identify each process in such a case. Therefore, in the process shown in FIG. 4, the process sequencer 103 registers the type of process and an ID for identifying the read process when registering the read request for the data to be transferred, and the process target. The block to be transferred is registered as an area in the storage medium 101. When there are a plurality of storage media, the processing sequencer 103 also registers information for specifying the storage media. Then, the process sequencer 103 determines which process has been completed based on the ID specified at the time of registration.

  The IO scheduler 104 records the process registered by the process sequencer 103. Then, the recorded process is selected according to a predetermined algorithm, and the medium control unit 105 is caused to execute the extracted process. As a predetermined algorithm, for example, there is a registered order. Alternatively, a process that minimizes the moving distance from the current position of the magnetic head or optical head of the storage medium 101 to the position to be processed may be selected first. Further, when the storage 11 is a disk array device including a plurality of recording media 101 and the medium control unit 105 is provided corresponding to each storage medium, the IO scheduler 104 stores the memory to be processed. A process for the storage medium 101 that is handled by the medium control unit 105 corresponding to the medium 101 is selected.

  When the processing being executed by the medium control unit 105 is completed, the IO scheduler 104 takes out the next processing. If the process is a read process, the medium control unit 105 causes the storage medium 101 to read data from a specified area in the storage medium 101, and reads the read data into a specified area in the buffer memory 106. To write to. If the process is a write process, the medium control unit 105 causes the storage medium 101 to write the data in the specified area of the buffer memory 106 to the specified area in the storage medium 101. Then, when the process is completed, the medium control unit 105 notifies the process sequencer 103 of the completion of the process.

  When the data input from the outside is a control system message such as a command, a reception completion message, or a reception completion message, the communication unit 102 passes the input data to the processing sequencer 103. If the data is to be written to the storage 11, the processing sequencer 103 is inquired as to where to store the data. Then, the data is stored in an area in the buffer memory 106 designated by the processing sequencer 103.

  The communication unit 102 also performs processing for transmitting a command or completion notification designated from the processing sequencer 103 to the designated relay device or host. Further, processing for transmitting the data in the buffer memory 106 designated by the processing sequencer 103 to the designated relay device or host is also performed.

  The host 10 recognizes that the data transfer from the storage 11 to the storage 12 is completed when the completion notification is received from the storage 11, that is, when the data transfer to the relay device 15 is completed.

  Next, the operation of the relay device 15 will be described. In the relay device 15, data (command or transferred data) from the storages 11 and 12 is received by the communication unit 150. When receiving the write command transmitted from the storage 11 in step S101, the communication unit 150 passes the write command to the relay processing unit 151.

  As shown in FIG. 5, the relay processing unit 151 reserves an area necessary for storing data sent from the storage 11 in the buffer memory 152 (step S120). The relay processing unit 151 reserves a data storage area based on the data amount information included in the write command. Then, the communication unit 150 is instructed to send a reception preparation completion notification to the storage 11 (step S121). The communication unit 150 transmits a reception preparation completion message to the storage 11 in response to the instruction. Further, the communication unit 150 is instructed to send a write command to the storage 12 (step S122). The communication unit 150 transmits a write command to the storage 12 in response to the instruction. This write command is a write command for writing data received from the storage 11 to the storage 12.

  Next, the process waits for data to arrive from the storage 11 (step S123). When the data arrives and receives an inquiry from the communication unit 150 about the area of the buffer memory 152 where the data is to be stored, the communication unit 150 determines the area secured in step S120. (Step S124). In addition, waiting for completion of storage of data in the buffer memory 152 (step S125), and when the communication unit 150 notifies that all data has been stored in the buffer memory 152, notifies the storage 11 of completion of the write command. The communication unit 150 is instructed to do so (step S126). The communication unit 150 transmits a reception completion message to the storage 11 in response to the instruction.

  Then, it waits for a preparation completion message (response to the write command transmitted in step S122) from the storage 12 (step S127). When the communication unit 150 notifies that the preparation completion message from the storage 12 has been received, the communication unit 150 causes the data stored in the buffer memory 152 to be transmitted to the storage 12 (step S128). Thereafter, the process waits for a reception completion message transmitted from the storage 12 (step S129). When the communication unit 150 notifies the reception completion message from the storage 12, the processing is terminated.

  When receiving a command from the outside, the communication unit 150 passes the received command to the relay processing unit 151. Further, when various messages are received from the outside, the relay processing unit 151 is notified of the reception. Further, when data is received, the relay processing unit 151 is inquired about the area of the buffer memory 152 where the data is to be stored, and the data is stored in the designated area in the buffer memory 152. Further, in response to an instruction from the relay processing unit 151, the designated message is transmitted to the designated storage. Further, when an area in the buffer memory 152 and a destination storage are designated, the data in the area is transmitted to the designated storage.

  When the relay processing unit 151 determines the storage of the data transmission destination, if the transmission destination storage (the storage 12 in this example) is fixedly determined in advance, no selection process is performed. When the data destination storage is determined according to the transfer source storage (storage 11 in this example), processing for selecting the destination storage is performed according to the transfer source storage. In addition, prior to transmitting data in the destination storage, the destination may be specified from the data source storage. The data stored in the buffer memory 152 is discarded when transmission of data to the destination storage (storage 12 in this example) is completed.

  When the communication unit 102 of the storage 12 receives the write command transmitted by the relay device 15 in step S <b> 122, the communication unit 102 passes the write command to the processing sequencer 103. The processing sequencer 103 secures an area necessary for storing data sent from the relay device 15 in the buffer memory 106. Then, the communication unit 102 is caused to transmit a preparation completion message to the relay device 15. This preparation completion message is a message that the relay device 15 waits in step S127. Thereafter, when data is sent from the relay device 15, the communication unit 102 of the storage 12 inquires the processing sequencer 103 about the area where the data is to be stored, and stores it in the buffer memory 106. When all the data is stored, a reception completion message is transmitted to the relay device 15. This reception completion message is a message that the relay device 15 waits in step S129. Further, after storing data in the buffer memory 106, the processing sequencer 103 of the storage 12 stores the type of processing (in this case, writing), the processing identification ID, and the processing target storage for the IO scheduler 104. Information indicating an area in the medium 101 and information indicating an area of the buffer memory 106 to be processed are registered. The IO scheduler 104 instructs the medium control unit 105 to write data to be written in response to the write request. The medium control unit 105 performs data writing processing from the buffer memory 106 to the storage medium 101 in accordance with the registered contents.

  In this embodiment, the relay device 15 is installed outside the range where the influence of the disaster is expected to spread when a disaster such as an earthquake occurs at the installation location of the storage 11. Further, the relay device 15 is installed at a position where the data transfer time between the storage 11 and the relay device 15 is shorter than the data transfer time when data is directly transferred between the storage 11 and the storage 12. Has been. Further, when transferring data from the storage 11 to the storage 12 for backup, the host 10 using the storage 11 transfers data from the storage 11 to the storage 12 when the data transfer to the relay device 15 is completed. Recognizes that data transfer is complete.

  Therefore, even if the storage 11 is damaged, the relay device 15 is not damaged, and the data transfer time from the storage 11 to the relay device 15 is short, so that the fault tolerance is improved. Further, since the host 10 can start the next process without waiting for the data to be stored in the storage 12, the delay in the process associated with the data transfer is improved.

Embodiment 2. FIG.
In the first embodiment, the relay device 15 receives all of one block of data from the storage 11 and then starts transmitting data to the storage 12. However, the relay device 15 receives one block of data from the storage 11. Data transmission to the storage 12 may be started without waiting for completion. FIG. 6 is a flowchart illustrating the operation of the relay processing unit 151 according to the second embodiment that performs control to start transmission of data to the storage 12 without waiting for completion of reception of data from the storage 11. The configuration of the data replication system and the configurations of the storages 11 and 12 and the relay device 15 are the same as those in the first embodiment (see FIGS. 1 to 3).

  As in the case of the first embodiment, the relay processing unit 151 receives the write command transmitted by the storage 11 in step S101. Then, the relay processing unit 151 secures an area necessary for storing data sent from the storage 11 in the buffer memory 152 as shown in FIG. 6 (step S140). The relay processing unit 151 secures a data storage area based on data amount information included in the write command from the storage 11. Then, the communication unit 150 is instructed to send a notification of reception preparation completion to the storage 11 (step S141). Further, the communication unit 150 is instructed to send a write command to the storage 12 (step S142). This write command is a write command for writing data received from the storage 11 to the storage 12.

  Next, it waits for data to arrive from the storage 11 (step S143), and when the data arrives and receives an inquiry from the communication unit 150 about the area of the buffer memory 152 where the data is to be stored, the area secured in step S140 is the communication unit 150. (Step S144). Next, the storage unit 12 waits for a preparation completion message to be transmitted (step S145). When the communication unit 150 notifies that the preparation completion message has been received from the storage unit 12, the communication unit 150 receives the buffer memory. The data stored in 152 is transmitted to the storage 12 (step S146).

  While performing the process of step S146, the data from the storage 11 is stored in the buffer memory 152, and the data read from the buffer memory 152 is transmitted to the storage. Control is performed so that the reading position (reading address) does not overtake the storage position (storage address). That is, when the reading position catches up with the storage position, reading of data from the buffer memory 152 is stopped.

  Then, it waits for the storage of data in the buffer memory 152 to be completed (step S147). When the communication unit 150 notifies that all the data has been stored in the buffer memory 152, the storage 11 is completed for the write command. Is notified to the communication unit 150 (step S148). Thereafter, the process waits for a reception completion message transmitted from the storage 12 (step S149). When the communication unit 150 notifies that the reception completion message has been received from the storage 12, the processing ends.

  In this embodiment, data transmission from the relay device 15 to the storage 12 can be completed earlier than in the first embodiment. Note that the operations of the storages 11 and 12 are the same as those of the first embodiment.

Embodiment 3 FIG.
In the first and second embodiments, the backup of the data in the storage 11 is realized. However, the data in the storage 11 may be transferred to the storage 12 by mirroring. FIG. 7 is a flowchart showing the operation of the processing sequencer 103 in the storage controller 100 of the storage 11 in the third embodiment, that is, when mirroring is performed. The configuration of the data replication system and the configurations of the storages 11 and 12 and the relay device 15 are the same as those in the first embodiment (see FIGS. 1 to 3). In the storage 11, the operations of the communication unit 102, the IO scheduler 104, and the medium control unit 105 are the same as those in the first embodiment.

  The backup shown in the first and second embodiments is started when the host 10 outputs a copy command to the storage 11. The mirroring shown as the third embodiment is started when the host 10 outputs a write command instructing the storage 11 to write data.

  When the communication unit 102 of the storage 11 receives a write command from the host 10, the communication unit 102 passes the write command to the processing sequencer 103. Then, as shown in FIG. 7, the processing sequencer 103 secures an area necessary for storing data received from the host 10 in the buffer memory 106 (step S160). Further, the communication unit 102 is instructed to send a notification of completion of preparation to the host 10 (step S161). The communication unit 102 sends a notification of preparation completion to the host 10 in response to the instruction.

  The processing sequencer 103 waits for data to arrive from the host 10 (step S162). When the data arrives and receives an inquiry from the communication unit 102 about the area of the buffer memory 106 where the data is to be stored, the processing sequencer 103 secures it in step S160. The communication unit 102 is notified of the completed area (step S163). Next, the communication unit 102 is instructed to send a write command to the relay device 15 (step S164). The communication unit 102 transmits a write command to the relay device 15 in response to the instruction.

  Next, the processing sequencer 103 waits for the storage of data from the host 10 to the buffer memory 106 (step S165), and when the communication unit 102 notifies that all the data has been stored in the buffer memory 106, The processing sequencer 103 instructs the IO scheduler 104 to specify the processing type (in this case, writing), the processing identification ID, information indicating the area in the storage medium 101 to be processed, and the processing target. Information indicating the area of the buffer memory 106 is registered. The IO scheduler 104 instructs the medium control unit 105 to write data to be written in response to the write request. The medium control unit 105 performs a data writing process from the buffer memory 106 to the storage medium 101 in accordance with the registered content (step S166).

  Then, the processing sequencer 103 waits for a reception preparation completion message (response to the write command sent in step S164) from the relay device 15 (step S167). When the communication unit 102 is notified that the reception preparation completion message from the relay device 15 has been received, the processing sequencer 103 transmits the data stored in the buffer memory 106 to the relay device 15. (Step S168). Thereafter, the communication unit waits for a reception completion message from the relay device 15 and a write completion notification from the medium control unit 105 (step S169), and that the communication unit has received the reception completion message from the relay device 15 When notified from 102 and a write completion notification from the medium control unit 105 is received, the host 10 is notified of the completion of writing (step S170), and the process is terminated. When the host 10 receives the completion notification from the storage 11, that is, when the data transfer to the relay device 15 and the data writing to the storage medium 101 are completed, the host 10 recognizes that the mirroring is completed.

  The operation in which the relay device 15 relays data sent from the storage 11 and the operation in which the storage 12 stores data sent from the relay device 15 in the storage medium 101 are the same as in the first embodiment.

  Also in this embodiment, even if the storage 11 is damaged, the relay device 15 is not damaged, and the data transfer time from the storage 11 to the relay device 15 is short, so that the fault tolerance is improved. In addition, the time from when the host 10 transmits the write command to when the next process is started can be shortened.

Embodiment 4 FIG.
FIG. 8 is a block diagram showing a fourth embodiment of the data replication system according to the present invention. In the data replication system shown in FIG. 8, the storage 11 is locally connected to the host 10 that uses the storage 11. The storage 11 is connected to the relay devices 15-1 to 15-n via the network 13. In addition, the relay devices 15-1 to 15-n are connected to the storage 12 via the network 14. The configurations of the storages 11 and 12 are the same as the configurations of the storages 11 and 12 of the first embodiment, and the configurations of the relay devices 15-1 to 15-n are the relay devices of the first embodiment. The configuration is the same as 15. In addition, the operation in which each relay device 15-1 to 15-n transfers the data received from the storage 11 to the storage 12, and the operation in which the data received by the storage 12 is stored in the storage medium 101 are described in the first embodiment. The operations of the relay device 15 and the storage 12 are the same.

  Similar to the first embodiment, the storage 11 actively transmits data. That is, the host 10 does not read data from the storage 11 and transmits data to the relay apparatuses 15-1 to 15-n, but the storage 11 directly transmits data to the relay apparatus 15. The storage 12 also receives data directly from the relay device 15. In addition, the relay devices 15-1 to 15-n are installed outside the range in which the influence of the disaster is expected to spread when a disaster such as an earthquake occurs at the installation location of the storages 11 and 12. Furthermore, the data transfer time between the storage 11 or storage 12 and the relay devices 15-1 to 15-n is shorter than the data transfer time when data is directly transferred between the storage 11 and the storage 12. In addition, relay devices 15-1 to 15-n are installed.

  Next, the operation of the data replication system will be described. FIG. 9 is a flowchart showing the operation of the processing sequencer 103 in the storage controller 100 of the storage 11. Here, an example in which the storage 11 as the data transfer source transfers data to the storage 12 as the data transfer destination to perform backup will be described.

  When the processing sequencer 103 receives a copy command from the host 10 via the communication unit 102, first, as shown in FIG. 9, the processing sequencer 103 sets the first block in the range of data to be copied as a block to be transferred (step). S200). Next, a relay device to be used is selected from the relay devices 15-1 to 15-n (step S201). Here, it is assumed that the relay device 15-1 is selected. Then, the communication unit 102 is instructed to transmit a write command to the selected relay device 15-1 (step S202). Also, a request for reading data to be transferred is registered in the IO scheduler 104 (step S203). The operations of the IO scheduler 104 and the medium control unit 105 when a read request is registered in step S203 are the same as the operations in step S102 described in the first embodiment.

  Then, the processing sequencer 103 waits for both a reception preparation completion message from the relay device 15-1 input via the communication unit 102 and a read completion notification from the medium control unit 105 (step S204). When both are input, the communication unit 102 is instructed to transfer the data stored in the buffer memory 106 to the relay device 15-1 (step S205). Then, the processing sequencer 103 waits for a reception completion message from the relay device 15-1 input via the communication unit 102 (step S206).

  The communication unit 102 transmits the data stored in the buffer memory 106 in step S <b> 203 to the relay device 15-1, and upon receiving a reception completion message from the relay device 15-1, outputs the reception completion message to the processing sequencer 103. To do. When receiving the reception completion message, the processing sequencer 103 checks whether or not the transfer of all data to be copied to the relay device 15 has been completed (step S207). If not completed, the next block in the range of the data to be copied is set as the block to be transferred (step S209), and the process returns to step S201.

  If the transfer of all data to be copied has been completed, the processing sequencer 103 instructs the communication unit 102 to output a completion notification to the host 10 (step S208), and ends the processing.

  Note that the data amount of one block is an amount set in the system in advance. Further, the processing sequencer 103 may change the data amount of one block according to the capacities of the buffer memory 106 of the transfer destination storage 12 and the buffer memory 152 of the relay device 15.

  As a method of selecting a relay device in step S201, for example, the relay devices 15-1 to 15-n are sequentially selected, or a relay device with a light load (not used for other data transfer) is selected. There is a method of selecting using a random number.

  Alternatively, data transfer may be performed in parallel using a plurality of relay devices simultaneously. FIG. 10 is a flowchart showing the operation of the storage 11 when a plurality of relay devices are used simultaneously. Here, a case where one block of data is transferred is taken as an example.

  In the storage 11, when the processing sequencer 103 sets the first block in the range of data to be copied as a block to be transferred (step S 220), the storage 11 sends to all the relay devices 15-1 to 15 -n. Data is transferred (step S221).

  FIG. 11 is a flowchart specifically showing the process of step S221. The processing sequencer 103 transmits data to each of the relay devices 15-1 to 15-n according to the flowchart shown in FIG. Here, a case where data is transmitted to the relay device 15-1 will be described as an example. The processing sequencer 103 selects data to be transmitted to the relay device 15-1 from the data stored in the buffer memory 106 (step S240). Then, the communication unit 102 is instructed to send a write command to the relay device 15-1 (step S241). Also, a read request for data to be transferred is registered in the IO scheduler 104 (step S242). The operations of the IO scheduler 104 and the medium control unit 105 when a read request is registered in step S242 are the same as the operations in step S102 described in the first embodiment. The processing sequencer 103 waits for both a reception preparation completion message from the relay device 15-1 input via the communication unit 102 and a read completion notification from the medium control unit 105 (step S243). If both are input, the communication unit 102 is instructed to transmit the data stored in the buffer memory 106 to the relay device 15-1 (step S244). Here, the case where data is transmitted to the relay device 15-1 has been described as an example. However, the processing sequencer 103 executes the processing of steps S240 to S244 for all the relay devices 15-1 to 15-n.

  Furthermore, when the processing sequencer 103 is notified from the communication unit 102 that a reception completion message has been received from one of the relay devices (step S222), it checks whether there is still untransferred data (step S223). In some cases, the communication unit 102 is instructed to send a write command to the relay apparatus that has transmitted the reception completion message (step S224). The specific process of step S224 is the process shown in FIG.

  If it is confirmed in step S223 that there is no untransferred data, the processing sequencer 103 waits for reception of a reception completion message from each relay device (step S225), and reception completion messages from all the relay devices that have transferred the data. If received (step S226), the communication unit 102 is instructed to output a completion notification to the host 10 (step S227), and the process is terminated.

  FIG. 12 is a timing chart showing an example of data transfer when one block of data is divided into five and three relay apparatuses 15-1, 15-2, and 15-3 are used. In response to the copy command from the host 10, the storage 11 transmits a write command to each of the relay devices 15-1, 15-2, and 15-3. Then, the data (1), (2), and (3) are transferred to the relay devices 15-1, 15-2, and 15-3 that have transmitted the reception preparation completion message.

  In the example illustrated in FIG. 12, since the relay apparatuses 15-2 and 15-3 have transmitted the completion notification first, the storage 11 transmits a write command to the relay apparatuses 15-2 and 15-3. Data (4) and (5) are transmitted. When the storage 11 confirms reception of completion notifications from all the relay apparatuses 15-1, 15-2, and 15-3, the storage 11 outputs completion notifications to the host 10.

  In this embodiment, data transfer can be executed using a plurality of paths. Therefore, it can be made difficult to be influenced by the usage rate of each route and the processing capability of the relay device. For example, in the example shown in FIG. 12, the route to the relay device 15-1 is congested or the processing capability of the relay device 15-1 is low. By using 15-3, the congestion of the route to the relay device 15-1 or the low processing capability of the relay device 15-1 does not significantly affect the data transfer from the storage 11.

  Here, a case has been described where different data is transmitted to each of the relay devices 15-1 to 15-n, but a write command of the same data is transmitted to each of the relay devices 15-1 to 15-n. Also good. For example, it is assumed that there are data A to C as data to be transmitted. The storage 11 transmits any write command of the data A to C to each relay device 15-1 to 15-n.

  In this case, the storage 11 provides an identifier for each identical data set to be transmitted to each relay apparatus 15-1 to 15-n, and adds the identifier to each write command. As the identifier, a numerical value indicating the order of each data set to be transmitted may be used. For example, if data A is first transmitted to each relay apparatus, an identifier “1” representing the first is added to each write command transmitted to each relay apparatus. Subsequently, if data B is transmitted to each relay device, an identifier “2” representing the second is added to each write command transmitted to each relay device. In the same way, identifiers are added to other data write commands.

  When each relay apparatus 15-1 to 15-n notifies the storage 11 of the completion of data transfer, it also notifies the identifier added to the write command. Accordingly, the storage 11 receives a response to which the same identifier is added from each relay device 15-1 to 15-n. If the identifier added to the response is the first identifier received, the storage 11 notifies the host 10 of the end of the data transfer. Thereafter, when a response to which the same identifier is added is received, no processing is performed on the response.

  Each relay device 15-1 to 15-n transmits a write command to the standby storage 12 with the identifier added. Therefore, the storage 12 receives a write command to which the same identifier is added from each relay device 15-1 to 15-n. If the identifier added to the write command is an identifier received for the first time, the storage 12 proceeds with processing according to the write command. Thereafter, when a write command to which the same identifier is added is received, only a transfer completion notification is sent to the relay apparatus that has transmitted the write command.

  If the write command of the same data is transmitted to each relay device in this way, the process can be advanced using the fastest path among the plurality of data transfer paths. Therefore, when the storage 11 performs mirroring or backup, the waiting time for a response from the relay device can be shortened. As a result, the time until the host 10 starts the next process is shortened.

  In the first embodiment to the fourth embodiment, the relay device 15 is provided with a nonvolatile storage device, and the relay device 15 records the command and data received from the storage 11 in the nonvolatile storage device. The command recorded in the volatile storage device may be issued to the storage 12 at an arbitrary timing. As the nonvolatile storage device provided in the relay device 15, for example, a single magnetic disk device, an optical disk device, or a disk array device that is a set of magneto-optical disk devices may be used. Alternatively, a battery-backed memory may be used.

  The relay device 15 may issue a write command to the storage 12 when, for example, the communication amount of the network 14 to which the relay device 15 and the storage 12 are connected falls below a predetermined threshold. The write command is issued when a predetermined time has elapsed since the command and data were recorded in the non-volatile storage device, when a predetermined time has been reached, or when the storage 12 is requested to issue a write command, etc. It may be.

  Thus, if the command and data from the storage 11 are recorded on the nonvolatile storage device, the command and data are not lost. Then, the write command can be issued to the storage 12 at an arbitrary timing without interlocking with reception of the write command from the storage 11. As a result, the storage 12 need not always be operated. Similarly, the network 14 to which the relay device 15 and the storage 12 are connected need not always be connected, and the operation cost of the network 14 is reduced.

  Normally, as shown in the first to fourth embodiments, a write command is issued to the storage 12 in conjunction with the reception of the write command from the storage 11, and the communication amount of the network 14 is predetermined. When the threshold is exceeded, commands and data may be recorded in the nonvolatile storage device. In this case, the traffic on the network 14 can be averaged. Further, a line having a low operation cost can be used as the network 14, and the cost of the entire data replication system can be reduced.

Embodiment 5. FIG.
FIG. 13 is a block diagram showing a fifth embodiment of the data replication system according to the present invention. In the data replication system shown in FIG. 13, the storage 20 is locally connected to the host 10 that uses the storage 20. The storage 20 is connected to the storage 21 via the network 13.

  The storages 20 and 21 are, for example, a single magnetic disk device, an optical disk device, or a magneto-optical disk device. As the storages 20 and 21, a disk array device which is a single magnetic disk device, an optical disk device or a set of magneto-optical disk devices can be used. The host 10 and the storage 20 are connected by SCSI, Fiber channel, Ethernet (registered trademark), or the like.

  FIG. 14 is a block diagram illustrating a configuration example of the storage 20 illustrated in FIG. The configuration of the storage 21 is also as shown in FIG. As shown in FIG. 14, the storage 20 includes a storage controller 200 and a storage medium 101 that is a storage body. The storage controller 200 is issued by a communication unit 204 that communicates with the host 10 and other storages, a processing sequencer 201 that manages the sequence of each process, an IO scheduler 104 that controls the order of processing instructions for the storage medium 101, and an IO scheduler 104 In the medium processing unit 105 that controls the operation of the storage medium 101 according to the processing instruction to be performed, the buffer memory 106 that temporarily stores data from the host 10 or the like to the storage medium 101 and data from the storage medium 101 to the host 10 or the like A redundancy unit 202 that makes the data to be sent redundant, and a restoration unit 202 that restores the original data from the redundant data sent from another storage are included. The processing sequencer 103 is realized by a CPU that operates according to a program, for example.

  Next, the operation of the data replication system of the fifth embodiment will be described with reference to the flowcharts of FIGS. FIG. 15 is a flowchart showing the operation of the processing sequencer 201 in the storage controller 200, and FIG. 16 is a flowchart showing the operation of the communication unit 204.

  When the processing sequencer 201 receives a copy command from the host 10 via the communication unit 204, as shown in FIG. 15, first, the first block in the range of data to be copied is set as a block to be transferred (step S300). Next, the communication unit 204 is instructed to transmit a redundant write command to the copy destination storage designated by the copy command (step S301). The redundant write command is a command for instructing writing based on redundant data. The redundancy write command includes information indicating that the data is made redundant together with information indicating the block size, that is, the data amount. Further, the processing sequencer 201 registers a read request for data to be transferred with the IO scheduler 104. In response to the read request, the IO scheduler 104 instructs the medium control unit 105 to read data to be transferred. The medium control unit 105 causes the data to be transferred to be output from the storage medium 101 to the buffer memory 106 according to the read instruction (step S302).

  Then, the processing sequencer 201 waits for both a reception preparation completion message from the storage 21 input via the communication unit 204 and a read completion notification from the medium control unit 105 (step S303). If both are input, the redundancy unit 202 is caused to make the data stored in the buffer memory 106 redundant (step S304). Redundancy is performed by creating redundant data from original data (hereinafter referred to as original data). In the following description, a set of original data and newly created redundant data is referred to as a redundant data group. The processing sequencer 201 instructs the communication unit 204 to transfer the redundant data group to the storage 21. In response to this instruction, the communication unit 204 transmits the redundant data group to the storage 21 (step S305).

  However, the communication unit 204 transmits the original data and the redundant data to the storage 21 as a set of separate data. The “data group” is a minimum data transmission unit of various data transfer protocols. Specific examples of the “data group” include, for example, a segment in TCP or UDP, a packet in Internet protocol, a frame in a physical layer of Fiber Channel protocol or Ethernet (registered trademark), and the like. Therefore, for example, when transmitting according to the Internet protocol, the communication unit 204 transmits the original data and the redundant data in separate packets. Generally, error detection data for detecting an error (data corruption) that occurs during transmission is added inside a data transmission unit such as a packet. The redundant data created in step S304 is data transmitted as a group of independent data, and is different from error detection data added to a packet or the like.

  Further, as will be described later, the original data is divided, and redundant data is created based on the divided original data. In step S305, the communication unit 305 transmits the divided original data as individual data transmission units. For example, it is assumed that the original data is divided into N pieces and m redundant data are created from the N pieces of data. In this case, the communication unit transmits the original data and redundant data in N and m packets (or segments, frames, etc.), respectively. If the communication unit 204 transmits the original data and the redundant data as separate data sets, the communication unit 204 waits for a reception completion message from the storage 21 input via the communication unit 204 (step S306).

  The communication unit 204 transmits all the redundant data groups (original data and redundant data) to the storage 21, and outputs a reception completion message to the processing sequencer 201 when receiving a reception completion message from the storage 21. When receiving the reception completion message, the processing sequencer 201 confirms whether or not the transfer of all data to be copied to the relay device 15 has been completed (step S307). If not completed, the block next to the copy target data range is set as the data transfer target block (step S309), and the process returns to step S301.

  If the transfer of all data to be copied has been completed, the processing sequencer 201 instructs the communication unit 204 to output a completion notification to the host 10 (step S308), and ends the processing.

  Note that the data amount of one block is an amount set in the system in advance. Further, the processing sequencer 201 may change the data amount of one block in accordance with the capacity of the buffer memory 106 of the transfer destination storage 21.

  If the data input from the outside is a control system message such as a command, a reception completion message, or a reception completion message, the communication unit 204 passes the input data to the processing sequencer 201. If the data is to be written to the storage 20, the processing sequencer 201 is inquired about the location where the data is to be stored. Then, the data is stored in an area in the buffer memory 106 designated by the processing sequencer 201, or the data is stored in the area in the designated buffer memory 106 after the restoration unit 203 restores the data.

  The communication unit 204 also performs processing for transmitting a command or completion notification designated by the processing sequencer 201 to the designated storage or host. Further, processing for transmitting data in the buffer memory 106 designated by the processing sequencer 201 to the designated storage or host is also performed. In addition, a process of receiving redundant data from the redundancy unit 202 and transmitting the received data to the designated storage or host 10 is also performed.

  Note that when the command is transmitted to another storage, the communication unit 204 adds a command identifier for distinguishing each command to the command. As an example of the command identifier, there is an issue number (issue ID described later) which is a numerical value obtained by adding 1 each time a command is transmitted. Further, when transmitting a redundant data group, a command identifier of a command related to the transmission is added. For example, when a redundant write command is transmitted, a command identifier for identifying the redundant write command is added. Then, the same command identifier as that of the redundant write command is added to the redundant data group (original data and redundant data) corresponding to the redundant write command.

  Further, the communication unit 204 adds a storage identifier as a transmission source to each command and data.

  Next, the operation of the redundancy unit 202 will be described. The redundancy unit 202 acquires designated data from the buffer 106 and creates a redundant data group using a prescribed redundancy method. The redundancy unit 202 divides the designated original data into N pieces of data. Then, m redundant data are created from the original data divided into N pieces. N and m are natural numbers. The N + m data groups become redundant data groups. Identification information for use at the time of restoration is added to the redundant data group.

  Hereinafter, a specific example of redundancy will be described. The redundancy unit 202 divides the original data into N by, for example, dividing the original data into N equal parts from the head. When this method is adopted, a number indicating the number of data from the beginning is used as identification information. When the original data is divided into N pieces of data, the redundancy unit 202 creates parity data by parity calculation in the same manner as in RAID 3 and RAID 5 devices, and makes the parity data redundant data. That is, redundant data is created so that the sum of the values of the corresponding bits is always an odd number (or necessarily an even number) in each piece of N data and redundant data. In this case, one piece of redundant data may be created. If redundant bits are created in this way, even if one of the divided N original data is missing, the missing data can be restored.

  The redundancy method is not limited to the above parity operation, and the number of redundant data is not limited to one. For example, redundant data may be created by a double parity operation. Further, ECC (Error Correcting Code) based on the original data may be used as redundant data. There are various redundancy methods as described above. In the following description, a case where one piece of redundant data is created by parity calculation will be described as an example.

  Next, the operation of the storage 21 that is the data transfer destination will be described. The redundant write command is sent from the storage 20 to the communication unit 204 in the storage controller 200 of the storage 21. Upon receiving the redundant write command, the communication unit 204 passes the redundant write command to the processing sequencer 201 and instructs the start of processing based on the redundant write command (redundant write processing).

  FIG. 16 is a flowchart showing the operation of the processing sequencer 201 in the storage 21. The processing sequencer 201 can perform a plurality of redundant write processes in parallel, and identifies each process using the identifier of the storage that issued the command and the command identifier added to the command.

  In the redundant write process, the process sequencer 201 first secures an area in the buffer 106 where the transmitted data can be stored (step S320). Next, the communication unit 204 is caused to transmit a reception preparation completion notification to the storage 20 that has transmitted the redundant write command (step S321). Then, it waits for data to arrive from the storage 20 (step S322). When the communication unit 204 notifies that the data has arrived, the identifier of the data transfer source storage 20 and the command identifier added to the data are designated. Then, the communication unit 204 is instructed to send the data to the restoration unit 203 (step S323).

  Next, the processing sequencer 201 determines the number of data sent in response to the redundant write command (step S324). Since the command identifier corresponding to the redundant write command is added to each data, the processing sequencer 201 confirms whether or not each received data is data transmitted corresponding to the redundant write command. be able to. When the number of data is less than the predetermined number (n), the process proceeds to step S322, and when the number is the predetermined number (n), the process proceeds to step S325 (step S324). This predetermined number is the number of data from which the original data can be restored, and differs depending on the redundancy scheme. For example, when one redundant data is created from the original data divided into N pieces by parity calculation, N + 1 pieces of data are sent from the storage 20. In this case, if N pieces of data are received, the original data can be restored. Accordingly, in step S324, it may be determined whether or not the number of received data is less than N.

  In step S325, the processing sequencer 201 designates the identifier and command identifier of the data transfer source storage 20 and the area in the buffer 106 secured in step S320, and instructs the restoration unit 203 to restore the data. Then, the communication unit 204 is instructed to send a response to the data transfer source storage 20 (step S326). Further, the IO scheduler 104 is instructed to write the data in the area in the buffer 106 secured in step S320 to the area in the medium designated by the redundant write command (step S327), and the processing is performed. finish. The operations of the IO scheduler 104 and the medium control unit 105 when this instruction is issued to the IO scheduler 104 are the same as the operations in step S102 described in the first embodiment.

  Note that the processing sequencer 201 instructs the communication unit 204 to discard the data related to the redundant write command that has arrived at the communication unit 204 after the start of the process of step S325. For example, when the original data can be restored from N pieces of data, there is no need for the data that has reached the (N + 1) th piece. Data that arrives at the storage 21 after the necessary data arrives is discarded.

  Next, processing of the restoration unit 203 will be described. The restoration unit 203 has a buffer memory, and stores the data passed together with the identifier and command identifier of the data transfer source storage 20 in the buffer memory. When the identifier and command identifier of the data transfer source storage 20 and the area for storing the restored data in the buffer memory 106 are designated and restoration is instructed, the restoration unit 203 reads the data transfer source from the internal buffer. The data corresponding to the identifier and command identifier of the storage 20 is collected. Then, based on the identification information added to each data, the data before redundancy is restored from the data group.

  For example, it is assumed that one redundant data is created by parity calculation from original data divided into N pieces, as in RAID 3 and RAID 5 devices. Then, it is assumed that k-th data among the divided N original data does not arrive at the storage 21, but other N-1 data and one redundant data (a total of N data) have arrived. . In this case, the kth data can be created by ensuring that the sum of the values of the corresponding bits in the kth data and the N pieces of data that have arrived is always odd (or always even). . Then, the original data before the division may be restored using the created kth data and other data that has arrived at the storage 21. Since each piece of divided data is added with identification information (for example, a number indicating the number of data from the beginning), it can be restored from the divided state to the original data before the division. Further, when all of the original data divided into N pieces arrives at the storage 21 and the redundant data does not arrive, the original data before division may be restored from the N pieces of arrived data.

  The restoration unit 203 stores the restored data in the designated area in the buffer memory 106. It should be noted that the area storing the data used in the restoration process in the buffer inside the restoration unit 203 is reused for storing other data after the restoration process is completed.

  As described above, the redundancy method is not limited to a method of creating one redundant data by parity calculation. The transfer destination storage 21 may be controlled so that it can be restored by a method corresponding to the redundancy method. For example, if ECC is used as redundant data, control may be performed so that restoration is performed by a method corresponding to ECC.

  Further, here, the case where the original data is divided into N pieces to create redundant data has been described as an example, but the redundancy unit 202 of the storage 20 does not divide the original data and divides the original data in step S304. The same data may be duplicated and the duplicated data may be used as redundant data. In step S305, the original data may be transmitted as one piece of data (for example, one packet, segment, or frame), and redundant data that is a duplicate of the original data may be sent as one piece of data. . Since the original data and the redundant data are the same data and are not divided, the transfer destination storage 21 may write the received data to the storage medium 101 as it is when either one is received. In addition, data received late may be discarded. Therefore, the processing sequencer 201 of the storage 21 may move to the next step S325 immediately after moving to step S324 for the first time. Furthermore, the restoration unit 203 may move to the next step S326 without performing the restoration process in step S325.

  In this way, if the original data and the redundant data are transmitted as the same data as a group of two separate data, even if one of the data is lost during the transmission process, the storage 21 is Data can be received as it is.

  Further, here, a case is shown in which the same data as the original data is duplicated without dividing the original data, and the duplicated data is used as redundant data. When creating redundant data, the redundancy unit 202 of the storage 20 may divide the original data into N pieces and duplicate the same data for each of the divided N pieces of data. In this case, N data, which is a copy of the N data after division, becomes redundant data. In step S305, the original data may be transmitted as a group of N data (for example, N packets, segments, or frames), and the redundant data may be transmitted as a group of N data. In this case, 2 × N pieces of data are transmitted to the storage 21. If the storage 21 receives a set of data (each data from the first to the Nth data after the division) that can restore the original data among the 2 × N pieces of data, the data Restore the original data based on That is, the processing sequencer 201 of the storage 21 may confirm whether or not all of the divided first to Nth data have been received in step S324. If all the data has been received, the process proceeds to step S325, and if there is data that has not been received, the process may return to step S322.

  If the duplicate data is transmitted as redundant data in this way, even if a part of 2 × N data is discarded in the transmission process, the storage 21 performs processing from the first data to the Nth data after division. Data can be restored when all are received. The redundancy unit 202 of the storage 20 may divide the original data and then create a copy of each data after the division, or may duplicate the same data as the original data before the division, Data and replicated data may be divided respectively.

  According to the fifth embodiment, redundant data is created, and original data and redundant data are transmitted separately. Therefore, even if a part of data is lost in the transmission process, the original data can be restored. Or the original data which the transmission source tried to transmit can be received as it is. As a result, when a part of the original data is lost in the transmission process, it is not necessary to notify the transfer source storage 20 that the original data has not arrived and to send the data to the storage 20 again. It can be shortened.

  In the data replication system shown in FIG. 13, both original data and redundant data included in a redundant data group are transmitted / received via the network 13. The original data and the redundant data may be transmitted via separate networks. FIG. 17 is a block diagram illustrating a configuration example of a data replication system that transmits original data and redundant data via different networks. In the configuration example illustrated in FIG. 17, the storage 20 is connected to the storage 21 via the network 13 and is also connected to the storage 21 via the network 14. The configuration is the same as that shown in FIG. 13 except that the storage 20 and the storage 21 are connected via the two networks 13 and 14. The configuration of the storages 20 and 21 shown in FIG. 17 is the same as the configuration shown in FIG. However, the communication unit 204 is connected to the two networks 13 and 14.

  Also, the operation of the storage 20 shown in FIG. 17 is the same as the operation of the storage 20 shown in FIG. However, when the communication unit 204 transmits the redundant data group to the storage 21 in step S305 (see FIG. 15), the communication unit 204 transmits the original data and the redundant data via separate networks. For example, the original data is transmitted as a single data transfer unit (for example, as one packet) to the storage 21 via the network 13a, and redundant data that is a copy of the original data is transmitted to the storage 21 via the network 14. To do.

  According to such a configuration, even if a failure occurs in one of the networks 13 and 14 or the transfer speed of one of the networks is slow, the original data or the original data is transmitted via the other network. The same redundant data can be transmitted. Therefore, fault tolerance can be further improved when replicating the original data is redundant data. Even when the redundant data is generated by dividing the original data into N pieces, the divided original data and redundant data may be transmitted via different networks.

Embodiment 6 FIG.
In the fifth embodiment, the backup of the data in the storage 20 is realized. However, the data in the storage 20 may be transferred to the storage 21 by mirroring. FIG. 18 is a flowchart showing the operation of the processing sequencer 201 in the storage controller 200 of the storage 20 in the sixth embodiment, that is, when mirroring is performed. The configuration of the data replication system and the configurations of the storages 20 and 21 are the same as those in the fifth embodiment (see FIGS. 13 and 14). In the storage 20, the operations of the communication unit 204, the IO scheduler 104, and the medium control unit 105 are the same as those in the fifth embodiment.

  The storage 20 starts mirroring when it receives a write command from the host 10. When the communication unit 204 of the storage 20 receives a write command from the host 10, the communication unit 204 passes the write command to the processing sequencer 204. Then, as shown in FIG. 18, the processing sequencer 201 secures an area necessary for storing data received from the host 10 in the buffer memory 106 (step S340). Further, the communication unit 204 is instructed to send a notification of completion of preparation to the host 10 (step S341). The communication unit 204 sends a notification of preparation completion to the host 10 in response to the instruction.

  Then, the processing sequencer 201 waits for data to arrive from the host 10 (step S342). When the data arrives and receives an inquiry about the area of the buffer memory 106 where the data is to be stored from the communication unit 204, the processing sequencer 201 secures it in step S340. The communication unit 204 is notified of the completed area (step S343). Next, the communication unit 204 is instructed to send a redundant write command to the storage 21 (step S344). The communication unit 204 transmits a redundant write command to the storage 21 in response to the instruction.

  Next, the processing sequencer 201 waits for completion of storage of data from the host 10 in the buffer memory 106 (step S345), and when the communication unit 204 notifies that all the data has been stored in the buffer memory 106, The processing sequencer 201 sends a processing type (in this case, writing), a processing identification ID, information indicating an area in the storage medium 101 to be processed, and a processing target to the IO scheduler 104. Information indicating the area of the buffer memory 106 is registered. The IO scheduler 104 instructs the medium control unit 105 to write data to be written in response to the write request. The medium control unit 105 performs data writing processing from the buffer memory 106 to the storage medium 101 according to the registered contents (step S346).

  Then, the processing sequencer 201 waits for a reception preparation completion message (response to the redundant write command sent in step S344) from the storage 21 (step S347). When the communication unit 204 is notified that a reception preparation completion message from the storage 21 has been received, the redundancy unit 202 is instructed to make the data stored in the buffer memory 106 redundant. In response to the instruction, the redundancy unit 202 performs redundancy as in the case of the fifth embodiment (step S348). In step S348, the original data may be divided into N pieces, and redundant data may be created from the divided data. Alternatively, the original data may not be divided, and the duplicated data may be used as redundant data. Further, the same data as each data obtained by dividing the original data may be duplicated, and each duplicated data may be used as redundant data. The same data as the original data may be duplicated, and the original data and the duplicated data may be divided.

  Subsequently, the processing sequencer 201 instructs the communication unit 204 to transmit the data group made redundant by the redundancy unit 202 toward the storage 21 (step S349). In response to this instruction, the communication unit 204 transmits the redundant data group (the divided original data and redundant data) to the storage 21. After that, a reception completion message is transmitted from the storage 21 and a write completion notification from the medium control unit 105 is waited (step S350), and the reception completion message from the storage 21 is received from the communication unit 204. When the notification is received and the writing completion notification is received from the medium control unit 105, the host 10 is notified of the completion (step S351), and the processing is terminated. The operation when the storage 21 receives a redundant data group from the storage 20 is the same as that of the fifth embodiment.

  Similarly to the case illustrated in FIG. 17, the storage 20 and the storage 21 may be connected by the networks 13 and 14. In step S349, when the communication unit 204 transmits the redundant data group, the original data and the redundant data may be transmitted via different networks.

  In the data replication system according to the first to fourth embodiments, as shown in the fifth embodiment or the sixth embodiment when the storage and the relay device transmit and receive data, as shown in the fifth embodiment or the sixth embodiment. You may make it transmit / receive the data group made redundant. In that case, also in the first to fourth embodiments, the same effects as those of the fifth or sixth embodiment can be obtained.

  In the fifth embodiment and the sixth embodiment, the data transfer processing means is realized by the processing sequencer 201 and the communication unit 204. The redundancy means 202 is realized by the redundant stock 202. The restoration means is realized by the restoration unit 203. The storage processing unit is realized by the processing sequencer 201, the IO scheduler 104, and the medium control unit 105.

Embodiment 7 FIG.
FIG. 19 is a block diagram showing a seventh embodiment of the data replication system according to the present invention. In the data replication system shown in FIG. 19, a storage 301 is locally connected to a host (higher level apparatus) 300 that uses the storage 301. Further, the storage 302 is locally connected to a host 303 that uses the storage 302. The storage 301 is connected to the storage 302 via the network 13. Further, the host 300 and the host 303 are connected to be communicable. The host 300 and the host 303 are preferably connected by a dedicated line, but may be connected by a network other than the dedicated line (for example, the Internet).

  The storages 301 and 302 are, for example, single magnetic disk devices, optical disk devices, or magneto-optical disk devices. As the storages 301 and 302, a single magnetic disk device, an optical disk device, or a disk array device that is a set of magneto-optical disk devices can be used. The hosts 300 and 303 and the storages 301 and 302 are connected by SCSI, Fiber channel, Ethernet (registered trademark), or the like. In the system shown in FIG. 19, the host 300 is a normal host that operates when no failure occurs in the system, and the host 303 is a standby host that operates when a failure occurs in the host 300. Suppose there is.

  The hosts 300 and 303 perform processing according to application programs (hereinafter referred to as applications) held by the hosts 300 and 303 themselves. This application uses data in the storages 301 and 302 as a processing target. For example, when storing data such as the deposit amount of a bank customer in the storages 301 and 302, the hosts 300 and 303 update the data in the storages 301 and 302 in accordance with the deposit data management application. The host that actually performs the operation is described below as the operation of the application.

  In this embodiment, the data in the normal storage 301 is mirrored to the standby storage 303 that exists at a remote location. However, when the normal host 300 writes data to the storage 301, the data is not written to the storage medium of the standby storage 303 but is held in the synchronization buffer memory provided in the standby storage 303. . Then, at the timing designated by the host 300, the data in the synchronization buffer memory of the standby storage 303 is written into a predetermined storage medium. For example, when the host starts the process X, it is assumed that data A and B must be written in the storage. In this case, when the normal host 300 writes data A to the storage 301, the standby storage 302 simply holds the data A in the synchronization buffer memory, and writing to the storage medium of the standby storage 302 is not possible. Don't do it. After the host 300 writes the data B to the storage 301, the host 300 designates the write timing to the storage medium of the standby storage 302, and at this timing, the standby storage 302 stores the data A and B in the synchronization buffer memory. Write to storage media.

  The timing at which the application can resume its operation as long as it is in the data state is referred to as a resumable point. That is, the resumable point is a timing at which the state of written data is in a state where processing by an application can be resumed. In the above example, the period from when data A and B are written to the normal storage 301 until the next data is written is a resumable point at which the process X can be started.

  FIG. 20 is a block diagram illustrating a configuration example of the host 300. In the host 300, one or a plurality of applications operate. Here, two applications 310a and 310b are illustrated. The applications 310 a and 310 b access the data in the storage 301 using the IO management unit 311. Further, the IO management unit 311 includes a resumable point notifying unit 312 for notifying the storage 301 of resumable points. In accordance with the applications 310a and 310b, the resumable point notifying unit 312 performs a process of notifying the storage 301 of a resumable point at the resumable point (resumable point notifying process). A host monitoring unit 313 that monitors the state of the host 300 is also provided. The configuration of the host 303 is the same as the configuration of the host 300.

  The applications 310a and 310b in the present embodiment are applications having a resume function. In other words, this is an application that realizes a function that can resume processing if the data recording state of the storage medium 101 of the storage is in a predetermined state.

  FIG. 21 is a block diagram illustrating a configuration example of the storage 301 illustrated in FIG. The configuration of the storage 302 is also as shown in FIG. As shown in FIG. 21, the storage 301 includes a storage controller 320 and a storage medium 101 that is a storage body. The storage controller 320 is issued by a communication unit 322 that communicates with the host 300 and other storages, a processing sequencer 321 that manages the sequence of each process, an IO scheduler 104 that controls the order of processing instructions for the storage medium 101, and an IO scheduler 104 A medium processing unit 105 that controls the operation of the storage medium 101 according to the processing command to be transmitted, a buffer memory 106 that temporarily stores data from the host 300 to the storage medium 101 and data from the storage medium 101 to the host 300, and other storages. A synchronization buffer memory 322 for temporarily storing received data is included. The processing sequencer 321 is realized by, for example, a CPU that operates according to a program. The operations of the IO scheduler 104 and the medium control unit 105 are the same as the operations of the IO scheduler 104 and the medium control unit 105 in the first embodiment.

  A semiconductor memory may be used as the synchronization buffer memory 322, or a larger capacity storage device such as a magnetic disk device, an optical disk device, or a magneto-optical disk device may be used.

  When the applications 310 a and 310 b read data from the storage 301, the host 300 requests the storage 301 to read data. The processing sequencer 321 of the storage 301 copies the requested data from the storage medium 101 to the buffer memory 106 by the IO scheduler 104 or the like. Then, the data is transmitted to the host 300.

  Next, an outline of an operation when the host 300 writes data to the storage 301 will be described. The host 300 outputs data to the storage 301 by outputting a write command to the storage 301. If a resumable point is reached after outputting the write command once or a plurality of times, a resumable point notification command is output to the storage 301 to notify that the point has become a resumable point.

  When the storage 301 receives the write command, the storage 301 writes data to the storage medium 101 in accordance with the write command. Also, a delayed write command is output to the storage 302 and the data is held in the synchronization buffer memory 323 of the standby storage 302. The delayed write command is a command for holding data to be written to the storage medium 101 in the synchronization buffer memory 323 and instructing to write to the storage medium 101 when a later-described delayed data reflection command arrives.

  When the storage 301 receives the resumable point notification command, the storage 301 transmits a delayed data reflection command (delayed write execution request) to the storage 302. The delayed data reflection command is a command for instructing to write the data stored in the synchronization buffer memory to the storage medium 101. When the standby storage 302 receives the delayed data reflection command, the standby storage 302 writes the data stored in the synchronization buffer memory 323 to the storage medium 101. By such an operation, the storage medium 101 of the storage 302 is always kept in a state where processing can be started.

  The storage 301 adds a synchronization ID and an issue ID to the delayed write command and the delayed data reflection command and transmits the command. FIG. 22 is an explanatory diagram illustrating an example of a synchronization ID and an issue ID. The storage 301 updates the synchronization ID every time a resumable point notification command is received from the host 300. Therefore, the same synchronization ID is added to the delayed write command output from one resumable point to the next resumable point. However, the storage 301 adds the synchronization ID immediately before the update to the delayed data reflection command issued when the resumable point notification command is received. The updated synchronization ID is added to the delayed write command issued thereafter. In the example illustrated in FIG. 22, “24” and “25” that are the updated synchronization IDs are added to the delayed write command, and “23” that is the synchronization ID before the update is added to the delayed data reflection command that is output most recently. And “24” are added.

  The storage 301 updates the issue ID every time a delayed write command or delayed data reflection command is issued (that is, every time a delayed write command or delayed data reflection command is created and transmitted). The example shown in FIG. 22 shows an example in which the issue ID of the delayed write command or the delayed data reflection command is increased by 1.

  When the storage 301 receives a write command from the host 300, the storage 301 transmits a delayed write command to the storage 302, and then transmits data corresponding to the delayed write command. The data transmission order may not be the same as the delayed write command issue order. For example, it is assumed that the storage 301 issues a delayed write command to write data p and q, and then issues the next delayed write command to write data r. In this case, the order in which the storage 301 transmits each data is not limited to the order of p, q, r. The transmission of the data r may be started before the transmission of the data p and q is completed. However, when the delayed data reflection command is transmitted to the storage 302, the delayed data reflection command is transmitted after the transmission of the delayed write command data issued so far is completed.

  The storage 301 also adds a location (offset address, sector number, block number, etc.) and data size to which data specified by the delayed write command is written to the delayed write command.

  Next, an operation when the applications 310a and 310b write data to the storage 301 will be described. FIG. 23 is a flowchart showing the operation of the processing sequencer 321 in the storage controller 320.

  When the applications 310 a and 310 b write data to the storage 301, they output a write command to the communication unit 322 in the storage controller 320 of the storage 301. Upon receiving the write command, the communication unit 322 passes the write command to the processing sequencer 321 and instructs the processing sequencer 321 to start the write processing.

  When receiving the write command, the processing sequencer 321 secures an area necessary for storing data sent from the host 300 in the buffer memory 106 as shown in FIG. 23 (step S400). Then, the communication unit 322 is instructed to send a notification of completion of preparation to the host 300 (step S401). In response to the instruction, the communication unit 322 transmits a preparation completion notification to the host 300.

  Next, the process waits for data to arrive from the host 300 (step S402). When the data arrives and receives an inquiry from the communication unit 322 about the area of the buffer memory 106 where the data is to be stored, the communication unit 322 allocates the area secured in step S400. (Step S403). Further, the communication unit 322 is instructed to designate the synchronization ID to the storage 302 and transmit a delayed write command (step S404). The communication unit 322 transmits a delayed write command to the storage 302 in response to the instruction. As already described, the synchronization ID and the issue ID are added to the delayed write command. The same synchronization ID is added to each delayed write command output from one resumable point to the next resumable point. Also, the issuance ID is updated every time a delayed write command or delayed data reflection command is output (see FIG. 22).

  Then, the storage sequencer 321 waits for the storage of data from the host 300 to the buffer memory 106 to be completed (step S405), and is notified from the communication unit 322 that all data has been stored in the buffer memory 106. For the IO scheduler 104, the processing type (in this case, writing), the processing identification ID, information indicating the area in the storage medium 101 to be processed, and the buffer memory 106 to be processed Information indicating the area is registered. The IO scheduler 104 instructs the medium control unit 105 to write data to be written in response to the write request. The medium control unit 105 performs data writing processing from the buffer memory 106 to the storage medium 101 in accordance with the registered contents (step S406).

  Then, it waits for a preparation completion message (response to the delayed write command) to be transmitted from the storage 302 (step S407). When the communication unit 322 is notified that the preparation completion message has been received from the storage 302, the processing sequencer 321 causes the communication unit 322 to transmit the data stored in the buffer memory 106 to the storage 302 (step S408). . Thereafter, the communication unit 322 waits for a reception completion message from the storage 302 and a write completion notification from the medium control unit 105 (step S409), and the communication unit 322 receives the reception completion message from the storage 302. When the notification is received and the writing completion notification is received from the medium control unit 105, the host 300 is notified of the completion (step S410), and the processing is terminated.

  The order of data to be transmitted in step S408 may be different from the order in which the corresponding delayed write commands are issued. Assume that the storage 301 continuously receives write commands from the host 300 and issues delayed write commands continuously. In this case, the storage 301 may start transmitting data corresponding to a later delayed write command (step S408) before completing transmission of data corresponding to the previously issued delayed write command (step S408). Good.

  Next, processing when the storage 302 receives a delayed write command will be described. FIG. 24 is a flowchart showing an operation when the processing sequencer 321 in the storage controller 320 of the storage 302 receives the delayed write command.

  The processing sequencer 321 first secures an area necessary for the received data in the synchronization buffer memory 323 (step S420), and transmits a notification of reception preparation completion to the request transmission source (the storage 301 in this example). 322 is instructed. In response to the instruction, the communication unit 322 sends a reception preparation completion message to the transmission source (step S421). Next, it waits for the arrival of data (step S422), and when the data arrives and receives an inquiry about the storage location of the data from the communication unit 322, the communication unit 322 is notified of the area of the synchronization buffer memory 323 secured in step S420 ( Step S423). Then, it waits for the storage of data from the request destination to the synchronization buffer memory 323 to be completed (step S424), and when the data storage is completed, the processing based on the delayed write command is completed at the request destination. The communication unit 322 is instructed to notify this. In response to the instruction, the communication unit 322 notifies the request transmission destination that the processing based on the delayed write command has been completed (step S425). Then, the processing sequencer 321 ends the processing.

  Further, when the processing sequencer 321 writes data to the storage medium 101, the processing sequencer 321 has the same synchronization ID as the data already stored in the synchronization buffer memory 323, and overlaps even if only a part of the storage area is stored. Is controlled so as not to write an overlapping portion already stored in the buffer memory 323 for synchronization. That is, with respect to data to be written in the same area, control is performed so that data in processing based on the subsequent delayed write command becomes valid.

  For example, it is assumed that a delayed write command for instructing data writing to a certain area in the storage medium 101 reaches the storage 302, and a delayed write command for instructing writing to the same area arrives. Assume that these two delayed write commands have the same synchronization ID. In this case, it means that data is written once and rewritten over the resumable point in the normal storage 301. Therefore, the processing sequencer 321 may not hold the data written first in the same area in the synchronization buffer memory 323 of the standby storage 302. When the processing sequencer 321 newly receives a delayed write command from the storage 301, the processing sequencer 321 determines whether data to be overwritten exists in the synchronization buffer memory 323. When the data to be overwritten is specified, the data is deleted from the synchronization buffer memory 323.

  Next, the operation of resumable point notification processing in which the applications 310a and 310b notify the storage 301 of resumable point notification will be described. Here, a case where the application 310a performs resumable point notification processing is taken as an example.

  When the application 310a notifies the resumable point, the application 310a instructs the IO management unit 311 to notify the resumable point. When notification of resumable points is instructed from the application 310 a in the IO management unit 311, the resumable point notification unit 312 issues a resumable point notification command to the storage 301.

  When the resumable point notification command arrives at the storage 301, the resumable point notification command is input to the communication unit 322 in the storage controller 320 of the storage 301. Upon receiving the resumable point notification command, the communication unit 322 passes the resumable point notification command to the processing sequencer 321 and instructs the start of the resumable point notification process.

  FIG. 25 is a flowchart showing the operation of the processing sequencer 321. The processing sequencer 321 first holds the value of the internal synchronization ID at that time, and then updates the value of the synchronization ID (step S440). Holding the value of the synchronization ID is, for example, saving it in a register. Next, the communication unit 322 is instructed to transmit the delayed data reflection command specifying the synchronization ID held in step S440 to the storage 302. In response to the instruction, the communication unit 322 transmits a delay data reflection command to the storage 302 (step S441). Then, it waits for a message indicating completion of processing based on the delayed data reflection command from the storage 302 (step S442). When the communication unit 322 notifies that the completion message has arrived from the storage 302, the host 300 is notified of the completion of the processing based on the resumable point notification command via the communication unit 322 (step S443), and the processing ends. .

  Note that, as shown in FIG. 22, an updated synchronization ID is added to each delayed write command output after transmission of one delayed data reflection command instead of the synchronization ID stored in the register or the like.

  Next, an operation when a delayed data reflection command of the storage 302 is received will be described. When the delayed data reflection command arrives at the storage 302, the delayed data reflection command is input to the communication unit 322 of the storage controller 320 of the storage 302. Upon receiving the delay data reflection command, the communication unit 322 passes the delay data reflection command to the processing sequencer 321 and instructs the start of the delay data reflection process.

  FIG. 26 is a flowchart showing the operation of the delay data reflection process of the process sequencer 321 in the storage controller 320 of the storage 302. The processing sequencer 321 determines whether the value of the synchronization ID added to the arrived delayed data reflection command is the next value of the previously processed delayed data reflection command (step S460). If it is not the next value, it waits for the arrival of the delayed data reflection command whose synchronization ID is the next value of the synchronous ID of the previously processed delayed data reflection command. For example, in the example shown in FIG. 22, it is assumed that the previously processed delayed data reflection command is the delayed data reflection command with the synchronization ID “23”. Thereafter, when the delay data reflection command with the synchronization ID “25” is received, the arrival of the delay data reflection command with the synchronization ID “24” is awaited. Then, after performing the processing of steps S461 to S468 for the delayed data reflection command with the synchronization ID “24”, the processing after step S461 is performed for the delayed data reflection command with the synchronization ID “25”.

  If the synchronization ID is the next value of the previous synchronization ID, the value of the issued ID of the delayed data reflection command processed last time and the issue ID added to the delayed data reflection command processed this time are stored. A search is performed to determine whether all data corresponding to the delayed write command having an issue ID between values is recorded in the synchronization buffer memory 323 (step S461). For example, among the commands shown in FIG. 22, it is assumed that the issue ID of the delayed data reflection command processed last time is “71” and the issue ID added to the delayed data reflection command processed this time is “76”. In this case, it is confirmed whether or not all the data corresponding to the delayed write commands having the issue IDs “71” to “75” are recorded in the synchronization buffer memory 323. However, data deleted as data to be overwritten does not have to be included in the recording confirmation target. If there is omission, it waits until all data corresponding to each delayed write command arrives at the standby storage 302 and is recorded in the synchronization buffer memory 323.

  In step S463, the processing sequencer 321 searches the synchronization buffer memory 323, and searches for a delayed write command that matches the synchronization ID specified by the delayed data reflection command and is not being written. For example, when the delayed data reflection command with the issue ID “76” shown in FIG. 22 is received and processed, the delayed write that matches the synchronization ID “24” added to the command and is not being written is written. Search for commands. If it is not found as a result of the search, there is no delayed write command for which data write processing to the storage medium 101 has not started. In this case, the process proceeds to step S465. When a delayed write command to be searched is found, there is a delayed write command for which data write processing to the storage medium 101 has not started. In this case, the process proceeds to step S464.

  In step S464, an instruction to write the corresponding data in the synchronization buffer memory 323 to the location (offset address) where the data on the storage medium 101 specified by the delayed write command found in the search in step S462 is to be written ( Write request) is registered in the IO scheduler 104. In response to a write request, the IO scheduler 104 instructs the medium control unit 105 to write data to be transferred. The medium control unit 105 outputs data from the synchronization buffer memory 323 to the storage medium 101 in accordance with the write instruction. Then, the processing sequencer 321 sets the searched area in the synchronization buffer memory 323 as a writing process, and returns to Step S462.

  In step S465, it is determined whether data output processing is being performed from the synchronization buffer memory 323 to the storage medium 101. If the process is in progress, the process proceeds to step S466. If the process has already been completed, the process proceeds to step S468. In step S466, the process waits until one of the data output processes to the storage medium 101 is completed, the area in the synchronization buffer memory 323 that is the target of the completed process is set to an unused state (step S467), and the process proceeds to step S465. Return. In step S468, the communication unit 322 is instructed to notify the transmission source of the delayed data reflection command (storage 301 in this example) of the completion of processing based on the delayed data reflection command. In response to the instruction, the communication unit 322 notifies the transmission source of the delayed data reflection command of the completion of processing. Also, the processing sequencer 321 updates the synchronization ID and issue ID held as information of the previously processed delayed data reflection command to those of the processed delayed data reflection command, and ends the processing.

  Next, the operation when a disaster occurs will be described. When the host 300 becomes unusable due to a disaster or the like, the storage 302 is changed from the standby system to the normal system. Further, the host 303 takes over processing from the host 300.

  The operation of the host 303 from failure detection to application restart will be described. First, the host monitoring unit 313 in the host 303 detects an abnormality of the host 300. When the host monitoring unit 313 detects an abnormality of the host 300, the host monitoring unit 313 issues a delayed data discard command to the storage 302 that is a standby system. Then, waiting for a response to the delayed data discard command, and receiving the response, the host monitoring unit 313 of the standby host 303 executes the applications 310a and 310b of the standby host. Thereafter, data writing and reading from the host 303 are performed on the storage 302.

  Note that the host monitoring unit 313 in the host 303 communicates with the host 300 constantly or periodically. When the host monitoring unit 313 cannot communicate with the host 300 for a certain period of time or when an abnormality is reported from the host 300, the host monitoring unit 313 recognizes that a disaster has occurred in the host 300. In order for the host 303 to reliably recognize the abnormality of the host 300, the host 303 and the host 300 are preferably connected by a dedicated line.

  Next, an operation when the standby storage 302 receives a delayed data discard command from the standby host 303 will be described. When the delayed data discard command arrives at the storage 302, the delayed data discard command is input to the communication unit 322 in the storage controller 320 of the storage 302. Upon receiving the delayed data discard command, the communication unit 322 passes the delayed data discard command to the processing sequencer 322 and instructs the start of the delayed data discard process.

  FIG. 27 is a flowchart showing the delayed data discarding process executed by the processing sequencer 322. The processing sequencer 322 searches the synchronization buffer memory 323 (step S480). In step S480, data that does not need to be moved to the storage medium 101 is searched because the delayed data reflection command has not yet arrived. As a result of the search, the processing sequencer 322 proceeds to step S483 if it is not found, and proceeds to step S482 if found (step S481). For example, among the commands shown in FIG. 22, the delayed data reflection command of issue ID “76” has not been received, and the data of some delayed write commands of issue IDs “72” to “75” are synchronized. If stored in the buffer memory 323, the process proceeds to step S482. If the delayed data reflection command with the issue ID “76” has been received and the data of each delayed write command has been transferred to the storage medium 101, the process advances to step S483.

  In step S482, the processing sequencer 322 makes the area in the synchronization buffer memory 323 in which the data found in step S480 is recorded unused, and returns to step S480. Then, in the next step S481, the process proceeds to step S483. In step S483, if the delayed data reflection process is being performed, the process proceeds to step S484. If the delayed data reflection process is not being performed, the process proceeds to step S485.

  In step S484, the delayed data reflection process being processed is completed, and if completed, the process proceeds to step S485. In step S485, the communication unit 322 is instructed to notify the issuer of the delayed data discard command (host 303 in this example) of the completion of the processing based on the delayed data discard command. In response to the instruction, the communication unit 322 notifies the issuer of the delayed data discard command of the completion of the processing, and the processing sequencer 322 ends the processing.

  In this embodiment, when the normal host 300 outputs a write command to the storage 301, the standby storage 302 receives the data written from the storage 301, but does not record it in the storage medium 101, but for synchronization. Records in the buffer memory 323. When the storage 302 receives the resumable point notification command, the storage 302 moves the data from the synchronization buffer memory 323 to the storage medium 101. When the storage 302 receives the delayed data discard command, the storage 302 does not use the synchronization buffer memory 323. Put it in a state. Accordingly, the storage medium 101 of the storage 302 is always kept in a state where the processing of the applications 310a and 310b can be resumed. As a result, when an abnormality occurs in the host 300, the host 303 can immediately continue the processing.

  In addition, the normal storage 301 does not transmit the data written in the storage medium 101 between the resumable point and the resumable point to the standby storage 302 at a time, but transmits it at every timing when the write command is received. . Therefore, since a large amount of data is not transmitted at a time, the data transfer time with the host 302 can be reduced.

  Further, each command received by the standby storage 302 from the storage 301 is managed by the synchronization ID and the issue ID, and the arrival order of the data transmitted by the normal storage between the resumable point and the resumable point is in an arbitrary order. Good. Therefore, since the order in which the delayed write command transmits data corresponding to the delayed write command is not limited, the design is facilitated.

  In the present embodiment, the case where both the synchronization ID and the issue ID are added to the delayed write command and the delayed data reflection command has been described. However, only the issue ID may be added. In this case, when the processing sequencer 321 of the standby storage 302 receives the delayed data reflection command, the processing sequencer 321 may perform the following processing instead of steps S460 and S461. When receiving the delayed data reflection command, the processing sequencer 321 receives all the commands to which the issue ID between the issued ID of the delayed data reflection command processed last time and the issued ID of the received delayed data reflection command is added. Check if it is. If all commands to which each issue ID is added have been received, the processing after step S462 (see FIG. 26) is performed. If not all commands to which the issue IDs are added have been received, the processing after step S462 is started when all the commands are received.

  For example, it is assumed that, among the commands shown in FIG. 22 (assuming that no synchronization ID is added), the delayed data reflection command with the issue ID “71” is the delayed data reflection command processed last time. After that, when the delayed data reflection command with the issue ID “76” is received, it is confirmed whether or not each command with the issue IDs “72” to “75” has been received, and after receiving all these commands, step S462. Subsequent processing is started. When the delayed data reflection command with the issue ID “80” is received before the issue ID “76”, the control sequencer 321 determines whether or not each command with the issue IDs “72” to “79” has been received. Check. When all commands with issue IDs “72” to “76” are received, processing of a delayed data reflection command with issue ID “76” is performed. Thereafter, when all the commands with the issue IDs “77” to “79” are received, the delayed data reflection command with the issue ID “80” is processed.

  If a synchronization ID is also added to each command, the data to be overwritten can be specified when a delayed write command is received. On the other hand, when the synchronization ID is not used, the processing sequencer 321 specifies data to be overwritten immediately before starting the processing of step S462, and deletes the data from the synchronization buffer memory 323.

Embodiment 8 FIG.
FIG. 28 is a block diagram showing an eighth embodiment of a data replication system according to the present invention. In this embodiment, data is transferred mainly by the storage, and data in the storage in the normal system is mirrored to a remote location. Further, the system is provided with a standby storage capable of storing a snapshot of at least one generation before. The snapshot one generation before is information on the address of the storage medium 101 in which data was stored at the latest resumable point.

  In the data replication system shown in FIG. 28, a storage 400 is locally connected to a host 300 that uses the storage 400. Further, the storage 401 is locally connected to the host 303 that uses the storage 401. The storage 301 is connected to the storage 302 via the network 13. Further, the host 300 and the host 303 are communicably connected so that the host 303 monitors the state of the host 300. The host 300 and the host 303 are preferably connected by a dedicated line, but may be connected by a network other than the dedicated line (for example, the Internet). The configurations of the hosts 300 and 303 are the same as the configurations of the hosts 300 and 303 in the seventh embodiment (see FIG. 20).

  The storages 400 and 401 are, for example, a single magnetic disk device, an optical disk device, or a magneto-optical disk device. As the storages 400 and 401, a single magnetic disk device, an optical disk device, or a disk array device that is a set of magneto-optical disk devices can be used. The hosts 300 and 303 and the storages 400 and 401 are connected by SCSI, Fiber channel, Ethernet (registered trademark), or the like. In the system shown in FIG. 28, the host 300 is a normal host that operates when there is no failure in the system, and the host 303 is a standby host that operates when a failure occurs in the host 300. Suppose there is.

  In this embodiment, the applications 310 a and 310 b running on the host 300 are points that can be resumed in the storage 301 as long as the applications can resume operation if they are in the data state. In order to notify, resumable point notification processing is performed. Note that the operations of the host 300 and the storage 400 when the applications 310a and 310b read data from the storage 400 are the same as the operations in the normal data read processing. In the storage 400, the operations of the IO scheduler 104 and the medium control unit 105 are the same as the operations of the IO scheduler 104 and the medium control unit 105 in the first embodiment.

  The applications 310a and 310b in the present embodiment are applications having a resume function. That is, it is an application that can resume processing if the data recording state of the storage medium 101 of the storage is in a predetermined state.

  FIG. 29 is a block diagram showing a configuration example of the storage 400 shown in FIG. The configuration of the storage 401 is also as shown in FIG. As shown in FIG. 29, the storage 400 includes a storage controller 410 and a storage medium 101 that is a storage body. The storage controller 410 is issued by a communication unit 412 that communicates with the host 300 and other storages, a processing sequencer 411 that manages the sequence of each process, an IO scheduler 104 that controls the order of processing instructions for the storage medium 101, and an IO scheduler 104 A medium processing unit 105 that controls the operation of the storage medium 101 in accordance with a processing instruction to be performed; a buffer memory 106 that temporarily stores data from the host 300 to the storage medium 101 and data from the storage medium 101 to the host 300; It includes an LBA management unit 413 to perform, and an address table storage unit 414 that stores an address table used for managing logical block addresses. The processing sequencer 321 is realized by, for example, a CPU that operates according to a program.

  A semiconductor memory may be used as the synchronization buffer memory 322, or a larger capacity storage device such as a magnetic disk device, an optical disk device, or a magneto-optical disk device may be used.

  The storages 400 and 401 manage data according to the physical address space of the storage medium 101. On the other hand, when receiving a request for data write processing, data read processing, or the like from another device, a write area or a read area is designated by an address in the logical address space. The logical address space at the resumable point is called a snapshot logical address space. In addition, the logical address at the latest time point is called the latest logical address space. The storages 400 and 401 proceed with the processing using the snapshot logical address space information and the latest logical address space information. The storages 400 and 401 may manage a plurality of snapshot logical address spaces.

  An example of the address table stored in the address table storage unit 414 is shown in FIG. The address table 420 includes a plurality of entries 420-0 to 420-n. Each entry 420-0 to 420-n corresponds to a logical block obtained by dividing the logical address space into fixed-length data.

  FIG. 31A is an explanatory diagram of entries. Information constituting each entry 420-0 to 420-n includes a physical block number in the storage corresponding to the latest logical address space and a physical block number corresponding to the snapshot logical address space. Each physical block number is set with a flag for recording that the block number has changed since the most recent snapshot was created. The physical block number is a unique number assigned to each physical block when the address space of the storage 400 or 401 is divided into fixed-length data. It is not necessary to provide a physical block number flag corresponding to the snapshot logical address space.

  FIG. 31B shows an example of the initial state of one entry. FIG. 31B shows that data is stored at an address corresponding to the physical block number “aaa”. Since the data has not been changed since the resumable point, the physical block number corresponding to the latest logical address space is also “aaa”. Assume that a write to an address corresponding to the physical block number “aaa” is requested. In this case, writing is performed to another address, and the physical block number “bbb” corresponding to the address is held as the physical block number corresponding to the latest logical address space. In addition, the flag corresponding to the new logical address space is also updated to “changed”. Thereafter, when a resumable point is reached, information corresponding to the latest logical address space is copied to information corresponding to the snapshot logical address space as shown in FIG. Further, the flag corresponding to the latest logical address space is set to “no change”.

  In this embodiment, information corresponding to the snapshot logical address space is used as a snapshot.

  As a result, until the next resumable point, the data to be newly written is written to an address different from the address to be written. Accordingly, data at the point of the resumable point remains. By managing the writing of data to the standby storage in this way, even if an abnormality occurs in the host 300, the host 303 can immediately resume processing.

  The address table storage unit 414 also stores an unused block table in which block numbers of physical blocks not included in any logical address space are recorded. As the address table storage unit 414, a nonvolatile semiconductor memory, a magnetic disk device, an optical disk device, a magneto-optical disk, or the like is used. In addition, a part of the storage area of the storage medium 101 may be used as the address table storage unit 414.

  Next, an operation when the host 300 writes data to the storages 400 and 401 will be described. When writing data to the storage 400, the host 300 outputs a write command to the communication unit 412 in the storage controller 410 of the storage 400. Upon receiving the write command, the communication unit 412 passes the write command to the processing sequencer 411 and instructs the processing sequencer 411 to start the write processing. The writing process executed by the processing sequencer 411 differs depending on whether or not the data transfer destination storage is set. In the storage 400, the storage 401 is set as the data transfer destination, and no data transfer destination is set in the storage 401.

  FIG. 32 is a flowchart showing the operation of the processing sequencer 411 in the storage controller 410 when the data transfer destination is set. In other words, this embodiment is a flowchart showing the operation of the processing sequencer 411 in the storage 400. If the data transfer destination is set, the processing sequencer 411 first secures an area necessary for the received data in the synchronization buffer memory 323 (step S500). In addition, the logical address of the write destination specified by the write command and the write processing are specified, and the LBA management unit 413 converts the logical address into a physical address (step S501). The LBA management unit 413 assigns a physical block number corresponding to the latest logical address space shown in FIG. 31 and returns a physical address corresponding to the physical block number to the processing sequencer 411. Details of the operation of the LBA management unit 413 in step S501 will be described later.

  Next, the processing sequencer 411 instructs the communication unit 412 to send a notification of completion of preparation to the host 300 (step S502). The communication unit 412 transmits a notification of preparation completion to the host 300 in response to the instruction. Then, it waits for data to arrive from the host 300 (step S503), and when the data arrives and receives an inquiry from the communication unit 412 about the area of the buffer memory 106 where the data is to be stored, the communication unit 412 identifies the area secured in step S500. (Step S504). If the resumable point notification process is being performed, the process proceeds to step S513, and if not, the process proceeds to step S506 (step S505). Here, “when resumable point notification processing is performed” refers to a case where a resumable point is notified from the host 300 and predetermined processing is performed on the storage 401. Specifically, if processing in steps S562 to S564 described later is performed, the process proceeds to step S513, and if not, the process proceeds to step S506.

  In step S506, the processing sequencer 411 instructs the communication unit 412 to issue a write command to the data transfer destination storage (storage 401 in this example). This write command specifies the write destination with a logical address. The communication unit 412 transmits a write command to the data transfer destination storage in response to the instruction. Then, the storage sequencer 411 waits for completion of storage of data from the host 300 in the buffer memory 106 (step S507). When the communication unit 412 notifies that all data has been stored in the buffer memory 106, the processing sequencer 411 An instruction (write request) for writing the data stored in the buffer memory 106 is registered in the IO scheduler 104 in the area corresponding to the physical address converted in step S501. The IO scheduler 104 instructs the medium control unit 105 to write data to be written in response to the write request. The medium control unit 105 performs a data writing process from the buffer memory 106 to the storage medium 101 in accordance with the registered content (step S508).

  The storage unit 401 waits for a preparation completion message to be transmitted from the storage 401 (step S509). When the communication unit 412 notifies that the preparation completion message has been received from the storage 401, the communication unit 412 is notified of the buffer memory. The data stored in 106 is transmitted to the storage 401 (step S510). After that, a reception completion message is transmitted from the storage 401 and a write completion notification from the medium control unit 105 is awaited (step S511), and the reception completion message from the storage 401 is received from the communication unit 412. When the notification is received and the writing completion notification is received from the medium control unit 105, the host 300 is notified of the completion (step S512), and the processing is terminated.

  In step S513, the processing sequencer 411 waits for the completion of storage of data from the host 300 in the buffer memory 106, and when the communication unit 412 notifies that all the data has been stored in the buffer memory 106, the processing sequencer 411 For 104, an instruction (write request) to write the data stored in the buffer memory 106 is registered in the area corresponding to the physical address converted in step S501 (step S514). Next, the process waits until the resumable point notification process is completed (step S515). When the resumable point notification process is completed, the communication unit 412 is instructed to issue a write command to the data transfer destination storage (step S516). ). This write command specifies the write destination with a logical address. After step S516, the process proceeds to step S509.

  Note that, in the processes in steps S506 to S508 and the processes in steps S513 to S516, the order of the write command issuance process for the storage 401 and the write process to the storage medium 101 is reversed. This is due to the following reason. In steps S506 to S508, the write command is transmitted to the storage 401 and the write process to the storage medium 101 is advanced during the period until the response is returned. I decided to do it first. On the other hand, in steps S513 to S516, the entry is changed in the storage 401 while waiting for completion of the resumable point notification process. Therefore, the write command cannot be transmitted to the storage 401 until the resumable point notification process is completed. Therefore, the writing process to the storage medium 101 is advanced until the end of the resumable point notification process is awaited.

  FIG. 33 is a flowchart showing the operation of the processing sequencer 411 in the storage controller 410 when the data transfer destination is set. In other words, this embodiment is a flowchart showing the operation of the processing sequencer 411 in the storage 401. More specifically, it is a flowchart showing the operation of the processing sequencer 411 of the storage 401 that has received the write command transmitted in steps S506 and S516.

  When receiving the write command, the processing sequencer 411 first secures an area necessary for the received data in the buffer memory 106 (step S520). Also, the logical address of the write destination specified by the write command and the write processing are specified, and the LBA management unit 413 converts the logical address into a physical address (step S521). This process is the same as step S501 (see FIG. 32).

  Next, the processing sequencer 411 instructs the communication unit 412 to send a notification of preparation completion to the data transfer source storage (storage 400 in this example) (step S522). In response to the instruction, the communication unit 412 transmits a preparation completion notification to the data transfer source storage. Then, it waits for data to arrive from the data transfer source storage (step S523). When the data arrives and receives an inquiry from the communication unit 412 about the area of the buffer memory 106 where the data is to be stored, the area secured in step S520 is obtained. The communication unit 412 is notified (step S524).

  Further, it waits for the storage of data from the data transfer source storage to the buffer memory 106 (step S525), and when the communication unit 412 notifies that all the data has been stored in the buffer memory 106, the processing sequencer 411 registers an instruction (write request) for writing the data stored in the buffer memory 106 in the area corresponding to the physical address converted in step S <b> 521 to the IO scheduler 104. The IO scheduler 104 instructs the medium control unit 105 to write data to be written in response to the write request. The medium control unit 105 performs data writing processing from the buffer memory 106 to the storage medium 101 according to the registered contents (step S526).

  Then, it waits for a write completion notification from the medium control unit 105 (step S527). When the write completion notification is received from the medium control unit 105, it notifies the data transfer source storage (step S528) and ends the processing. .

  The storages 400 and 401 both receive a write command with a specified logical address, and start a write process. However, data is not written to a physical address corresponding to the logical address, but data is written to another physical address (the physical address assigned in steps S501 and S521). Therefore, since the data at the resumable point can be kept in the storage medium 101, even if an abnormality occurs in the normal host 300, the standby host 303 can restart the processing in a short time.

  Next, conversion processing from a logical address to a physical address by the LBA management unit 413 will be described. Note that the conversion process by the LBA management unit 413 differs depending on whether the write process is designated or not. If the write process is not designated, the LBA management unit 413 calculates a logical block number from the logical address when the logical address is designated and receives an instruction for the conversion process, and the address table storage unit 414 The physical block number corresponding to the logical block number in the latest logical address space is acquired. Further, the physical address is calculated from the physical block number and the logical block address, and the calculated physical address is notified to the processing sequencer 411. The logical block number is a value obtained by dividing the logical block address by the block length (the decimal part is rounded down). When calculating a physical address, first, a product of a physical block and a block length (referred to as P) is obtained. Further, the remainder (Q) of the value obtained by dividing the logical address by the block length is obtained. The sum of P and Q is calculated as a physical address.

  FIG. 34 is a flowchart showing the operation of the LBA management unit 413 when it is specified that the write process is performed. That is, it is a flowchart showing the operation of the LBA management unit 413 in steps S501 and S521. If it is specified that the process is write processing, the LBA management unit 413 calculates a logical block number from the logical address (step S540), and then the logical block number calculated in step S540 from the address table storage unit 414. The entry corresponding to is acquired (step S541). In step S541, an entry in the latest logical address space is acquired. Then, a flag paired with the physical block number corresponding to the latest logical address space in the entry is confirmed (step S542). If the flag is “changed”, the process proceeds to step S546. If the flag is “not changed”, the process proceeds to step S543.

  When the flag is “no change”, as shown in FIG. 31B, the physical block number corresponding to the latest logical address space is equal to the physical block number corresponding to the snapshot logical address space. is there. Hereinafter, description will be made with reference to FIGS. In step S543, the LBA management unit 413 obtains an unused physical block number from the unused block table in the address table storage unit 414 (step S543). Here, it is assumed that a physical block number “bbb” is obtained. Further, the acquired physical block number information is deleted from the unused block table information registered in the address table storage unit 414. Then, the unused block table information after deletion is registered again in the address table storage unit 414.

  Further, the data in the area corresponding to the physical block (aaa) corresponding to the latest logical address space in the entry obtained in step S541 is copied to the area corresponding to the physical block number (bbb) obtained in step S543. Information (copy request) is registered in the IO scheduler 104. The IO scheduler 104 instructs the medium control unit 105 to copy the target data in response to the copy request. The medium control unit 105 performs a copy process between designated areas in the storage medium 101 (step S544). When the copy process is completed, the LBA management unit 413 changes the physical block number (aaa) corresponding to the latest logical address space of the entry obtained in step S541 to the physical block number (bbb) obtained in step S543, and after the change The flag set with the physical block number is updated to “changed”. Further, the changed entry is stored in the address table storage unit 414 (step S545). Then, control goes to a step S546.

  In step S546, the LBA management unit 413 calculates a physical address from the physical block number corresponding to the latest logical address space of the entry, notifies the calculated physical address to the processing sequencer 411 (step S547), and ends the process.

  After step S542, when the processing after step S543 is performed, the physical block numbers shown in FIG. 31B are changed as shown in FIG. However, when the processing up to step S547 is completed, the same data as the physical address corresponding to “aaa” is stored in the physical address corresponding to “bbb”. The normal storage 400 rewrites the data at the physical address corresponding to “bbb” in step S508 or step S514 (see FIG. 32). The standby storage 401 rewrites the data of the physical address corresponding to “bbb” in step S526 (see FIG. 33).

  Next, an operation when the applications 310a and 310b notify the storage 400 of a resumable point notification will be described. When the applications 310 a and 310 b notify the storage 400 of a resumable point notification, the applications 310 a and 310 b instruct the IO management unit 311 (see FIG. 20) to notify the resumable point. Then, the resumable point notification unit 312 in the IO management unit 311 issues a resumable point notification command to the storage 400.

  When the resumable point notification command arrives at the storage 400, the resumable point notification command is input to the communication unit 412 in the storage controller 410. Upon receiving the resumable point notification command, the communication unit 412 passes the resumable point notification command to the processing sequencer 411 and instructs the processing sequencer 411 to start the resumable point notification process.

  FIG. 35 is a flowchart showing the operation of the processing sequencer 411 that has received the resumable point notification command. When the process sequencer 411 receives the resumable point notification command, as shown in FIG. 35, the process sequencer 411 first checks whether there is any uncompleted write process requested to the other storage (step S560). If there is, the process is executed in step S561. That is, it waits for all the write processing requested to the other storage to be completed (step S561). The processing sequencer 411 manages list information indicating whether or not a response to each issued write command has been completed. In step S560, a determination may be made based on this information.

  The processing sequencer 411 sends the snapshot creation command (snapshot creation request) to the data transfer destination storage (storage 401 in this example) to the communication unit 412 in a state where all the write processing requested to the other storage has been completed. ) To send. In response to the instruction, the communication unit 412 transmits a snapshot creation command to the data transfer destination storage (step S562). The snapshot creation command copies information corresponding to the latest logical address space to information corresponding to the snapshot logical address space, and updates the flag corresponding to the latest logical address space to “no change”. It is a command to request After step S562, the processing sequencer 411 waits for a response from the transfer destination storage (step S563). When the communication unit 412 notifies that the response has arrived, the processing sequencer 411 notifies the communication unit 412 of completion to the host 300. (Step S564), and the process ends.

  Next, the operation when the storage 401, which is the data transfer destination storage, receives the snapshot creation command will be described. When the snapshot creation command arrives at the storage 401, the snapshot creation command is input to the communication unit 412 of the storage controller 410 of the storage 401. Upon receiving the snapshot creation command, the communication unit 412 passes the snapshot creation command to the processing sequencer 411 and instructs the start of the snapshot creation processing.

  In the snapshot creation process, the process sequencer 411 performs the following process for all entries in the address table in the address table storage unit 414. That is, if “changed” is recorded in the flag corresponding to the latest logical address space, the physical block number corresponding to the snapshot logical address space of the entry is stored in the address table storage unit 414 as an unused block. Register in the unused block table. Further, the physical block number corresponding to the latest logical address space is copied to the block number corresponding to the snapshot logical address space. At this time, the flag value is also copied. Thereafter, the flag corresponding to the latest logical address space is initialized to “no change”.

  For example, in the entry shown in FIG. 31C, “changed” is recorded in the flag corresponding to the latest logical address space. In this case, the physical block number “aaa” corresponding to the snapshot logical address space is registered in the unused block table as an unused block. Then, the physical block number “bbb” corresponding to the latest logical address space is copied to the block number corresponding to the snapshot logical address space. Then, the flag “changed” corresponding to the latest logical address space is also copied. Further, the flag corresponding to the latest logical address space is initialized to “no change”. Then, the entry state changes from the state shown in FIG. 31C to the state shown in FIG.

  If the flag corresponding to the latest logical address space is “no change”, it means that the data stored in the address corresponding to the physical block number has not been changed since the immediately preceding resumable point. Therefore, no processing is performed on the entry.

  After completing the processing for all entries for which the flag corresponding to the latest logical address space is “changed”, the processing sequencer 411 transmits a response to the snapshot creation command using the communication unit 412.

  Next, the operation when a disaster occurs will be described. When the host 300 becomes unusable due to a disaster or the like, the storage 401 is changed from the standby system to the normal system. Further, the host 303 takes over processing from the host 300.

  The operation of the host 303 from failure detection to application restart will be described. First, the host monitoring unit 313 (see FIG. 20) in the host 303 detects an abnormality of the host 300. When the host monitoring unit 313 detects an abnormality in the host 300, it issues a snapshot restoration command to the storage 401 that is the standby system. The host monitoring unit 313 then executes the applications 310a and 310b when waiting for a response to the snapshot restoration command and receiving the response. Thereafter, writing and reading of data from the host 303 is performed on the storage 401.

  Note that the host monitoring unit 313 in the host 303 communicates with the host 300 constantly or periodically. When the host monitoring unit 313 cannot communicate with the host 300 for a certain period of time or when an abnormality is reported from the host 300, the host monitoring unit 313 recognizes that a disaster has occurred in the host 300.

  Next, an operation when the storage 401 receives a snapshot restoration command will be described. When the snapshot restoration command arrives at the storage 401, the snapshot restoration command is input to the communication unit 412 in the storage controller 410 of the storage 401. Upon receiving the snapshot restoration command, the communication unit 412 passes the snapshot restoration command to the processing sequencer 411 and instructs the start of the snapshot restoration process.

  In the snapshot restoration process, the process sequencer 411 performs the following process on all entries in the address table 414 in the address table storage unit. That is, when “changed” is recorded in the flag corresponding to the latest logical address space, the unused block table in the address table storage unit 414 is set to the physical block number corresponding to the latest logical address space as an unused block. Register with. As a result, the area corresponding to the physical block number is released as an unused state. Next, the storage information (physical block number and flag corresponding to the latest logical address space) indicating the data storage status in the storage medium is returned to the state at the time of the previous snapshot creation. That is, the physical block number corresponding to the snapshot logical address space is copied to the block number corresponding to the latest logical address space. Further, the flag corresponding to the latest logical address space is changed from “changed” to “not changed”.

  For example, assume that an entry is in the state shown in FIG. The flag corresponding to this latest logical address space is “changed”. Accordingly, the processing sequencer 411 registers the physical block number “bbb” corresponding to the latest logical address space as an unused block in the unused block table. Next, the physical block number corresponding to the snapshot logical address space is copied to the block number “bbb” corresponding to the latest logical address space. As a result, the block number corresponding to the latest logical address space is “aaa”. Further, the flag corresponding to the latest logical address space is changed from “changed” to “not changed”. As a result, the storage information (physical block number and flag corresponding to the latest logical address space) returns to the state at the time of the previous snapshot creation. Thus, by performing the snapshot restoration process, the entry is restored to the state at the immediately preceding resumable point.

  If the flag corresponding to the latest logical address space is “no change”, it means that the data stored in the address corresponding to the physical block number has not been changed since the immediately preceding resumable point. Therefore, no processing is performed on the entry.

  In the present embodiment, the standby storage 401 stores the data received from the storage 400 at an address corresponding to the newly assigned physical block number. When the storage 401 receives the snapshot creation command at the resumable point, the storage 401 holds the information of the newly allocated physical block as a snapshot. In addition, when an abnormality occurs in a normal host before the resumable point, the newly assigned physical block number is set to an unused state. If the newly assigned physical block number is set to an unused state, the host 303 can resume the processing of the applications 310a and 310b. That is, even if an abnormality occurs in the host 300, the time until the host 303 resumes the process can be short.

  Further, the normal storage 400 does not transmit the data written in the storage medium 101 between the resumable point and the resumable point to the standby storage 401 at a time, but at each timing when the write command is received from the host 300. Send to. Therefore, since a large amount of data is not transmitted at a time, the data transfer time with the host 401 can be reduced.

  Next, a modification of the eighth embodiment will be described. In the processing examples shown in FIGS. 32 and 35, when the resumable point notification processing has been performed, the issue of the write command is awaited (see step S505). If there is a write process that has not been completed, a snapshot creation command is awaited (see step S560). On the other hand, in this modification, when the storage 400 receives a write command from the host 300, the storage 400 transmits the write command to the storage 401 regardless of the state of the resumable point notification process. That is, immediately after step S504 shown in FIG. 32, the processing after step S506 is started. Further, when the storage 400 receives a resumable point notification from the host 300, a snapshot creation command is immediately issued to the storage 401 regardless of the status of the write process. That is, when the resumable point notification is received, the processing after step S562 shown in FIG. 35 is immediately started.

  In this modification, the storage 400 adds an issue ID to all commands transmitted to the storage 401. The standby storage 401 holds processed issue ID information. The processed issue ID information is information indicating up to which command of each command received from the storage 400 has been processed. For example, if the content of the processed issue ID information is “53”, it means that the processing for the commands up to the issue ID “53” has been completed.

  FIG. 36 is a flowchart showing the operation of the processing sequencer 411 of the storage 401 that has received the write command in the modification of the eighth embodiment. The processing sequencer 411 first secures an area necessary for received data in the buffer memory 106 (step S580). In addition, the logical address of the write destination specified by the write command and the write processing are specified, and the LBA management unit 413 converts the logical address into a physical address (step S581). In step S581, the LBA management unit 413 performs processing on the entry in the same manner as in step S521, and passes the physical address to the processing sequencer 411.

  Next, the processing sequencer 411 instructs the communication unit 412 to send a notification of completion of preparation to the data transfer source storage (storage 400 in this example) (step S582). In response to the instruction, the communication unit 412 transmits a preparation completion notification to the data transfer source storage. Then, it waits for data to arrive from the data transfer source storage (step S583). When the data arrives and receives an inquiry from the communication unit 412 about the area of the buffer memory 106 where the data is to be stored, the area secured in step S580 is obtained. The communication unit 412 is notified (step S584).

  Further, it waits for the storage of data from the data transfer source storage to the buffer memory 106 (step S585), and when the communication unit 412 notifies that all the data has been stored in the buffer memory 106, the processing sequencer Step 411 waits until the value of the processed issuance ID information becomes the previous value of the issuance ID added to the write command to be processed (step S586). For example, it is assumed that the write command with the issue ID “54” is received and the operations from step S580 to step S585 are performed. At this time, each command (write command, snapshot creation command, etc.) up to the issue ID “53” is not necessarily received. In this case, in step S586, each command up to the issue ID “53” is received, and the process for the issue ID “54” is interrupted until all the processes corresponding to each command are completed. When the processed issue ID information becomes “53”, the process for the issue ID “54” is resumed, and the process proceeds to step S587.

  When the value of the processed issuance ID information becomes the previous value of the issuance ID, the processing sequencer 411 stores the IO scheduler 104 in the buffer memory 106 in an area corresponding to the physical address converted in step S581. An instruction (write request) for writing the read data is registered. The IO scheduler 104 instructs the medium control unit 105 to write data to be written in response to the write request. The medium control unit 105 writes data from the buffer memory 106 to the storage medium 101 according to the registered contents (step S587).

  Then, it waits for a write completion notification from the medium control unit 105 (step S588), and when it receives a write completion notification from the medium control unit 105, it notifies the data transfer source storage of the completion (step S589). In step S589, the issue ID added to the write command is reflected in the processed issue ID information. For example, if the processing up to step S589 is performed for the write command of the issue ID “54”, the information of the processed issue ID information is updated from “53” to “54”.

  In the modification of the eighth embodiment, when the storage 401 receives a snapshot creation command, the operation is as follows. That is, when a snapshot creation command arrives at the storage 401, the snapshot creation command is input to the communication unit 412 of the storage controller 410 of the storage 401. Upon receiving the snapshot creation command, the communication unit 412 passes the snapshot creation command to the processing sequencer 411 and instructs the start of the snapshot creation processing.

  In the snapshot creation process, the process waits until the processed issue ID information becomes the value immediately before the issue ID added to the request to be processed. That is, as in the case of step S586, the process waits without proceeding with the process according to the snapshot creation command until all the processes according to the command issued immediately before the snapshot creation command are completed.

  When the value of the processed issue ID information becomes the previous value of the issue ID, the processing sequencer 411 performs the following processing on all entries in the address table in the address table storage unit 414. That is, when “changed” is recorded in the flag corresponding to the latest logical address space, the unused block table in the address table storage unit 414 is set to the physical block number corresponding to the latest logical address space as an unused block. Register with. Next, the physical block number corresponding to the snapshot logical address space is copied to the block number corresponding to the latest logical address space. Further, the flag corresponding to the latest logical address space is changed from “changed” to “not changed”.

  If the flag corresponding to the latest logical address space is “no change”, it means that the data stored in the address corresponding to the physical block number has not been changed since the immediately preceding resumable point. Therefore, no processing is performed on the entry.

  If all the above processes are completed, the processing sequencer 411 changes the value of the processed issuance ID information to the issuance ID added to the snapshot creation command. Then, a response to the snapshot creation command is transmitted to the storage 400 using the communication unit 412.

  This modification is different from the already described eighth embodiment in that the storage 401 advances the processing of each command in the order of the issue ID. However, there is no difference in the processing for the entry itself. Therefore, even if an abnormality occurs in the host 300, the time until the host 303 resumes processing can be short. In addition, the data transfer time between the storages 400 and 401 can be shortened.

Embodiment 9 FIG.
FIG. 37 is a block diagram showing a ninth embodiment of a data replication system according to the present invention. In the data replication system shown in FIG. 37, a normal storage 400 is locally connected to a normal host 600 that uses the storage 400. A standby storage 401 is locally connected to a standby host 601 that uses the storage 401. The storage 400 is connected to the storage 401 via the network 13. The host 600 is connected to the host 601 via the network 602.

  The configurations and operations of the storages 400 and 401 are the same as the configurations and operations of the storages 400 and 401 having the snapshot function in the eighth embodiment (see FIG. 29). Further, in place of the storages 400 and 401, the storages 301 and 302 (see FIG. 21) in the seventh embodiment may be used.

  The networks 13 and 602 may be a single network. However, the network 602 to which the hosts 600 and 601 are connected is preferably a dedicated line.

  FIG. 38 is a block diagram illustrating a configuration example of the host 600. As shown in FIG. 38, one or a plurality of applications operate on the host 600. Here, two applications 603a and 603b are illustrated. The applications 603 a and 603 b use the IO management unit 311 to access data in the storage 400. Further, the IO management unit 311 includes a resumable point notifying unit 312 for notifying the storage 400 of resumable points. In addition, an execution image transfer unit 604 that transfers execution images of the applications 603a and 603b to the host 601 and notifies the resumable point notification unit 312 of the resumable point is provided.

  The applications 603a and 603b in the present embodiment are applications that do not have a resume function. In other words, the application does not have a function that can resume processing if the data recording state of the storage medium 101 of the storage is in a predetermined state.

  FIG. 39 is a block diagram illustrating a configuration example of the host 601. As illustrated in FIG. 39, the host 601 includes an execution image storage unit 606 that stores an execution image, an execution image reception unit 605 that receives the execution image transmitted from the host 600 and stores the execution image in the execution image storage unit 606, A host monitoring unit 608 that monitors the status of the host 600 and an application resuming unit 607 that resumes an application based on the execution image in the execution image storage unit 606 are provided. The execution image storage unit 606 is a medium for storing data such as a volatile semiconductor memory, a nonvolatile semiconductor memory, a magnetic disk, a magneto-optical disk, and an optical disk.

  The execution image is information indicating the processing execution status of the active system, and is information necessary for causing a process executed by each application to be executed on another host. The execution image includes, for example, data in a virtual address space of each process, a register value as process management information, a program counter value, a process state, and the like. The file used by the process, information for restoring communication between processes, and the like are included. When a single process is realized by a plurality of threads by parallel processing, the execution value includes a register value of each thread, a value of a program counter, a program status word, various flags, and the like. Further, the execution image may include the contents of a communication buffer in the kernel used by the target process. A thread is a processing unit in parallel processing.

  Next, the operation will be described. First, the operation at the time of execution image transfer will be described. For example, an execution image transfer instruction is input from the user to the execution image transfer unit 604 of the normal host 600. The timing at which this instruction is input may be any timing. When the instruction is input, the execution image transfer unit 604 acquires execution images of the applications 603a and 603b. Next, the execution image transfer unit 604 instructs the resumable point notification unit 312 of the IO management unit 311 to notify the storage 400 of the resumable point. Then, the resumable point notifying unit 312 notifies the storage 400 of the resumable point. Next, the execution image transfer unit 604 transfers the acquired execution image to the execution image reception unit 605 of the host 601. In the host 601, when the execution image receiving unit 605 receives the execution image from the execution image receiving unit 606 of the host 600, the execution image receiving unit 605 stores the received execution image in the execution image storage unit 606.

  The storage 400 receives the resumable point notification at the time when the transfer of the execution image is instructed, and operates according to the notification. Therefore, the standby storage 401 creates a snapshot when execution image transfer is instructed in the normal host.

  When the same storages 301 and 302 (see FIG. 19) as in the seventh embodiment are used as the storage, the standby storage 401 indicates the state of the storage medium when execution image transfer is instructed in the normal host. Maintain the state at.

  The capacity of the execution image may be several hundred MB or more, and it takes time to transfer the execution image to the host 601. When the applications 603a and 603b advance the processing during the transfer of the execution image, the values of the memory and the register change, and the contents of the execution image change during the transfer. Therefore, the applications 603a and 603b stop processing during execution image transfer.

  Alternatively, the execution image information may be stored in a memory or a magnetic disk (not shown in FIG. 37) provided in the host 600, and the stored execution image may be transmitted to the host 601. Once the storage is completed, even if the process is advanced, the contents of the stored execution image are maintained. Therefore, in this case, the applications 603a and 603b of the host 600 can proceed with processing. The time for storing the execution image in the memory or the like is shorter than the transfer time of the execution image. Accordingly, if the execution image is stored in a memory or the like and then transferred, the time for stopping the processing can be shortened.

  Here, the execution image transfer unit 604 transfers the execution image in accordance with an instruction from the user. The timing at which the execution image transfer unit 604 starts the execution image transfer process is not limited to the time when an instruction is input from the user. For example, the execution image transfer process may be started at regular time intervals. Alternatively, the execution image transfer process may be started again when the execution image transfer ends. Alternatively, the transfer process may be started when the application becomes idle or when the idle time of the application exceeds a certain value.

  Next, the operation when a disaster occurs will be described. When the host 600 becomes unusable due to a disaster or the like, the storage 401 is changed from the standby system to the normal system. In addition, the host 601 takes over processing from the host 600.

  The operation of the host 601 from failure detection to application restart will be described. First, the host monitoring unit 608 in the host 601 detects an abnormality of the host 600. For example, the host monitoring unit 608 communicates with the host 600 constantly or periodically, and recognizes that an abnormality has occurred when communication with the host 600 is disabled for a certain period of time. Alternatively, when an abnormality is reported from the host 600, it may be recognized that a disaster has occurred in the host 600.

  When the host monitoring unit 608 detects an abnormality of the host 600, it issues a snapshot restoration command to the storage 401 that is the standby system. Then, the host monitoring unit 608 waits for a response to the snapshot restoration command, and upon receiving the response, the host monitoring unit 608 instructs the application resuming unit 607 to resume the application. The application resuming unit 607 uses the execution image stored in the execution image storage unit 606 to restart the operation of the application application. In this example, the storage 400 and 401 similar to the fifth embodiment are used as the storage. When the storages 301 and 302 (see FIG. 19) similar to those of the seventh embodiment are used as the storage, the host monitoring unit 608 may output a delayed data discard command instead of the snapshot restoration command.

  Note that, when resuming the processing of an application, the application resuming unit 607 changes the information at the time of restoration if there is information in the execution image that needs to be changed by the operating host.

  In the ninth embodiment, when the transfer of the execution image is instructed in the normal host, the normal storage receives the resumable point notification. Therefore, the standby storage system creates a snapshot at the time when transfer of the execution image is instructed in the normal system host. Alternatively, the state of the storage medium at the time when transfer of the execution image is instructed in the normal host is maintained. Further, the standby host holds the execution image of the normal host when the execution image transfer is instructed. Therefore, even if the restart function is not realized by the application, the standby storage can reproduce the normal storage state at a predetermined time, and the standby host can reproduce the execution image at that time. Processing can be resumed quickly in the standby system.

  In the ninth embodiment, the execution image transfer unit 604 may transfer only the difference from the previously transferred execution image, instead of transferring all execution images. In this way, if only the part changed from the previously transferred execution image is transferred, the transfer time can be shortened.

  Computer systems using virtual memory often use a conversion table between logical addresses and physical addresses that have a high probability of being used. This conversion table is called TLB (Translation Look-aside Buffer). In the TLB, for each address to be converted, update information indicating whether or not the data at that address has been rewritten is managed. This update information is generally called a dirty flag or dirty bit. For example, in "Quantitative approach to computer architecture design, implementation and evaluation (David Patterson, John Hennessy, Yuji Tomita and two other translations, Nikkei BP, February 1994, p.442-443)" This update information is described as a dirty bit. The execution image transfer unit 604 may transmit information corresponding to the dirty flag indicating that there has been a change as an execution image.

  In virtual storage, placing a page necessary for executing a program in a real storage device is called page-in, and removing a page from the real storage device and placing it in a virtual storage device is called page-out. When the execution image transfer unit 604 is instructed to transfer the execution image, the execution image transfer unit 604 transmits the page-in memory data to the host 601, and then the page-out memory data at the timing when the page-out occurs. May be transmitted to the host 601. The paged-out memory data is not changed until the next page-in. Therefore, the transmission amount of the execution image can be reduced by transmitting the execution image in this way.

  40 and 41 are block diagrams illustrating a configuration example in the case where the seventh to ninth embodiments are applied to a client server system. Usually, the host 300 functions as a server for each of the clients 500-1 to 500-n. When the above occurs in the host 300 or the storage 400 due to a disaster or the like, the host 303 operates. After the host 303 starts operation, each client requests the host 303 for processing using the host 303 as a server.

  As shown in FIG. 40, when the host 300 and the host 303 are directly connected, the host 303 is connected to the host 300 as described in the seventh to ninth embodiments. Monitor status. As shown in FIG. 41, when the host 300 and the host 303 are not directly connected, the clients 500-1 to 500-n monitor the status of the host 300, and when the above occurs in the host 300. The host 303 may be notified of the above occurrence.

  In the seventh to ninth embodiments, similarly to the first to fourth embodiments, data may be transmitted to the standby storage via the relay device. Good. In that case, the effects of the first to fourth embodiments can also be obtained. In addition, when data is transmitted / received between storages or between a storage and a relay device, a redundant data group is transmitted / received as shown in the fifth embodiment or the sixth embodiment. May be. In that case, the same effect as the fifth embodiment or the sixth embodiment can be obtained.

It is a block diagram which shows 1st Embodiment of a data replication system. It is a block diagram which shows the structural example of a storage. It is a block diagram which shows the structural example of a relay apparatus. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows operation | movement of a relay process part. It is a flowchart which shows operation | movement of the relay process part of 2nd Embodiment. It is a flowchart which shows operation | movement of the process sequencer of 3rd Embodiment. It is a block diagram which shows 4th Embodiment of a data replication system. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows operation | movement of a storage. It is a flowchart which shows the process of step S221 concretely. It is a timing diagram showing an example of data transfer. It is a block diagram which shows 5th Embodiment of a data replication system. It is a block diagram which shows the structural example of a storage. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows operation | movement of a communication part. It is a block diagram which shows the other structural example of a data replication system. It is a flowchart which shows operation | movement of the process sequencer of 6th Embodiment. It is a block diagram which shows 1st Embodiment of a data replication system. It is a block diagram which shows the structural example of a host. It is a block diagram which shows the structural example of a storage. It is explanatory drawing which shows the example of synchronous ID and issue ID. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows operation | movement when a process sequencer receives a delay write command. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows the operation | movement of a delay data reflection process. It is a flowchart which shows a delay data discard process. It is a block diagram which shows 8th Embodiment of a data replication system. It is a block diagram which shows the structural example of a storage. It is explanatory drawing which shows an example of an address table. It is explanatory drawing of an entry. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows operation | movement of a LBA management part. It is a flowchart which shows operation | movement of a process sequencer. It is a flowchart which shows operation | movement of a process sequencer. It is a block diagram which shows 9th Embodiment of a data replication system. It is a block diagram which shows the structural example of a host. It is a block diagram which shows the structural example of a host. It is a block diagram which shows the structural example at the time of applying this invention to a client server system. It is a block diagram which shows the structural example at the time of applying this invention to a client server system.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Host 11,12 Storage 13,14 Network 15 Relay apparatus 101 Storage medium 102 Communication part 103 Processing sequencer 104 IO scheduler 105 Medium control part 106 Buffer memory 151 Relay processing part 150 Communication part 152 Buffer memory

Claims (13)

In a data transmission / reception method for transmitting data from a transmission source that transmits data to a transmission destination that receives data,
The data transmission / reception method characterized in that the transmission source creates at least one redundant data for error correction from the transmitted original data, and transmits the original data and the redundant data in different data transmission units. In the transmission destination, before completing the reception of all of the data group that is a set of the original data and the redundant data, at the stage of receiving a part of the data group that can partially perform error correction processing on the original data, The data transmission / reception method according to claim 1, wherein error correction processing is executed. 3. The transmission source according to claim 1 or 2, wherein the transmission source divides the original data into divided data, and creates redundant data that can restore the original data even if one or more of the divided data is lost. Data transmission / reception method.   4. The data transmission / reception method according to claim 3, wherein parity data or ECC is used as redundant data. The data transmission / reception method according to any one of claims 1 to 3, wherein duplicate data of transmission data is used as redundant data. The data transmission / reception method according to any one of claims 1 to 5, wherein the original data and the redundant data are transmitted to different communication networks. In a data replication system that mirrors or backs up data in the first storage to the second storage via a communication network,
The first storage includes data transfer processing means for controlling data transfer, and redundancy means for creating redundant data for error correction from the transmitted original data,
The data transfer processing means transmits the original data and the redundant data created by the redundancy means in separate data transmission units. The second storage includes data restoring means for performing error correction processing using redundant data received from the first storage, and storage processing means for storing data restored by the data restoring means in a storage medium,
The data restoration means can execute partial error correction processing on the original data before completing reception of all of the data group that is a set of original data and redundant data from the first storage. The data replication system according to claim 7, wherein error correction processing is executed when a part of the data is received. The redundancy means in the first storage divides the original data into divided data, and creates redundant data that can restore the original data even if one or more of the divided data is lost. The data replication system according to claim 8.   The data duplication system according to claim 9, wherein the redundancy means uses parity data or ECC as redundant data. The data duplication system according to any one of claims 7 to 9, wherein the redundancy unit creates duplicate data of the original data as redundant data. The data replication system according to any one of claims 7 to 11, wherein the data transfer processing means sends the original data and the redundant data to different communication networks. A computer provided in the first storage in the data replication system that mirrors or backs up data in the first storage to the second storage via a communication network.
Creating at least one redundant data for error correction from the transmitted original data;
A program for duplicating data in the storage for executing processing for transmitting original data and redundant data in separate data transmission units.

Images