KR101470857B1 - Network distributed file system and method using iSCSI storage system - Google Patents

Abstract

Asymmetric Clusters A system and method are provided for increasing the efficiency of a distributed file system using an iSCSI storage system in a distributed file system.
The network distributed file system according to the present invention is driven including a metadata server M, a data server D and an application server 0 and stores the metadata server M and the data server D as block I / Wherein the iSCSI gateway further comprises a plurality of iSCSI gateways for processing SCSI commands based on an iSCSI interface, wherein the iSCSI gateway functions as a standard interface for block data input / output requested outside the network- A SCSI CDB processing module for supporting an iSCSI interface in a distributed file system, and a block layout manager for managing data block information distributed in a plurality of physical disk spaces.

Description

[0001] The present invention relates to a network distributed file system and method using an iSCSI storage system,

The present invention relates to a system and method for increasing the efficiency of a distributed file system using an iSCSI storage system in an asymmetric cluster distributed file system.

In recent years, with the rapid development of the Internet technology, the transmission of multimedia data such as photographs and moving images has been dramatically increased. Therefore, a company providing a multimedia data transmission service using the Internet has generated a very large amount of data. Therefore, it is urgent to introduce a system capable of effectively distributing and managing a large amount of data.

Recently, advances in storage system or file system technology have greatly improved the scalability and performance of storage systems. In terms of the file system structure, some systems are making efforts to improve the scalability and performance of distributed storage systems by constructing a so-called asymmetric cluster file system that separates the data input / output path of the file and the metadata management path of the file.

This structure allows the client system to directly access the storage devices and also increases storage scalability by avoiding bottlenecks caused by frequent file accesses.

Enterprise-class storage solutions such as IBM's StorageTank and Panasas ActiveScale Storage Cluster, Cluster Filesystems' Luster, Hadoop's DFS and Google's Google Filesystem were developed based on this structure.

In such a network-based distributed file system environment, an application server, a metadata server, and a data server communicate with each other via a network to provide data input / output. In order for the application server to access a specific file, the metadata server obtains the location information of the block in which the actual data of the file is stored, accesses the data server storing the actual data using the location information, Data is obtained.

However, in the distributed file system of the conventional asymmetric structure, only the file IO is provided, and the block IO is not provided.

It is an object of the present invention to provide a distributed file system capable of supporting Block IO as well as file IO by solving the problems of the conventional asymmetric cluster distributed file system.

A network distributed file system driven by including a metadata server (M), a data server (D), and an application server (0)

The distributed file system includes a plurality of (i. E., SCSI) interfaces for processing SCSI (Small Computers System Interface) commands on an iSCSI (Internet Small Computer System Interface) interface by utilizing the metadata server M and the data server D as block input / Gt; iSCSI < / RTI > gateway (Gateway) and an iSCSI control server (Control Server)

The iSCSI gateway includes a SCSI target core serving as a standard interface for block data input / output requested outside the network-based storage, SCSI middleware for supporting the iSCSI interface in the distributed file system, And a block layout manager (Block Layout Manager) for managing distributed data block information.

Preferably, the control server includes a function of configuring an iSCSI volume using a storage space of the distributed file system and controlling a plurality of iSCSI gateways.

In addition, the SCSI target core includes iSCSI target management, network portal management, target portal group management, LUN mapping, Challenge Handshake Authentication Protocol (CHAP) based iSCSI authentication management, multiple connection type iSCSI session management, (Protocol Data Unit) and a command process.

In addition, the SCSI middleware is a SCSI CDB module for processing a SCSI CDB (Command Descriptor Block). The SCSI middleware decodes the CDB, identifies and processes the commands according to the standard of the corresponding SCSI command, and interlocks with the block layout manager, Or outputting the command.

Also, the block layout manager manages the location, storage unit, and number of file chunks constituting the iSCSI volume, and stores the location, the storage unit, and the number information of the file fragment and the information of the block input / output received from the SCSI CDB processing module And a function of performing block data input / output by mapping block address and input / output size information as parameters.

In addition, the meta information about the file managed by the metadata server includes at least one of a file name, a directory structure, a data piece list constituting a file, and position information on a data server.

In addition, the network distributed file method using the iSCSI storage system is driven by the network distributed file system using the iSCSI storage system,

The iSCSI control server requesting a metadata server to generate an iSCSI volume;

Assigning an ID of an iSCSI volume to be created to the metadata server and allocating a chunk to be stored in a data server distributed on the network to a size of the chunk; And

When the metadata server returns iSCSI volume ID information to the iSCSI management server, transmitting the returned iSCSI volume ID information to a block layout manager associated with the iSCSI volume.

Also, the network distributed file method using the iSCSI storage system is:

Receiving a request to read SCSI block data from the application server in the iSCSI gateway in an iSCSI standard manner, the SCSI target core analyzes and decomposes the iSCSI PDU to deliver a SCSI CDB to the SCSI middleware;

The SCSI middleware analyzing SCSI commands and transferring LBA (Logical Block Address) and read size information of a block to be read to a block layout manager;

The block layout manager calculates a chunk ID and an offset constituting the corresponding iSCSI volume using the LBA address, transfers the size of the data to be read to the metadata server, and stores the chunk ID, offset, Arranging a list of all related chunk file locations of a data server area distributed in a network on the basis of size information of the data in the order of IDs and transmitting the sorted list to the block layout manager;

Wherein the block layout manager reads data from the offset of the first chunk file to the last chunk file based on the chunk file location list and writes the data to the buffer located in the block layout manager.

The block layout manager transfers the location on the buffer of the buffer to the SCSI middleware, the SCSI middleware generates a SCSI CDB including the contents of the buffer, and transmits the SCSI CDB to the SCSI target core; And

The SCSI target core may include generating an iSCSI PDU including a SCSI CDB and transmitting the generated iSCSI PDU to an application server in an iSCSI standard manner.

Also, the network distributed file method using the iSCSI storage system is:

The block layout manager recording data in a buffer located in the block layout manager 1-3 by a size of data to be written from the offset of the first chunk file to the last chunk file based on the chunk file location list;

The block layout manager transfers the write completed information to the SCSI middleware 1-2 and the SCSI middleware 1-2 generates the SCSI CDB including the information that the data write has been completed As shown in FIG.

The distributed file system according to the present invention can support not only the file IO but also the block IO, so that data transfer and storage can be efficiently performed.

FIG. 1 is a diagram showing a schematic configuration of an iSCSI storage system and a network distributed file system using the storage system according to a preferred embodiment of the present invention.
2 is a schematic diagram illustrating an iSCSI Volume formation procedure for supporting Block IO.
3 is a view schematically showing a procedure of processing a data read in the block IO.
Fig. 4 is a view schematically showing a procedure of processing data write (Data Write) in block IO.

Preferred embodiments of an iSCSI storage system and a network distributed file system using the same according to a preferred embodiment of the present invention will be described below with reference to the accompanying drawings.

The network distributed file system according to the present embodiment is a storage system that creates a logical iSCSI volume (LUN) and provides it to an iSCSI initiator through an iSCSI target to process a block IO.

Also, according to a preferred embodiment of the present invention, the expansion of the storage capacity is provided in the form of adding nodes of the distributed file system.

Also, according to a preferred embodiment of the present invention, the overall input / output bandwidth expansion is provided in the form of adding an iSCSI target of a gateway type.

According to a preferred embodiment of the present invention, more than two iSCSI gateways are used for ensuring storage availability and fail-over is automatically provided to other iSCSI gateways when an iSCSI gateway fails.

FIG. 1 is a diagram showing a schematic configuration of an iSCSI storage system and a network distributed file system using the storage system according to a preferred embodiment of the present invention.

1, a network distributed file system according to a preferred embodiment of the present invention includes a metadata server M, a data server D, and an application server (hereinafter, referred to as " 0). Specifically, the metadata server M performs a control path function in the distributed file system. In addition, the metadata server M manages meta information (file attribute, path, path list of data pieces on the path (D), replication status, etc.) in the file. The data server D also functions as a data path in the distributed file system. That is, when a file is divided into data fragments by the metadata server M, a function of distributing and storing substantially separated data fragments and a function of dispersing and transmitting separated data fragments are performed.

The distributed file system according to the preferred embodiment of the present invention is a system for processing SCSI commands based on an iSCSI interface by utilizing the metadata server M and the data server D as block input / output storage spaces. Also, according to a preferred embodiment of the present invention, an iSCSI initiator is formed in the application server (0).

In addition, the network distributed file system according to the preferred embodiment of the present invention further includes an expandable iSCSI storage system, which includes a plurality of iSCSI gateways 1 and an iSCSI control server 2. Each of the plurality of iSCSI gateways 1 includes a SCSI target core 1-1, a SCSI middleware 1-2, and a block layout manager set 1-3.

The iSCSI control server 2 and the plurality of iSCSI gateways 1 constituting the scalable iSCSI storage system and the metadata server M, the data server D and the application server 3 are connected to each other via Ethernet , And is configured to extensively provide block I / O services to a plurality of existing iSCSI initiators 3.

According to a preferred embodiment of the present invention, the metadata server M and the data server D, which are used as a base storage of the scalable iSCSI storage system, support distributed data processing for each iSCSI gateway 1, It is desirable to provide scalability. In addition, the iSCSI gateway 1 is preferably composed of a plurality of iSCSI gateways 1 in order to satisfy the performance requirements of a plurality of iSCSI initiators using the block data input / output service.

The iSCSI gateway 1 according to the preferred embodiment of the present invention comprises a SCSI target core 1-1, a SCSI middleware 1-2 and a block layout manager set 1-3. And the like.

First, the SCSI target core 1-1 functions as a standard interface for block data input / output requested outside the storage based on the Ethernet network. Specifically, the SCSI target core 1-1 can perform iSCSI target management, network portal management, target portal group management, LUN mapping, iSCSI authentication management based on Challenge Handshake Authentication Protocol (CHAP) ) Management, an iSCSI protocol data unit (PDU), and a command processing function. On the other hand, the mapping between the iSCSI volume (LUN) on the UCS-FS and the target is achieved through control of the SCSI target core of the iSCSI volume manager.

Meanwhile, the iSCSI interface function encapsulates the CDB to generate a PDU in order to exchange SCSI commands and parameters encapsulated in the form of Ethernet-based TCP / IP, and transmits the promised It provides only the role of delivering according to the procedure. Therefore, in order to support the iSCSI interface to the distributed file system, a file-based solution, a module for processing SCSI CDB is essential. The SCSI middleware 1-2 decodes the CDB as a module configured according to this purpose and processes the corresponding command in accordance with the standard of the corresponding SCSI command. The SCSI command can be largely classified into a data input / output command and a non-input / output command, and interlocking with the block layout manager is determined according to the above-mentioned classification. The data input / output instruction is interlocked with the block layout manager. In addition, when the processing of the SCSI command is completed, the processed result is encapsulated in the CDB format and transferred to the SCSI target core.

The block layout manager 1-3 is a block for managing data block information distributed in a plurality of physical disk spaces. The storage space on the distributed file system is physically divided into data servers and stored in chunks of data. The set of data fragments is mapped to a single "file" on the distributed file system. The meta information of the file managed by the metadata server of the distributed file system includes a file name, a directory structure, a list of data fragments constituting the file, and location information on the data server. The iSCSI volume (LUN) of the scalable iSCSI storage system corresponds to any file specified in the storage space of the distributed file system. Therefore, in order to link (LUN mapping) the file and the iSCSI target, a block layout manager that manages the meta information of the iSCSI volume connected to the iSCSI target that the iSCSI gateway serves is essential. The block layout manager manages the location, storage unit, and number of the file chunks constituting the iSCSI volume, via the metadata server of the distributed file system, if necessary, in the internal storage. And performs block data input / output by mapping the information and the block address and input / output size information, which are parameters of the block input / output received from the SCSI middleware side.

Also, according to a preferred embodiment of the present invention, the iSCSI control server 2 performs control and monitoring of the iSCSI gateway 1. Specifically, the iSCSI control server 2 is a block that configures an iSCSI volume using the storage space of the distributed file system and controls a plurality of iSCSI gateways 1. [ The iSCSI control server 2 can perform iSCSI target management control, network portal management control, target portal group management control, LUN mapping control, and iSCSI authentication management control through the interface provided by the iSCSI gateway 1-1 .

ISCSI volume formation procedure to support block IO

First, an iSCSI volume formation procedure for supporting a block IO will be described in detail below with reference to FIG.

First, the iSCSI management server 2 requests the metadata server M to create an iSCSI volume. (ISCSI volume name / size)

Next, the metadata server M allocates the ID of the iSCSI volume to be created and assigns a chunk to be stored in the data server D distributed on the network to the size (Chunk ID, Chunk position, )do.

Next, when the metadata server M returns the iSCSI volume ID information to the iSCSI management server 2, the returned iSCSI volume ID information is transmitted to the block layout manager 1-3 associated with the iSCSI volume, The iSCSI volume creation procedure to support < RTI ID = 0.0 >

Block IO: Data Data Read Process

Next, a procedure for processing data reading in the block IO will be described below with reference to Fig.

First, the application server 0 requests the iSCSI gateway 1 to read SCSI block data in the iSCSI standard system.

Then, the SCSI target core 1-1 analyzes the iSCSI PDU, decomposes it, and transmits the SCSI CDB to the SCSI middleware 1-2.

Then, the SCSI middleware 1-2 analyzes the SCSI command and transfers the logical block address (LBA) of the block to be read and the size information of the data to be read to the block layout manager 1-3.

Then, the block layout manager 1-3 calculates the chunk ID and offset constituting the iSCSI volume using the LBA address, and transmits the size of the data to be read to the metadata server M. [

Then, based on the chunk ID, the offset, and the size information of the data to be read, which are transmitted to the metadata server M, all the related chunk file location lists of the data server (D) (1-3).

Then, the block layout manager 1-3 reads the data from the offset of the first chunk file to the last chunk file based on the chunk file location list and reads the data by the size of the data to be stored in the buffer located in the block layout manager 1-3 Record the data.

Then, the block layout manager 1-3 transfers the location of the buffer on the memory to the SCSI middleware 1-2, and the SCSI middleware 1-2 generates the SCSI CDB containing the contents of the buffer.

Then, the generated SCSI CDB is delivered to the SCSI target core 1-1. The SCSI target core 1-1 generates an iSCSI PDU including the SCSI CDB, and then transmits the generated iSCSI PDU to the application target To the server (0).

Block IO: Data data write process

Next, a procedure of processing data data write (Data Write) in the block IO will be described below with reference to FIG.

First, the application server 0 requests the iSCSI gateway 1 to write SCSI block data in the iSCSI standard manner.

Then, the SCSI target core 1-1 analyzes the iSCSI PDU, decomposes it, and transmits the SCSI CDB to the SCSI middleware 1-2.

Then, the SCSI middleware 1-2 analyzes the SCSI command and transfers the logical block address (LBA) of the block to be written with data and the size information of the data writing to the block layout manager 1-3.

Then, the block layout manager 1-3 calculates the chunk ID and offset constituting the corresponding iSCSI volume by using the LBA address, and transfers the size of the data write to the metadata server M. [

Then, based on the chunk ID, offset, and size of data write information transmitted to the metadata server M, it is determined whether all relevant chunk files in the data server (D) area distributed on the network exist. If there is no chunk file A chunk is generated at the assigned position, and the chunk position list is sorted in the ID order and transmitted to the block layout manager 1-3.

Then, the block layout manager 1-3 writes data to the buffer located in the block layout manager 1-3 by the amount of data writing from the offset of the first chunk file to the last chunk file based on the chunk file location list.

Subsequently, the block layout manager 1-3 transmits the information to which the data writing has been completed to the SCSI middleware 1-2, and the SCSI middleware 1-2 includes the information that the data writing has been completed Create a SCSI CDB.

Then, the generated SCSI CDB is delivered to the SCSI target core 1-1. The SCSI target core 1-1 generates an iSCSI PDU including the SCSI CDB, and then transmits the generated iSCSI PDU to the application target To the server (0).

As described above, the iSCSI storage system and the network distributed file system using the same according to the preferred embodiment of the present invention have been described in detail with reference to the accompanying drawings. However, it will be understood by those skilled in the art that various modifications and variations can be made in the present invention. Accordingly, the scope of the present invention is limited only by the claims that follow.

Claims (9)

A network distributed file system driven by including a metadata server (M), a data server (D), and an application server (0)
The distributed file system includes a plurality of (i. E., SCSI) interfaces for processing SCSI (Small Computers System Interface) commands on an iSCSI (Internet Small Computer System Interface) interface by utilizing the metadata server M and the data server D as block input / Gt; iSCSI < / RTI > gateway (Gateway) and an iSCSI control server (Control Server)
The iSCSI gateway includes a SCSI target core serving as a standard interface for block data input / output requested outside the network-based storage, SCSI middleware for supporting the iSCSI interface in the distributed file system, And a block layout manager (Block Layout Manager) for managing distributed data block information. The network distributed file system using an iSCSI storage system according to claim 1, wherein the iSCSI control server includes a function of configuring an iSCSI volume using a storage space of a distributed file system and controlling a plurality of iSCSI gateways. 2. The method of claim 1, wherein the SCSI target core comprises at least one of an iSCSI target management, a network portal management, a target portal group management, a LUN mapping, a Challenge Handshake Authentication Protocol (CHAP) based iSCSI authentication management, , an iSCSI protocol data unit (PDU), and a command (Command) process. The network distributed file system using the iSCSI storage system. The SCSI middleware according to claim 1, wherein the SCSI middleware is a SCSI CDB module for processing a SCSI CDB (Command Descriptor Block). The SCSI middleware decodes the CDB, identifies and processes the command according to the standard of the SCSI command, , And a function of commanding data input or output. The network distributed file system using the iSCSI storage system. The method as claimed in claim 1, wherein the block layout manager manages a location, a storage unit, and a number of file chunks constituting an iSCSI volume, and transmits the location, the storage unit, and the number information of the file fragment to the SCSI CDB processing module And a function of performing block data input / output by mapping block address and input / output size information which are parameters of received block input / output. The method as claimed in claim 1, wherein the meta information about the file managed by the meta data server includes at least one of a file name, a directory structure, a list of data pieces constituting a file, and location information on a data server. Network distributed file system using system. A network distributed file method driven using a network distributed file system using an iSCSI storage system according to any one of claims 1 to 6,
The iSCSI control server requesting a metadata server to generate an iSCSI volume;
Assigning an ID of an iSCSI volume to be created to the metadata server and allocating a chunk to be stored in a data server distributed on the network to a size of the chunk; And
When the metadata server returns iSCSI volume ID information to the iSCSI management server, transmitting the returned iSCSI volume ID information to a block layout manager associated with the iSCSI volume. The method of claim 7, wherein the network distributed file method comprises:
Receiving a request to read SCSI block data from the application server in the iSCSI gateway in an iSCSI standard manner, the SCSI target core analyzes and decomposes the iSCSI PDU to deliver a SCSI CDB to the SCSI middleware;
The SCSI middleware analyzing the SCSI command and transferring the logical block address (LBA) of the block to be read and the size information of the data to be read to the block layout manager;
The block layout manager calculates a chunk ID and an offset constituting the corresponding iSCSI volume using the LBA address, transfers the size of the data to be read to the metadata server, and stores the chunk ID, offset, Arranging a list of all related chunk file locations of a data server area distributed in a network on the basis of size information of the data in the order of IDs and transmitting the sorted list to the block layout manager;
The block layout manager reading data from the offset of the first chunk file to the last chunk file based on the chunk file location list and writing the data in the buffer located in the block layout manager;
The block layout manager transfers the location on the buffer of the buffer to the SCSI middleware, the SCSI middleware generates a SCSI CDB including the contents of the buffer, and transmits the SCSI CDB to the SCSI target core; And
Wherein the SCSI target core further comprises generating an iSCSI PDU including a SCSI CDB and then transmitting the generated iSCSI PDU to an application server in an iSCSI standard manner. The method of claim 7, wherein the network distributed file method comprises:
The block layout manager recording data in a buffer located in the block layout manager 1-3 by an amount of data to be written from the offset of the first chunk file to the last chunk file based on the chunk file location list;
The block layout manager transfers the write completed information to the SCSI middleware 1-2 and the SCSI middleware 1-2 generates the SCSI CDB including the information that the data write has been completed Further comprising the steps of:

Images