CN102594852A - Data access method, node and system - Google Patents

Abstract

The invention discloses a data access method, device and system, wherein the method includes: receiving an access request including a block number identifier; a pre-established local block unit searches for the block number of the local block unit according to the block number identifier and provides The corresponding relationship of the storage node resource information of the storage service obtains the location of the storage node corresponding to the block number identification and the block data identification; the local block unit sends the block data access to the storage node according to the storage node location and the block data identification corresponding to the block number identification request; wherein, the storage node resource information includes the location information of the storage node providing the storage service and the block data identifier of the storage node. The present invention can solve the defect that each access of the existing distributed file system needs to be addressed through the metadata server, the metadata server becomes the bottleneck of the system, and the access efficiency is low.

Description

Data access method, node and system

technical field

The present invention relates to data service technology in the communication field, in particular, to a data access method, node and system.

Background technique

The most commonly used method of data storage system is the file system. For example, the file system in Figure 1 accesses files through path names, such as /home/myname/novel/novell.txt. The path name is in a namespace, and the path is in a tree In the space, the tree-like space represents a certain logical relationship, which is more user-friendly. The database system is another way of using the storage system. It does not rely on the path of the file system to reflect the logical relationship, and uses "tables" or other structures to represent the logic. In Figure 1, the database system uses "tables" to represent the logic.

As shown in Figure 1, the bottom layers of all data storage systems (file systems and database systems) that require random access are built on block devices. The so-called block device can access a specified area in the device by specifying a block number. The block number can be randomly addressed in a large addressing space, but the block number itself has no logical meaning. Users need to use A system with user logic such as the upper file system or database system.

File system access has a better logical structure, but block access can support various file systems and other storage systems, such as database systems, with greater flexibility. The file reading and writing process in Figure 1 is shown in Figure 2 and Figure 3. Both reading and writing are initiated by the client and sent to the file system layer. The file system client provides the application with operation interfaces such as file opening, positioning, and reading and writing. The file open operation will return a file descriptor, which is a serial number or a pointer, which is a unique identifier for the application to represent the file to be operated. For the file access operation of the application, the file system layer needs to be divided into one or more A deblocking device accesses some blocks at a time to read information or write data. Some blocks store metadata for managing and organizing files, and some blocks store user data. Different file system implementations use different methods to store and read these blocks. The file system translates into multiple block operations. The block device layer finally reaches the disk for operation, and optimization measures such as caching are omitted here. The client using the block device service obtains a storage area in the form of a disk, etc., and stores, reads, updates, and deletes storage blocks through Small Computer System Interface (SCSI) or Advanced Technology Attachment Interface (ATA) block operation instructions. Generally, users will create a file system on the disk for use.

With the development of cloud computing, the challenges of storage systems are capacity and cost, which requires both large-capacity storage space and lower storage system costs. Faced with this requirement, a large number of ordinary servers need to be used to build a large-capacity storage system, and the block access shown in Figures 1-3 is not suitable for distributed access.

The above Figures 1-3 are for accessing block devices through the file system or data system. The following Figures 4-6 are a parallel network file system (pNFS) provided for the large-capacity and low-cost requirements of cloud computing, that is, the user uses the file system Access the pNFS metadata server through the interface to obtain metadata, and then access the data from the distributed storage device (file storage server) through the block interface or file system interface, so as to improve the scalability and performance of the system by using distributed resources. Other distributed file systems such as LustreFS also use a similar architecture.

In Fig. 4, the file system layer of the metadata server stores the corresponding table of file descriptors and data location information (block numbers). In the distributed system in Fig. 4, the concept of blocks is not particularly emphasized, and the data location information needs to have The file storage node information and port number, the user accesses the corresponding file storage node according to the data location information, and the file storage node finds the file or data block according to the file name or block ID. Write access control flowchart.

The client of the distributed file system in Figure 5 and Figure 6 adopts different operations from the file system in Figure 1-3, it can perceive the distribution of the file system, and at least has the access method of the metadata server, regarding file creation, Operations such as opening and closing are all performed through the metadata server. The functions it undertakes can basically be compared to the file system layer of the local file system, and the operation messages are transmitted through the network between the two. As shown in Figure 5, the client application sends a file open access request to the metadata server through the file system interface. The file system of the metadata server finds the corresponding data location information according to the file descriptor in the access request and returns the information to the client application. The client application immediately The corresponding file storage node can be read according to the data location information. As shown in Figure 6, the client application sends a file creation request to the metadata server through the file system interface, and the file system of the metadata server creates a corresponding file descriptor. The specified storage space is returned to the client application, and the specified storage node and data description identifier are included when the space is allocated. The client application finds the corresponding storage node according to the specified allocated space, and writes to the file storage node.

When the metadata server node allocates space or returns data location information, it specifies the storage node for user data and the data description identifier, where the data description identifier is usually an address of a data location or a data number. The specific user data read and write operations occur on the storage node. These specific block data are similar to the blocks of the stand-alone file system, but in different implementations, they may be some files on the storage node or some centralized block storage. block on the device.

In the process of implementing the present invention, the inventors found that each access of the existing distributed file system needs to be addressed through the metadata server one or more times, and then access the storage nodes, and the metadata server becomes the bottleneck of the system.

Contents of the invention

The first purpose of the present invention is to propose a data access method to reduce the access overhead of the prior art through metadata server addressing.

The second object of the present invention is to propose a client node to reduce the access overhead of addressing through the metadata server in the prior art.

The third object of the present invention is to propose a storage control node, so as to reduce the access overhead of addressing through the metadata server in the prior art.

The fourth object of the present invention is to propose a data storage system to improve access efficiency.

In order to achieve the first purpose above, according to one aspect of the present invention, a data access method is provided, including: receiving an access request including a block number identifier;

The pre-established local block unit searches for the corresponding relationship between the block number of the local block unit and the resource information of the storage node providing the storage service according to the block number identifier, and obtains the storage node location corresponding to the block number identifier and the block data identifier;

The local block unit sends a block data access request to the storage node according to the location of the corresponding storage node identified by the block number and the block data identifier;

Wherein, the storage node resource information includes the location information of the storage node providing the storage service and the block data identifier of the storage node.

Preferably, before receiving the access request containing the block number identification, it may also include: sending a block data opening request containing the storage capacity to the control node; Customer node location information, allocation of storage node resource information corresponding to the size of the storage capacity; establishment of a local block device unit based on the storage node resource information.

Preferably, before receiving the access request containing the block number identification, it may also include: receiving a direct access request containing the block number identification through the block device interface and/or receiving the file name/table data access request through the file/database system interface, and parsing the file Name or the block number identifier corresponding to the table data.

Preferably. The local block unit is a virtual local block device driver that provides distributed storage services or a virtual machine that provides application services.

In order to achieve the above-mentioned second purpose, according to another aspect of the present invention, a client node is provided, including: one or more local block units, wherein the local block unit includes: an interface module, used to receive access request; the storage module is used to store the corresponding relationship between the block number of the local block unit and the resource information of the storage node providing the storage service; the resource information of the storage node includes the information of the storage node providing the storage service and the block data identification of the storage node; The processing module is used to search the storage unit according to the block ID, obtain the location of the storage node corresponding to the block ID and the block data ID, and send a block data access request to the storage node through the interface module.

In order to achieve the above third objective, according to another aspect of the present invention, a storage control node is provided, including: an interface module, configured to receive a block data activation/release request;

The control module is configured to allocate corresponding storage node resource information according to the storage resource status information of one or more storage nodes and/or the received client node location information of the block data activation request; or release the corresponding storage node according to the block data release request resource information;

A storage module, configured to store resource information of storage nodes allocated by one or more storage nodes;

Wherein, the storage node resource information includes storage node information and a block data identifier of the storage node.

To achieve the above fourth objective, according to another aspect of the present invention, a data storage system is provided, including a second-purpose client node and a third-purpose storage control node.

The data access method, device, and system of each embodiment of the present invention provide a lower-level local block unit, and the local block unit stores the corresponding relationship between the block number and the pre-allocated storage node resource information. Therefore, when performing data access, It does not need to be addressed through the metadata server in the prior art, so there is no performance bottleneck, the access overhead can be greatly reduced, and the access efficiency is improved.

Some embodiments of the present invention can directly use the block device interface or create its own file system interface to access the local block unit according to the application, so it has better system scalability. The local block unit in the present invention can be applied to a virtual machine, and users can only access allocated storage resources, and the multi-tenant isolation capability for public services is stronger and the security is higher.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

The technical solutions of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

Description of drawings

The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the description, and are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the attached picture:

Fig. 1 is according to the block device in the prior art and file system, the logical relationship diagram of database system;

Fig. 2 is the flow chart of read access operation for Fig. 1;

Fig. 3 is a flow chart of the write access operation for Fig. 1;

Fig. 4 is a logic diagram according to the parallel network file system (pNFS) in the prior art;

Fig. 5 is the flow chart of read access operation for Fig. 4;

Fig. 6 is a flow chart of the write access operation for Fig. 4;

FIG. 7 is a flowchart of Embodiment 1 of a data access method according to the present invention;

FIG. 8 is a flow chart of Embodiment 2 of the data access method according to the present invention;

FIG. 9 is a schematic diagram of the analysis of the data access method according to the present invention, and a schematic structural diagram of the data storage system of the present invention;

FIG. 10 is a schematic diagram of the data resource application opening/release process for FIG. 9;

Fig. 11 is a schematic diagram of an embodiment of the access process in the data access method of the present invention;

Fig. 12 is a flow chart of the read access operation for Fig. 11;

Fig. 13 is a flow chart of the write access operation for Fig. 11;

Fig. 14 is a schematic diagram of an embodiment of a storage control device according to the present invention;

Fig. 15 is a schematic diagram of an embodiment of a client node according to the present invention.

Detailed ways

The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described here are only used to illustrate and explain the present invention, and are not intended to limit the present invention.

method embodiment

FIG. 7 is a flow chart of Embodiment 1 of the data access method according to the present invention. As shown in FIG. 7, this embodiment includes:

Step S102: receiving an access request including a block number identifier;

Step S104: The pre-established local block unit searches for the corresponding relationship between the block number of the local block unit and the resource information of the storage node providing the storage service according to the block number identifier, and obtains the storage node location corresponding to the block number identifier and the block data identifier; where , the storage node resource information includes the location information of the storage node providing the storage service and the block data identifier of the storage node;

The local block unit in this embodiment can obtain the storage node resource information that can provide the server by applying for block data from a control node, and establish a local block unit in advance based on the applied block data, such as providing distributed storage services The virtual local block device driver is provided to the application program, and the application program can perform subsequent data access according to the unit;

Step S106: The local block unit sends a block data access request to the storage node according to the location of the storage node identified by the block number and the block data identifier.

Since the local block unit stores one or more storage node resource information, data access can be performed according to the storage node location and block data identifier in each storage resource information, and the storage node can also distinguish access from different client nodes according to the block data identifier. For details, refer to the subsequent embodiment in FIG. 9 . Since this embodiment provides a lower-level local block unit, which stores the corresponding relationship between the block number and the pre-allocated storage node resource information, it does not need to pass through the metadata server in the prior art when performing data access. etc. for addressing, so there is no performance bottleneck, which can greatly reduce access overhead and improve access efficiency.

FIG. 8 is a flow chart of Embodiment 2 of the data access method according to the present invention. As shown in FIG. 8, this embodiment includes:

Step S202: As shown in Figure 9, the client node sends a block data activation/release request including the size of the storage capacity to the control node;

Step S204: The storage control node (also called management node or control node) monitors the resource usage status of each storage node, and allocates one or more storage nodes to the client node according to the allocated space requested by the client node and the resource status information of each storage node Resources, including storage node location information (address/port number), block data identification, etc.;

The management node can further assign the resource information of the storage node closer to the client node according to the location of the client node sending the request;

The management node can also allocate one or more storage node resources to the client node in combination with the location of the client node and the resource status of the storage node;

Step S206: the client node acquires storage node resource information;

Step S208: Create a local block device driver according to the storage node resource information;

Step S210: The client node can directly access the local block device driver through the standard block device interface access command, and map it to access to the storage node; it can also create a file system or database system by itself, and use the file system interface or database system interface The file name or table data is parsed to obtain the block number identification corresponding to the local block device driver, which is further mapped into an access to a certain storage node by the local block device driver.

This embodiment combines the advantages of centralized access and distributed access. It can access block data and has a large access capacity. The client node obtains the allocation space through the control node and correspondingly increases the local block device driver. Users can use it according to the application Choose to use block devices directly or create your own local file system, so you have better system scalability. This embodiment adopts the added local block device driver, and does not need to search in the metadata server, so there is no performance bottleneck. This embodiment can also be applied to virtual machines, users can only access allocated storage resources, and the multi-tenant isolation capability for public services is stronger and the security is higher.

FIG. 9 is a schematic diagram of the analysis of the data access method according to the present invention, and a schematic structural diagram of the data storage system of the present invention.

The data storage system shown in Figure 9 includes a storage control node and a client node. The control node is a logical centralized management node, which provides an interface for applying for and releasing block data, and can perform global, centralized, Reliable management and control, providing an application interface with client nodes, and controlling the storage resource status of each storage node; storage nodes (also called service nodes) are mainly used to allocate specific block data and provide storage services. After the node allocates the storage resource of a storage node to the client node, the storage node can directly provide the storage resource on the node to the user without going through the control node.

As shown in Figure 9, in terms of physical deployment, the control node is not necessarily a single or independent node, but can be a cluster, or even reuse some storage nodes, which are located in the storage nodes to perform storage management and control functions. The storage node corresponds to each physical server with disk space that provides block device services. The control node receives and processes user applications, releases block devices, and adjusts the size of block devices; the storage node stores user data and performs operations such as adding, deleting, obtaining, and modifying data blocks.

The data allocation and access process will be described in detail below with reference to Figures 10-13.

Figure 10 is a schematic diagram of the data resource application opening/release process for Figure 9, as shown in Figure 10, the data storage resource application/release process occurs between the client node and the control node, including:

Step 1: The client node can apply for a block device of a specified size, such as sending a block data activation request with a specified storage capacity to the control node;

Step 2: Before the control node receives the activation request, it will check the resource usage such as the available space and effective status of each storage node, which can be actively monitored or reported by each storage node;

Step 3: Since the control node has stored the resource usage of each node, it allocates space on the storage node according to the storage resource status of the storage node, the location of the customer node, and the size of the requested storage capacity space, and provides the allocated storage space to For customer nodes, in order to achieve high availability, the control node can allocate resources of multiple storage nodes to the same customer as redundancy. In order to avoid the subsequent space being occupied, the control node can send the storage node in advance after allocation to reserve the corresponding space size;

If the client node requests 2G space, the management node allocates according to the resource usage of each storage node:

Storage node 1, port number 5, allocated block data identifier lv0 (corresponding to block data 1-10000 on storage node 1);

Storage node 2, port number 4, allocation block data identifier lv1 (corresponding to data 2000-12000 on storage node 2);...

In the specific allocation, multiple redundant storage nodes can be allocated to the same client node according to the situation.

Step 4: The control node sends the allocated storage node resource information to the client node, and the control node stores the storage node resource information allocated to the client node, such as storage node ID, location information, space size, block data area ID, etc. wait;

Step 5: According to the returned results, the client node includes the address of the storage node and the identification of the storage area specified for the specific service, such as the area identification lv0, lv1, etc. corresponding to the block data allocated in step 3. Based on these information, the storage node of the service Accesses from different client nodes can be distinguished, and the client node then builds a local block device driver for the resource information of these storage nodes and provides it to the application program, and the application process is completed. The control node manages each storage node to determine the size of the space that can be stored.

This embodiment only uses the application process as an example to illustrate, specifically, before the client node actively initiates the release process, these storage resources are always available to the user. The release process is similar to the application, so no examples are given here.

Figure 11 shows the architecture used by the local block device constructed in Figure 10 . After the storage resource is applied for in Figure 9 and Figure 10, the subsequent resource usage process, such as reading and writing, only occurs between the client node and the storage node (service node) that provides the service, which is different from the existing technology that must Depending on where the block device is located in the metadata server, there is no performance bottleneck in this embodiment.

As described in the embodiment of Figure 8, in Figure 11, after the client node establishes a local block unit, it can be used as a local block device driver, provided to the application program, and can directly pass the standard block device interface to the local block device For block number access or file access by creating a file system, database system, etc., it has better system scalability. A standard block device interface can be used between the local block unit in FIG. 11 and the upper layer. After the local block device receives the block access command, it can perform queuing and other processing to improve the access efficiency. For details, refer to the following figure 12. The processed block access command is mapped by the local block unit to the corresponding storage node access and sent to the corresponding Storage nodes.

This embodiment uses the added local block device driver for mapping access, which can be applied to virtual machine technology. The system shown in Figure 11 can provide a block device with permanent storage and high availability for the virtual machine of the big cloud elastic computing service. Serve.

In the way of distributed file system in the prior art, users can see other global resources, and there are hidden dangers in security and reliability. In the data storage access service provided by the present invention, users can only access allocated storage resources, and cannot see the status of other global storage resources. The multi-tenant isolation capability for public services is stronger, and the security and reliability are higher.

In the use of virtual machines in the prior art, each client virtual machine has only one disk. This disk actually comes from the "virtual machine image" file and runs the guest operating system. When the virtual machine is destroyed by the instance, all disks stored on the disk data no longer exists. After the present invention provides data storage services, the virtual machine can have several mapped local block devices to use. The local block device provided by the present invention is independent of the virtual machine image and has nothing to do with the actual operating system of the client. Even after the virtual machine is destroyed, it will still exist, and the block device can be re-acquired by the management node to be used by other virtual machines of the user. Even a local block device can be used by multiple virtual machines at the same time, which greatly improves the quality of service. The local block device of the service is installed on the host (Host) or server that provides the virtual machine service, and this block device is directly provided to the virtual machine specified by the user for use. The virtual machine itself does not need any changes or drivers to be installed. Use this "local" disk. The local block device allocated to the user by each embodiment of the present invention can allow the user to fully customize independently, and has better flexibility.

For Figure 11, if the client node does not create a file system or database system for the local block unit and directly accesses the block data, then according to the corresponding relationship between the block number stored in the local block unit and the storage node resource information according to the block number identification, the mapping is as follows: The process of accessing the block data of a certain storage node and returning the access result is relatively simple, so we will not draw an example here.

In order to compare with the centralized and distributed file system access in the prior art, the process of using the local block device of the present invention for file access is shown below in FIG. 12 and FIG. 13 .

FIG. 12 is a flow chart of the read access operation of FIG. 11 , and FIG. 13 is a flow chart of the write access operation of FIG. 11 .

In the embodiment shown in Figure 12 and Figure 13, the file system created by the client node stores the block number corresponding to the file descriptor. When accessing data through the file system, the application of the client node obtains the corresponding block number through the file name (descriptor) of the file system, and then accesses the block data to the local block device, and the local block device driver maps the block data For the block data access of the corresponding storage node, the data result is returned.

The following explains the meaning of the block number corresponding to the file descriptor and the block data identifier mapped by the local block device driver. The block number corresponding to the file descriptor is a number such as 0x0500001 on the local block device; the block data identifier is the storage node The identifier corresponding to one of the block devices on the network, such as lv0, lv1, and lv2.

For example, when a user accesses the block number 0x0500001 on his local block device through a file descriptor, according to the corresponding relationship between the block number stored by the local block device driver and the resource information of the storage node, the actual mapping is to access the storage node, such as host host07( 0x50001 on lv1 (block data area identifier) on address).

When performing write operations, the local block device driver can send them to all storage nodes allocated to the user according to the block number for parallel write operations. During the read operation, it can only be sent to the storage node corresponding to the block number.

Compared with the prior art, when the client application in Figure 12 and Figure 13 accesses the local block device driver, the process before reaching the local block device driver is similar to the scenario of the non-distributed application in Figure 2 and Figure 3, so it is similar to the traditional The application is compatible, but the subsequent access process is different: the local block device driver maps the access to the local block device to the storage node that provides the service according to the storage node resource information obtained at the time of application, as shown in Figure 12 and Figure 13, the local After the block device driver receives block data access, it can perform the following processing, including:

1. Cache block device data access before reaching the preset access number or time threshold;

2. Sort the cached block device access (such as sorting according to the accessed block data identifier, which is convenient to improve the access efficiency of accessing the same storage node);

3. According to the characteristics of user access, determine whether it is necessary to pre-read some of the user's unrequested data; for example, according to the user's access history, the continuous access can be pre-read; if the user access is non-continuous, the pre-read data is not required ;

4. Pack and compress block access operation instructions, and send them to the corresponding storage service nodes according to the information obtained during the application (select one for the read operation, and send the write operation to multiple service nodes concurrently).

Compared with the centralized file in the prior art, the above embodiments of the present invention can provide storage nodes in a distributed manner, and the local block device driver finds the corresponding storage node according to the pre-resource application; The file system and local block device of the node are similar to the functions of the existing centralized file system. At this time, the local block device of the client node is equivalent to a forwarder and forwarded to the storage node.

The above method embodiments have the following advantages:

1. Compared with the storage system of the distributed file system in the prior art, the file system layer of the distributed file system of the prior art has one more network access process of the metadata server node, but the present invention can provide a lower-level native Block device access interface, users can choose to directly use the local block device driver or create their own local file system according to the application, which provides flexibility and reduces the overhead of the file system layer.

2. Through the mapping of the client node, the client node maps the block data access to the physical storage resources of one or more storage nodes, and there is no need to request the control node during the access process; the scheduling management of the control node only plays a role in the allocation of resources and failures role, therefore, there is no performance bottleneck.

3. Compared with the centralized block device, the distributed architecture has better system scalability, and because it can centrally utilize the idle storage space in the cluster, it has better resource utilization and higher cost performance.

4. Since the user only accesses the resources assigned to the user when using storage resources, other

Storage resources are invisible to users and have stronger isolation capabilities for multi-tenants.

If the control node assigns multiple storage node resource information to the same client node as redundancy, the storage data of multiple storage nodes obtained through mapping is consistent, and when several of the storage nodes suddenly fail (failure), the remaining The information of valid storage nodes is still complete and can provide services. After the control node detects the failure of the storage node, it can notify the remaining valid nodes according to the resource usage of each storage node that has been stored before, and copy the information of the same storage data to the newly added storage node, and the mapping on the client node Point to the new storage node for reconfiguration to achieve high availability.

For example, the mapping relationship between users and storage nodes of each client node stored by the control node can be shown in Table 1 below:

Table 1 Mapping relationship between users and storage nodes (service nodes)

Specifically, the processing process after a certain storage node fails may include:

Step 1: Remove the failed storage node, add a new storage node, and the control node allocates the same size partition on the newly added storage node; and the control node notifies the valid node with the same data as the failed node, so that subsequent data recovery can be performed ;

Step 2: After receiving the notification from the control node, the active node still serving takes a complete snapshot of the required block data, and uses the snapshot data to restore the data in the newly added node;

Step 3: Since the data may be updated during the execution of step 2, an incremental snapshot is taken of the data required to be restored in the effective node in step 2 (updated data during the execution of step 2), and the snapshot data Restore data in the newly added node. If the size of the incremental snapshot is larger than the preset threshold, repeat step 3, otherwise go to step 4;

Step 4: If the size of the last incremental snapshot is smaller than the threshold but still not 0, you can lock and quickly recover the data of the last incremental snapshot so that the new storage node is completely consistent with the data of the storage node taking the snapshot. During the locking period, Can use the client node's local block device driver to cache data changes;

Step 5: Unlock the storage node in step 4 and restore normal services.

In the case of using multiple storage nodes in this embodiment, when some storage nodes fail, the remaining nodes can be used to continue to provide continuous services, and the data can be automatically copied to other storage nodes to maintain service availability. The scheduling management only plays a role in the allocation of resources and failures, and there is no performance bottleneck.

Fig. 14 is a schematic diagram of an embodiment of a client node according to the present invention. As shown in Figure 14, the client node in this embodiment includes one or more local block units, where the local block units include:

an interface module, configured to receive an access request including a block number identifier;

The storage module is used to store the corresponding relationship between the block number of the local block unit and the resource information of the storage node providing the storage service; the resource information of the storage node includes the information of the storage node providing the storage service and the block data identification of the storage node;

The processing module is used to search the storage module according to the block number identifier, obtain the storage node location corresponding to the block number identifier and the block data identifier, and send the block data access request to the storage node

Among them, the communication module is a module outside the local block unit of the client node, which is used to apply for a block data activation request to the control node through the network, and receive the storage node resource information allocated by the control node. The local block unit is based on the storage node resource information. A local block device driver is established by information. When the interface module of the local block unit receives a data access request, the corresponding storage node block data access request is sent to the communication module after processing by the processing module, and the access result is returned through the communication module.

The interface modules of the local block unit may include:

a direct access sub-module for receiving access requests directly including block number identification via a block device interface; and/or

The system access sub-module is used to receive the file name/table data access request through the file/database system interface, analyze the block number identification corresponding to the file name/table data, and send the access request including the block number identification to the processing module.

Processing modules may also include:

The block access processing sub-module is used to pre-process block data access to improve access efficiency. The pre-processing includes any combination of the following: when block data access is not greater than the preset access number or time threshold, cache block data access; Data access is cached and sorted; according to the characteristics of block data access, it is determined whether to pre-read block data that is not requested to be accessed; package and compress block data access.

For the specific access process of the client in this embodiment, reference may be made to the relevant descriptions of the above method embodiments, and the same or similar content will not be described repeatedly here.

Fig. 15 is a schematic diagram of an embodiment of a storage control node according to the present invention. As shown in Figure 15, the specific functions of this embodiment can refer to the control node functions in the above method embodiments, as shown in the figure, this embodiment includes:

The interface module is used to receive the block data activation/release request of the client node;

The control module is configured to allocate corresponding storage node resource information according to the storage resource status information of one or more storage nodes and/or the received client node location information of the block data activation request; or release the corresponding storage node according to the block data release request resource information;

A storage module, configured to store resource information of storage nodes allocated by one or more storage nodes;

Wherein, the resource information of the storage node includes location information of the storage node and a block data identifier of the storage node.

The node device may also include:

The monitoring module is used to find an effective storage node in the storage module that has the same storage data as the failed storage node after detecting the failure of the storage node, and copy the storage data of the effective storage node to the newly established storage node.

Preferably, the storage module also stores the corresponding relationship between the client node and the allocated storage node resource information, as shown in Table 1; the interface module further receives an acquisition request including the client node identifier; the control module further searches the storage module according to the client node identifier To store node resource information and return the interface module.

The storage control node in this embodiment may be located between the client node and one or more storage nodes, or located on a storage node. For the specific control process, reference may be made to relevant descriptions of the foregoing method embodiments, and repeated descriptions of the same or similar content will not be repeated here.

The techniques described herein may be implemented by various means. For example, these techniques may be implemented in hardware, firmware, software, or a combination thereof. For hardware implementation, the control module of FIG. 15 and the processing module of ±14 can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic devices (PLDs), field programmable gate arrays ( FPGA), processor, controller, microcontroller, microprocessor, electronic device, other electronic unit designed to perform the functions described herein, or a combination thereof.

For a firmware and/or software implementation, the techniques may be implemented with modules (eg, procedures, steps, flows, and so on) that perform the functions described herein. Firmware and/or software codes may be stored in memory and executed by processors. Memory can be implemented within the processor or external to the processor.

The present invention can have a variety of specific implementations in different forms. The technical solution of the present invention is illustrated by taking Fig. 7-Fig. 15 as an example in conjunction with the accompanying drawings. In a specific process or embodiment structure, those of ordinary skill in the art should understand that the specific implementation solutions provided above are just some examples of various preferred usages, any storage node resources are mapped to local block devices, and Implementations of mapping the access to the local block device to the access to the corresponding storage node shall fall within the protection scope of the technical solution of the present invention.

Those of ordinary skill in the art can understand that all or part of the steps for realizing the above-mentioned method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the It includes the steps of the above method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

Finally, it should be noted that: the above is only a preferred embodiment of the present invention, and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, it still The technical solutions recorded in the foregoing embodiments may be modified, or some technical features thereof may be equivalently replaced. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (15)

1. a data access method is characterized in that, comprising:

Reception comprises the access request of block number sign;

Pre-established local module unit is searched the block number of said local module unit according to said block number sign and the corresponding relation of the memory node resource information of stores service is provided, and obtains corresponding memory node position of said block number sign and blocks of data sign;

Memory node position and blocks of data sign that said local module unit is corresponding according to said block number sign are sent the blocks of data access request to said memory node;

Wherein, said memory node resource information comprises the blocks of data sign that the memory node of stores service positional information and said memory node are provided.

2. data access method according to claim 1 is characterized in that, receives the access request that comprises the block number sign and also comprises before:

Send the blocks of data that comprises the memory capacity size to Control Node and open request;

Receive said Control Node according to the storage resource status information of memory node and/or send the client node positional information that said blocks of data is opened request, distribute the corresponding memory node resource information of said memory capacity size;

Set up said local module unit according to said memory node resource information.

3. data access method according to claim 1 is characterized in that, receives the access request that comprises the block number sign and comprises before:

Receive the access request that comprises the block number sign through the block device interface; And/or

Receive filename/table data access request through file/Database Systems interface; Said file/Database Systems interface is resolved the corresponding block number sign of said filename/table data, sends the access request that comprises said block number sign to said local module unit.

4. data access method according to claim 1 is characterized in that, before said memory node sends the blocks of data access request, also comprises:

The blocks of data access request is carried out preliminary treatment, and said preliminary treatment comprises following combination in any:

When said blocks of data visit is not more than preset visit quantity or time threshold, the said blocks of data visit of buffer memory; Visit is carried out buffer memory and is sorted to blocks of data; According to the blocks of data accessing characteristic, determine whether to read in advance not ask the blocks of data of visiting; The said blocks of data visit of packing compression.

5. according to each described data access method of claim 1-4, it is characterized in that said local module unit is that the virtual local block device driving of distributed storage service is provided or is the virtual machine that application service is provided.

6. data access method according to claim 5 is characterized in that, also comprises:

After said virtual machine is destroyed, send the request of obtaining of the memory node resource information that comprises the client node sign to Control Node;

Receive the memory node resource information that said Control Node is returned according to said client node sign.

7. a client node is characterized in that, comprising:

One or more local module unit, wherein, said local module unit comprises:

Interface module is used to receive the access request that comprises the block number sign;

Memory module, the corresponding relation that is used to store the block number of local module unit and the memory node resource information of stores service is provided; Said memory node resource information comprises the blocks of data sign that the memory node of stores service information and said memory node are provided;

Processing module is used for searching said memory cell according to said block number sign, obtains corresponding memory node position of said block number sign and blocks of data sign, sends the blocks of data access request to said memory node.

8. client node according to claim 7 is characterized in that, said interface module comprises:

Directly the visit submodule is used for receiving the access request that comprises the block number sign through the block device interface; And/or

System's visit submodule is used for receiving filename/table data access request through file/Database Systems interface, resolves the corresponding block number sign of said filename/table data, sends the access request that comprises said block number sign to said processing module.

9. according to claim 7 or 8 described client nodes, it is characterized in that said processing module also comprises:

The block access processing sub is used for preliminary treatment is carried out in said blocks of data visit, and said preliminary treatment comprises following combination in any: when said blocks of data visit is not more than preset visit quantity or time threshold, and the said blocks of data visit of buffer memory; Said blocks of data visit to buffer memory is sorted; According to the blocks of data accessing characteristic, determine whether to read in advance not ask the blocks of data of visiting; The said blocks of data visit of packing compression.

10. according to claim 7 or 8 described client nodes, it is characterized in that said local module unit is the virtual machine that the virtual local block device driving of distributed storage service is provided or application service is provided.

11. a storage Control Node is characterized in that, comprising:

Interface module is used to receive blocks of data and opens/discharge request;

Control module is used for opening the client node positional information of request according to the storage resource status information of one or more memory node and/or the said blocks of data of reception, distributes the corresponding memory node resource information of the said blocks of data request of opening; Perhaps discharge request and discharge corresponding memory node resource information according to said blocks of data;

Memory module is used to store the memory node resource information that one or more memory node has distributed;

Wherein, said memory node resource information comprises the blocks of data sign of memory node information and said memory node.

12. storage Control Node according to claim 11 is characterized in that, also comprises:

Monitoring modular after being used to monitor memory node and losing efficacy, is searched in the said memory module effective memory node that identical storage data are arranged with the memory node of said inefficacy, and the storage data of said effective memory node are copied to newly-established memory node.

13. storage Control Node according to claim 11 is characterized in that, said memory module is also stored the corresponding relation of client node and the memory node resource information of having distributed;

Said interface module further receives the request of obtaining that comprises the client node sign;

Said control module further according to said client node sign from said memory module search to the memory node resource information, and return said interface module.

14. according to each described storage Control Node of claim 11-13, it is characterized in that, between client node and one or more memory node, perhaps be positioned on the said memory node.

15. a data-storage system is characterized in that, comprises each described storage Control Node of each described client node of claim 7-10 and claim 11-14.

Images