JP4380658B2 - Shared disk device - Google Patents

Description

  The present invention relates to a disk device that can store a large amount of data, and more particularly to a disk device that includes a fiber channel capable of high-speed data transfer at an interface with a host and is shared by a plurality of hosts.

  SCSI is often used as a means for connecting a magnetic disk device or disk array device to a host computer.

  In recent years, with an increase in the amount of data to be accumulated, a user's system is generally configured to install a file server or connect a plurality of servers and clients to a network. In such a situation, when using the database software of a certain manufacturer and using the workflow software of another manufacturer, it is necessary to prepare clients that run the software of each manufacturer. However, since the management method of data on the magnetic disk differs depending on the host file system in the SCSI interface, it is difficult for clients of different manufacturers to share the magnetic disk.

NFS, on the other hand, uses a network TCP / IP to share files from clients from different manufacturers. However, since a file operation command which does not depend on the manufacturer is used, there are problems such as a client-specific file attribute which cannot be used for mapping by the server, and a protocol overhead is large, resulting in low data transfer performance.

Thus, in order to use each feature of a plurality of protocols, a heterogeneous protocol mixed type optical network has been proposed. In Japanese Patent Laid-Open No. 9-51322, different optical wavelengths are assigned to each of a plurality of protocols, so that different protocols can be mixed in the same optical network.

  Japanese Patent Laid-Open No. 10-149334 discloses means for connecting a fiber channel interface adapter to a PCI bus of a personal computer. In this way, by using the TCP / IP protocol, it is possible to perform control in the same manner as NFS via conventional Ethernet (Ethernet is a registered trademark).

JP-A-9-51322 JP-A-10-149334

As described above, using NFS makes it possible to share a large-capacity magnetic disk. However, if shared, the data transfer performance will be degraded and the file attributes unique to each host can be used as they are. The problem of disappearing remains. JP-A-9-5132
If you use the method of assigning different optical wavelengths to multiple protocols using an optical network as described in Publication No. 2, you can use TCP / IP (i.e. NFS) and SCSI protocols at the same time. The above-mentioned problems can be solved by using them accordingly.

  However, the method of assigning the optical wavelength requires a filter with a special optical wavelength and is not a general standard, so that the range of use is limited.

Also, if you use a Fiber Channel interface, TCP / IP (i.e. NFS
) And SCSI protocols can be used and are widely available because they are ANSI standards. However, Japanese Patent Laid-Open No. 10-149334 does not define the use of multiple protocols at the same time. If a host-specific dedicated file can be accessed as it is as a shared file of TCP / IP and SCSI, there will be a problem of data corruption.

The object of the present invention is to use TCP / IP (that is, NFS) and SCSI protocols at the same time and to use file attributes specific to each host. It is an object of the present invention to provide a shared disk processing device that does not destroy data even when accessed.

  Another object of the present invention is to provide a shared disk processing apparatus that can simultaneously use a plurality of protocols while conforming to the ANSI standard.

In order to achieve the above object, the present invention provides a means for identifying TCP / IP and SCSI protocols, a partition for data accessed by the SCSI command and data accessed by the NFS command, and data destruction due to erroneous access. To prevent.

  Also, means for identifying the protocol of TCP / IP and SCSI, means for determining whether or not the command is an NFS command when this identification means is determined to be TCP / IP, and a case where the determination means determines NFS Further, a means for accessing the magnetic disk by converting the NFS file system into a SCSI command is used, and a means for accessing the magnetic disk using the SCSI command when the identification means determines SCSI.

  Further, as means for identifying the TCP / IP and SCSI protocols, means for referring to the source ID of the Fiber Channel header information and means for providing a table for determining the protocol to be used from the source ID are used.

  Further, when the means for determining whether or not the command is an NFS command, it is determined that the command is not NFS, and a means for executing a Telnet process and remotely logging in to a processor that controls the magnetic disk is used.

According to the present invention, a client connected to the same Fiber Channel can access the same file server by using the NFS and SCSI commands as needed, so when sharing data between clients, use the NFS command, You can use SCSI commands if you want fast access without sharing data.

  An embodiment of the present invention will be described below with reference to FIGS. FIG. 1 is a diagram showing a system configuration and processing procedure in which a file server 101 and clients 102 and 103 which are host devices are connected via a fiber channel interface 105. FIG. 2 is a structural diagram of a header used in the fiber channel.

First, the configuration of FIG. 1 will be described. In general, SCSI and Ethernet are often used for connection between a client and a server as a host, but recently, fiber channel has been attracting attention instead. Fiber Channel is a standard for transferring data at 100MB / s using optical fiber or copper wire, and defines the physical layer and the lower logical layer. For the upper logical layer, it is only a framework, and SCSI in Fiber Channel
-3 and TCP / IP protocols can be passed through, for example, the SCSI-3 protocol can be used to control like a conventional SCSI magnetic disk. The file server 101 is connected to the client 102 and the client 103 via the fiber channel optical cable 104.

The fiber channel interface 105, which is a command interface, converts the optical signal of the optical cable 104 into an electrical signal and selects data (command) addressed to the file server 101 from header information from the client. On the contrary, the other party is put into the header information for the client 102 and the client 103, and the electric signal is converted into an optical signal and sent out to the optical cable 104. The protocol determination unit 106 is means for identifying whether the protocol of data (command) sent from the contents of the header information is SCSI or TCP / IP.

The processing determination unit 107 is means for determining whether the data (command) sent by the TCP / IP protocol is an NFS command or other than an NFS command. NFS processing unit 108
Receives an NFS command, converts it into a SCSI command according to its own file system, writes data to the magnetic disk 110 as a recording medium, or reads data recorded on the magnetic disk 110. The magnetic disk interface 109 is an interface for writing data to or reading data from the magnetic disk 110 using a SCSI command.

  Processing when data in the file server 101 is read from the client 102 with a SCSI command will be described below. The client 102 sends a SCSI READ command to the fiber channel interface 105 of the file server 101 via the optical cable 104. The fiber channel interface 105 checks the contents of the destination ID 202 in FIG. 2 to select a READ command addressed to the file server 101 and transfers the READ command to the protocol determination unit 106.

  The protocol determination unit 106 checks whether the protocol 203 in FIG. 2 is SCSI or TCP / IP. In this case, a READ command is passed to the magnetic disk interface 109 for SCSI. In the SCSI READ command, the position of the file name and the like is managed by the client 102. Therefore, the block position on the magnetic disk is simply specified as the parameter of the READ command.

  The magnetic disk interface 109 issues a READ command to the magnetic disk 110 and reads the data of the block specified by the READ command. The read data is sent from the magnetic disk interface 109 to the fiber channel interface 105. The fiber channel interface 105 adds the header information in which the source ID 201 is the file server 101, the destination ID 202 is the client 102, and the protocol 203 is SCSI. Sent out. The client 102 picks up the data whose destination ID 202 is addressed to the client 102 from the optical cable 104, and receives the data read in response to the READ command.

Next, processing when data in the file server 101 is read from the client 103 by using an NFS command is shown below. In order to use data on a file server with NFS, it is necessary to export a file that the file server 101 discloses to the client 103 in advance, and the client 103 needs to NFS mount, but the description is omitted here. The client 103 issues an NFS READ command using TCP / IP and sends it to the fiber channel interface 105 of the file server 101 via the optical cable 104. The fiber channel interface 105 checks the contents of the destination ID 202 in FIG. 2 to select a READ command addressed to the file server 101 and passes the READ command to the protocol determination unit 106.

The protocol determination unit 106 checks the contents of the protocol 203 in FIG. 2 and passes a READ command to the processing determination unit 107 for TCP / IP here. The processing determination unit 107 determines whether the passed command is an NFS command. Since the READ command in this case is one of the NFS commands, the READ command is further passed to the NFS processing unit 108. Since the NFS processing unit 108 manages the position of the file name and the like in the NFS READ command, the file command or the file descriptor serving as a substitute for the file name and the offset from the beginning of the file are specified as the parameters of the READ command. . The NFS processing unit 108 converts the NFS READ command into a SCSI READ command, calculates the block position on the magnetic disk using the file name and the offset from the beginning of the file, and sends the SCSI READ command to the magnetic disk interface 109. And pass the block position.

The magnetic disk interface 109 issues a READ command to the magnetic disk 110 and reads the data of the block specified by the READ command. The read data is sent from the magnetic disk interface 109 to the NFS processing unit 108, converted into the NFS data format by the NFS processing unit 108, and sent to the fiber channel interface 105. In the fiber channel interface 105, the source ID 201 is added to the file server 101, the destination ID 202 is added to the client 103, and the protocol 203 is added to the TCP / IP header information and sent to the optical cable 104. The client 103 picks up the data whose destination ID 202 is addressed to the client 103 from the optical cable 104, and receives the data read in response to the READ command.

As described above, according to this embodiment, clients connected to the same Fiber Channel can access the same file server using NFS and SCSI commands as needed, so when sharing data between clients, If you use NFS commands and want to access them quickly without sharing data, you can use SCSI commands. In the present embodiment, the exchange of data from the client 102 to SCSI and from the client 103 to NFS has been described. However, any file system or both file systems may be used by any client.

  Next, an embodiment in which the position of the magnetic disk used for each client (protocol) is partitioned by partition and an identifier is provided for each partition will be described with reference to FIG. .

  FIG. 3 shows a table for managing clients connected to the optical cable 104. Item 301 is an ID of a client connected to the optical cable 104, item 302 is information on whether the protocol used by each client is TCP / IP, that is, NFS or SCSI, and item 303 is a magnetic disk used by each client (protocol). 110 represents an identifier of a partition. Since TCP / IP is a shared file system, it has the same partition identifier, and since SCSI is a dedicated file system, the same SCSI also has different partition identifiers.

  First, the protocol determination unit 106 checks whether the protocol 203 and the source ID 201 in FIG. 2 are compatible with the client ID 301 and the protocol 302 in FIG. In the case of NFS, the NFS processing unit 108 checks the block position on the magnetic disk 110 and the block position of the partition 303 whose protocol 302 is TCP / IP, thereby preventing access to an erroneous partition. In the case of SCSI, the protocol determination unit 106 checks the block position on the magnetic disk 110 and the block position of the partition 303 whose protocol 302 is SCSI, thereby preventing access to an erroneous partition.

  In this way, since access to an erroneous partition can be checked, data corruption caused by partition and protocol mismatch can be prevented.

  If one client uses only one protocol (for example, the client 103 shown in this figure), the header information from the fiber channel interface 105 is insufficient and the protocol information is missing. Even if it is not possible to determine with this, it is determined which client of the item 301 is from the source ID 201, and if it is the client 103, the protocol can be determined from the corresponding item 302 to be TCP / IP.

  Hereinafter, another embodiment of the present invention will be described with reference to FIG. FIG. 4 is a diagram showing another configuration of the file server.

  The fiber channel adapter 401 is an adapter that connects the optical cable 104 and the bus 407, and writes the data (command) of the optical cable 104 to the memory 403 via the bus 407 or sends the data of the memory 403 to the optical cable 104 via the bus 407. The CPU 402 is a processor that controls the entire file server according to a program written in a ROM (Read Only Memory) 404. The memory 403 is used not only as a temporary buffer when exchanging data between the optical cable 104 and the magnetic disk 406 but also as a temporary data storage location of the CPU 402.

The ROM 404 stores a program for the CPU 402. The programs to be stored are an operating system that is fundamental for the CPU 402 to execute various programs, and a processing procedure corresponding to the protocol determination unit 106, the processing determination unit 107, and the NFS processing unit 108. The SCSI adapter 405 is an adapter that connects the magnetic disk 406 and the bus 407 and corresponds to the magnetic disk interface 109 in FIG. 1, and writes data on the magnetic disk 406 to the memory 403 via the bus 407 or Data is written to the magnetic disk 406 via the bus 407.

The processing procedure itself is the same as the configuration of FIG. That is, client 102 (
The processing in the case of reading the data of the file server 101 with the SCSI command from (see FIG. 1) is as follows.

  The client 102 sends a SCSI READ command to the fiber channel adapter 401 of the file server 101 via the optical cable 104 in accordance with the fiber channel protocol. The fiber channel adapter 401 corresponding to the fiber channel interface 105 of the embodiment shown in FIG. 1 checks the contents of the destination ID 202 in FIG. 2, selects the READ command addressed to the file server 101, and writes it in the memory 403.

  The CPU 402 checks the content of the protocol 203 in FIG. 2 of the READ command written in the memory 403 according to the program that realizes the processing content of the protocol determination unit 106 written in the ROM 404. Here, a READ command is transferred from the memory 403 to the SCSI adapter 405 for SCSI.

The SCSI adapter 405 corresponding to the magnetic disk interface 109 shown in FIG. 1 issues a READ command to the magnetic disk 406 and reads the data in the block specified by the READ command. The read data is written from the SCSI adapter 405 to the memory 403. Further, the data is sent from the memory 403 to the fiber channel adapter 401, and the fiber channel adapter 401 sends the file ID to the file server 101, the destination ID 202 to the client 102, and the protocol 203 to the optical cable 104 with the SCSI header information added. The client 102 picks up the data whose destination ID 202 is addressed to the client 102 from the optical cable 104, and receives the data read in response to the READ command.

The access by the NFS command from the client 103 is the same, and the processing procedure corresponding to the protocol determination unit 106, the processing determination unit 107, and the NFS processing unit 108 is written in the ROM 404 as a program and corresponds to the protocol determination unit. If the program to be processed is identified as TCP / IP, it is determined whether the program corresponding to the processing determination unit is an NFS command. If it is an NFS command, the program is converted to a SCSI command by the program corresponding to the NFS processing unit. Sent to the adapter.

  To remotely log in to the file server 101 from the client 103 using Telnet, a Telnet command is issued using TCP / IP according to the fiber channel protocol and sent to the fiber channel adapter 401 of the file server 101 via the optical cable 104. . The fiber channel adapter 401 checks the contents of the destination ID 202 in FIG. 2 to select the Telnet command addressed to the file server 101 and writes it to the memory 403.

The CPU 402 checks the contents of the protocol 203 in FIG. 2 of the Telnet command written in the memory 403 in accordance with a program that realizes the processing contents of the protocol determination unit 106 in the ROM 404. Here, for TCP / IP, the CPU 402 checks the Telnet command written in the memory 403 in accordance with a program for realizing the processing contents of the processing determination unit 107 in the ROM 404. Since it is not an NFS command, the command is passed to the operating system processing in the ROM 404.

  Since a UNIX (UNIX is a registered trademark) type operating system supports a Telnet command as a standard, processing for Telnet can be realized without specially incorporating a program into the ROM 404.

As described above, according to the present embodiment, even when using a processor that does not use special hardware and uses a widespread processor, a client connected to the same Fiber Channel can properly use NFS and SCSI commands as necessary. You can use the NFS command to share data between clients, and you can use the SCSI command to access data at high speed without sharing data. Further, by remotely logging in via Telnet via fiber channel, various settings of the file server 101 can be made without having a line separate from the data path.

It is a figure which shows a system configuration | structure and a process sequence. It is a header structure figure of a fiber channel. It is a block diagram of a client management table. It is a system configuration diagram.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 ... File server, 102 ... Client, 103 ... Client, 104 ... Optical cable, 105 ... Fiber channel interface, 106 ... Protocol judgment part, 107 ... Processing judgment part, 108 ... NFS processing part, 109 ... Magnetic disk interface, 110 ... Magnetic Disk 201 ... Source ID 202 ... Destination ID 203 ... Protocol 301 ... Client ID 302 ... Protocol 303 ... Partition 401 ... Fiber Channel Adapter 402 ... CPU 403 ... Memory 404 ... ROM 405 ... SCSI adapter, 406... Magnetic disk.

Claims (5)

A magnetic disk for storing data;
A command interface connected to the client and receiving commands from the client;
A disk interface for writing or reading data to or from the magnetic disk in accordance with a command from the client;
An identification means for identifying whether the data from the command interface is a protocol of TCP / IP or SCSI;
A determination unit that determines whether the command is an NFS command when the identification unit determines TCP / IP,
When the determination means determines NFS, the NFS file system processing is performed and converted to a SCSI command, a partition to be written is determined, and processing means for accessing the magnetic disk,
2. The shared disk device according to claim 1, wherein the identification unit determines a partition to which the data is written when it is identified as a SCSI protocol. A command interface connected to the client and receiving commands from the client;
  An identification means connected to the command interface and identifying whether the data from the command interface is a protocol of TCP / IP or SCSI;
  A disk interface connected to the identification means and writing or reading in accordance with a command from the client;
  A magnetic disk connected to the disk interface and having a partition corresponding to the type of protocol used by the client;
  A determination unit that determines whether the command is an NFS command when the identification unit determines TCP / IP,
  Processing means for performing NFS file system processing when converted to a SCSI command and accessing the magnetic disk when the determination means determines NFS;
  The identification unit and the processing unit include a first identifier attached to the partition on the magnetic disk corresponding to each type of protocol used by the client, and a second identifier attached to the client. The shared disk device is characterized in that a partition to which the data is written is determined by the protocol and / or the identifier. A magnetic disk for storing data;
A command interface connected to the client and receiving commands from the client;
  A disk interface connected to the magnetic disk and writing or reading data to or from the magnetic disk in accordance with a command from the client;
  Identification means for identifying whether the data from the command interface is a TCP / IP or SCSI protocol;
  When it is determined that the identification unit is TCP / IP, a determination unit that determines whether the command is an NFS,
  When the determination means determines NFS, the NFS file system processing is performed and converted into a SCSI command, the partition for writing the data is determined, and the processing means for accessing the magnetic disk,
  2. The shared disk device according to claim 1, wherein the identification unit determines a partition to which the data is written when it is identified as a SCSI protocol. The shared disk device according to claim 2,
  The identification unit includes a table describing a relationship between the first identifier, the second identifier, and a protocol used by the client to which the second identifier is attached. apparatus. The shared disk device according to claim 1 or 3,
  The identification means includes a first identifier attached to a partition corresponding to each type of protocol used by the client, a second identifier attached to the client, and the second identifier. A shared disk device comprising a table describing a relationship with a protocol used by the client.

Images