Patent Application
20170177489
The present invention relates to the field of information technologies, and in particular, to a data deduplication system and method in a storage array.
A storage array generally includes one engine, and one engine includes two controllers, which is generally referred to as a dual-controller structure. As shown in
The foregoing data deduplication process in a storage array consumes CPU computing power of a controller and memory resources of the controller, and affects performance of the storage array severely.
Embodiments of the present invention provide a data deduplication method and a storage array.
According to a first aspect, an embodiment of the present invention provides a data deduplication method, where the method is applied to a storage array, where the storage array includes a switching device, a first controller, and a cache device, where the first controller and the cache device are connected to the switching device; and the switching device is connected to a storage unit in the storage array, and the method comprising:
receiving, by the first controller, an eigenvalue of a to-be-deduplicated data block from the cache device, and searching an eigenvalue index set of data blocks for the eigenvalue of the to-be-deduplicated data block;
when the eigenvalue of the to-be-deduplicated data block is not found in the eigenvalue index set of data blocks, obtaining, by the first controller, a cache address of the to-be-deduplicated data block in the cache device via the switching device;
sending, by the first controller, a read data instruction to a controller of a target storage unit via the switching device, where the read data instruction carries an identifier of the cache device and the cache address;
reading, by the controller of the target storage unit, the to-be-deduplicated data block from the cache address via the switching device using the identifier of the cache device and the cache address; and
storing, by the controller of the target storage unit, the to-be-deduplicated data block into the target storage unit.
With reference to the first aspect of the present invention, in a first possible implementation manner, the method further includes:
sending, by the controller of the target storage unit, a storage address in the target storage unit to the first controller via the switching device, where the storage address in the target storage unit includes an identifier of the controller of the target storage unit and a logical storage address for storing the to-be-deduplicated data block in the target storage unit; and
creating, by the first controller, an eigenvalue index of the to-be-deduplicated data block in the eigenvalue index set of data blocks, where the eigenvalue index of the to-be-deduplicated data block includes the eigenvalue of the to-be-deduplicated data block and the storage address in the target storage unit.
With reference to the first aspect of the present invention, in a second possible implementation manner, the storage array further includes a second controller, where the second controller is connected to the switching device, where the second controller stores an address of the to-be-deduplicated data block, and the second controller is a home controller of a target logical unit in which the to-be-deduplicated data block is located, and the receiving, by the first controller, an eigenvalue of a to-be-deduplicated data block from the cache device specifically includes:
sending, by the cache device, the eigenvalue of the to-be-deduplicated data block to the second controller via the switching device;
determining, by the second controller, that a home controller of the eigenvalue of the to-be-deduplicated data block is the first controller; and
sending, by the second controller, the eigenvalue of the to-be-deduplicated data block to the first controller via the switching device.
With reference to the second possible implementation manner of the first aspect of the present invention, in a third possible implementation manner, the method further includes: when the eigenvalue of the to-be-deduplicated data block is not found in the eigenvalue index set of data blocks, sending by the first controller, a notification to the second controller via the switching device, where the notification carries the storage address in the target storage unit; and
establishing, by the second controller according to the notification, a correspondence among the address of the to-be-deduplicated data block, the eigenvalue of the to-be-deduplicated data block, and the storage address in the target storage unit.
With reference to the second possible implementation manner of the first aspect of the present invention, in a fourth possible implementation manner, the method further includes: establishing, by the second controller, a correspondence among the address of the to-be-deduplicated data block, the eigenvalue of the to-be-deduplicated data block, and an address of the first controller.
According to a second aspect, an embodiment of the present invention provides a storage array, where the storage array includes a switching device, a first controller, and a cache device, where the first controller and the cache device are connected to the switching device; and the switching device is connected to a storage unit in the storage array;
the first controller is configured to receive an eigenvalue of a to-be-deduplicated data block from the cache device, and search an eigenvalue index set of data blocks for the eigenvalue of the to-be-deduplicated data block;
when the eigenvalue of the to-be-deduplicated data block is not found in the eigenvalue index set of data blocks, the first controller is further configured to obtain a cache address of the to-be-deduplicated data block in the cache device via the switching device;
the first controller is further configured to send a read data instruction to a controller of a target storage unit via the switching device, where the read data instruction carries an identifier of the cache device and the cache address;
the controller of the target storage unit is configured to read the to-be-deduplicated data block from the cache address via the switching device according to the identifier of the cache device and the cache address; and
the controller of the target storage unit is further configured to cause the storage of the to-be-deduplicated data block into the target storage unit.
With reference to the second aspect of the present invention, in a first possible implementation manner, the controller of the target storage unit is further configured to send a storage address in the target storage unit to the first controller via the switching device, where the storage address in the target storage unit includes an identifier of the controller of the target storage unit and a logical storage address for storing the to-be-deduplicated data block in the target storage unit; and
the first controller is further configured to establish an eigenvalue index of the to-be-deduplicated data block in the eigenvalue index set of data blocks, where the eigenvalue index of the to-be-deduplicated data block includes the eigenvalue of the to-be-deduplicated data block and the storage address in the target storage unit.
With reference to the second aspect of the present invention, in a second possible implementation manner, the storage array further includes a second controller, where the second controller is connected to the switching device; the second controller is configured to store an address of the to-be-deduplicated data block, and the second controller is a home controller of a target logical unit in which the to-be-deduplicated data block is located, and where the first controller receives the eigenvalue of the to-be-deduplicated data block from the cache device specifically includes:
The cache device further configured to send the eigenvalue of the to-be-deduplicated data block to the second controller via the switching device;
the second controller further configured to determine, that a home controller of the eigenvalue of the to-be-deduplicated data block is the first controller; and
the second controller further configured to send the eigenvalue of the to-be-deduplicated data block to the first controller via the switching device.
With reference to the second possible implementation manner of the second aspect of the present invention, in a third possible implementation manner, when the eigenvalue of the to-be-deduplicated data block is not found in the eigenvalue index set of data blocks, the first controller is further configured to send a notification to the second controller via the switching device, where the notification carries the storage address in the target storage unit, and
the second controller is further configured to establish a correspondence among the address of the to-be-deduplicated data block, the eigenvalue of the to-be-deduplicated data block, and the storage address in the target storage unit, according to the notification.
With reference to the second possible implementation manner of the second aspect of the present invention, in a fourth possible implementation manner, the second controller is further configured to establish a correspondence among the address of the to-be-deduplicated data block, the eigenvalue of the to-be-deduplicated data block, and an address of the first controller.
According to the data deduplication method and the storage array provided in the embodiments of the present invention, a controller is connected to a cache device via a switching device, a first controller receives an eigenvalue of a to-be-deduplicated data block from the cache device, and searches an eigenvalue index set of data blocks for the eigenvalue of the to-be-deduplicated data block, and when the same eigenvalue is not found, the first controller sends a cache address of the to-be-deduplicated data block in the cache device to a controller of a target hard disk, and the controller of the target hard disk reads the to-be-deduplicated data block from the cache address of the to-be-deduplicated data block. The cache device implements calculation of a fingerprint of the to-be-deduplicated data block, thereby saving computing resources of the controller. During the process of storing the to-be-deduplicated data block into the target hard disk, the first controller only provides the cache address of the to-be-deduplicated data block, and the controller of the target hard disk directly reads the to-be-deduplicated data block from the cache address, thereby saving computing resources and memory resources of the first controller and improving performance of the storage array.
To describe the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings required for describing the embodiments. The accompanying drawings in the following description show merely some embodiments of the present invention, and other drawings may still be derived from these accompanying drawings.
The following clearly describes the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention.
A storage array provided in the embodiments of the present invention, such as a storage array shown in
The cache device M may be specifically a storage device formed by a volatile storage medium or a non-volatile storage medium, such as a phase change memory (PCM), or may be another non-volatile storage medium that is suitable to be used as a cache device, which is not limited by the embodiments of the present invention. The cache device M is configured to cache data. The following describes the cache device M with reference to specific embodiments of the present invention. In the embodiments of the present invention, that the switching device A is a PCIe switching device, the switching device B is a PCIe switching device, and the hard disk is an SSD with a PCIe protocol interface is used as an example.
In the storage array shown in
The write data request received by the input/output manager A carries an address of to-be-written data. The address of the to-be-written data includes an identifier of a target LU of to-be-written data, a logical block address (LBA) of the to-be-written data, and a length of the to-be-written data. The input/output manager A sends the write data request to the controller A. The controller A receives the write data request, and determines, according to the identifier of the target LU of the to-be-written data in the address of the to-be-written data, whether the controller A is a home controller of the target LU.
When the controller A is the home controller of the target LU, that is, the target LU is generated by the controller A by virtualizing hard disks and provided for the host. The controller A determines a cache device used to cache the to-be-written data, which is the cache device M in this embodiment of the present invention. An implementation manner is: The controller A instructs, according to the write data request, the cache device M to allocate a cache address to to-be-written data, and the cache device M allocates a cache address according to a length of the to-be-written data. The controller A obtains the cache address allocated by the cache device M to the to-be-written data (the cache address allocated by the cache device M to the to-be-written data is hereinafter referred to as a cache address M, and in an implementation manner, the cache address includes a start address and a length). The controller A sends an identifier of the cache device M and the cache address M to the input/output manager A via the PCIe switching device A. The input/output manager A receives the identifier of the cache device M and the cache address M that are sent by the controller A, and writes the to-be-written data to the cache address M according to the identifier of the cache device M and the cache address M (or may directly write the to-be-written data to the cache address M). The controller A obtains only the cache address M allocated to the to-be-written data, and the input/output manager A directly writes the to-be-written data to the cache address M via the PCIe switching device A, which, compared with the prior art, saves CPU computing resources of the controller A and memory resources of the controller A and improves data writing efficiency.
The controller A establishes a correspondence among the address of the to-be-written data, the identifier of the cache device M, and the cache address M, and therefore, when reading the to-be-written data, the controller A sends the cache address M of the to-be-written data to the input/output manager A, and the input/output manager A may read the to-be-written data from the cache address M of the to-be-written data (or may directly read the to-be-written data from the cache address M of the to-be-written data), thereby saving CPU computing resources of the controller A and memory resources of the controller A and improving data reading efficiency.
After conditions are satisfied, if the storage array does not perform data deduplication, the cache device M stores the to-be-written data into a target SSD of the storage array. The target SSD refers to an SSD for storing the to-be-written data. A specific process of writing the to-be-written data into the target SSD may be: The controller A sends the identifier of the cache device M and the cache address M to a controller of the target SSD via the PCIe switching device A or the PCIe switching device B. The controller of the target SSD directly reads the to-be-written data from the cache address M via the PCIe switching device A or the PCIe switching device B according to the identifier of the cache device M and the cache address M, and stores the to-be-written data. The controller of the target SSD sends a storage address of the to-be-written data in the target SSD to the controller A via the PCIe switching device A or the PCIe switching device B. The storage address of the to-be-written data in the target SSD includes an identifier of the controller of the target SSD and a logical storage address for storing the to-be-written data in the target SSD. The controller A establishes a correspondence between the address of the to-be-written data and the storage address of the to-be-written data in the target SSD.
The foregoing process is specifically shown in
Step 301: A host sends a write data request to an input/output manager A.
The input/output manager A is an input/output receiving management device in a storage array, and is responsible for receiving a data operation request sent by the host and forwarding the data operation request to a controller. In an embodiment of the present invention, the host sends a write data request that carries an address of to-be-written data to the input/output manager A. For example, the small computer system interface (SCSI) protocol, that is, a SCSI protocol write data request, may be used for the write data request. Alternatively, other protocols may also be used, which is not limited by this embodiment of the present invention.
Step 302: Send the write data request to a controller A.
In this embodiment of the present invention, the input/output manager A generally communicates with a specific controller. The input/output manager A may establish a correspondence with a controller in multiple manners, for example, according to a load of the controller, or according to a specific path selection algorithm, which is not limited by the present invention. The input/output manager A receives the write data request, and sends the write data request to the controller A via a PCIe switching device A or a PCIe switching device B. In this embodiment of the present invention, that the input/output manager A receives a write data request and sends the write data request to the controller A via the PCIe switching device A is used as an example.
Step 303: The controller A obtains a cache address of to-be-written data.
The controller A receives the write data request sent by the input/output manager A, and determines a cache device used to cache the to-be-written data, which is a cache device M in this embodiment of the present invention. In an implementation manner, the cache device M allocates a segment of cache addresses to the controller A. In the segment of cache addresses, the controller A allocates a cache address M to the to-be-written data according to the length of the to-be-written data. In another implementation manner, the controller A sends an instruction to the cache device M via the PCIe switching device A or the PCIe switching device B, where the instruction carries the length of the to-be-written data and instructs the cache device M to allocate a cache address to the to-be-written data. The controller A obtains the cache address M.
Step 304: Send an identifier of the cache device M and the cache address M.
The controller A obtains the cache address M, and sends the identifier of the cache device M and the cache address M to the input/output manager A via the PCIe switching device A, where the identifier of the cache device M is a device address.
Step 305: The host sends the to-be-written data to the input/output manager A.
The input/output manager A receives the identifier of the cache device M and the cache address M that are sent by the controller A, and receives the to-be-written data sent by the host.
Step 306: Write the to-be-written data to the cache address M.
The input/output manager A directly writes the to-be-written data to the cache address M via the PCIe switching device A according to the identifier of the cache device M and the cache address M. The input/output manager A receives, via the PCIe switching device A, a response indicating that the to-be-written data is written successfully that is sent by the cache device M. The input/output manager A sends a response indicating that the write data request is completed to the host, and notifies the host that a write request operation is complete.
Step 307: Notify the controller A that the to-be-written data is written to the cache address M.
The input/output manager A writes the to-be-written data to the cache address M successfully, and notifies the controller A that the to-be-written data is written to the cache address M.
Step 308: The controller A establishes a correspondence among an address of to-be-written data, the cache device M, and the cache address M.
The controller A receives the notification sent by the input/output manager A, and establishes a correspondence among the address of the to-be-written data, the cache device M, and the cache address M.
The cache device M allocates the cache address M to the to-be-written data, thereby establishing a correspondence between the address of the to-be-written data and the cache address M. The cache device M may obtain the address of the to-be-written data from a cache address allocation instruction sent by the controller A, and after allocating the cache address M, the cache device M establishes a correspondence between the address of the to-be-written data and the cache address M. In another embodiment, the cache device M is an exclusive cache device of a target LU, and is only used to cache data of the target LU, and therefore, the cache device M saves the correspondence among the target LU, an LBA in the target LU, and the cache address by default. The cache device M saves the correspondence among the target LU, the LBA in the target LU and a segment of cache addresses of the cache device M by default. In this segment of cache addresses, the cache device M allocates the cache address M to the to-be-written data.
To improve reliability of the storage array and to cache multiple copies of the to-be-written data, in the prior art shown in
The input/output manager A receives the identifier of the cache device M and the cache address M, and the identifier of the cache device N and the cache address N. The input/output manager A directly writes the to-be-written data to the cache address M via the PCIe switching device A according to the identifier of the cache device M and the cache address M; and the input/output manager A directly writes the to-be-written data to the cache address N via the PCIe switching device A according to the identifier of the cache device N and the cache address N. The input/output manager A receives, via the PCIe switching device A, a response indicating that the to-be-written data is successfully written to the cache address M, and instructs the controller A to establish a correspondence among the address of the to-be-written data, the identifier of the cache device M, and the cache address M. Likewise, the controller A establishes a correspondence among the address of the to-be-written data, the identifier of the cache device N, and the cache address N.
In another embodiment, the controller A sends the identifier of the cache device M and the cache address M to the input/output manager A via the PCIe switching device A. The input/output manager A receives the identifier of the cache device M and the cache address M. The input/output manager A directly writes the to-be-written data to the cache address M via the PCIe switching device A or the PCIe switching device B according to the identifier of the cache device M and the cache address M. The controller A sends a write data instruction to the cache device M via the PCIe switching device A or the PCIe switching device B, where the write data instruction carries the identifier of the cache device N and the cache address N. The cache device M caches the to-be-written data, and the cache device M directly writes the to-be-written data to the cache address N via the PCIe switching device A or the PCIe switching device B according to the write data instruction.
The controller A only needs to obtain the cache address M and the cache address N that are allocated to the to-be-written data, so that the input/output manager A implements writing of the to-be-written data into the cache device M and the cache device N, thereby saving CPU computing resources of the controller A and memory resources of the controller A and improving data writing efficiency.
In another case, the input/output manager A receives a write data request of a host. The write data request carries an address of to-be-written data. The input/output manager A sends the write data request to the controller A by means of forwarding by the PCIe switching device A. The controller A receives the write data request sent by the input/output manager A, and determines, according to an identifier of a target LU that is carried in the write data request, that the controller A is not a home controller of the target LU. A specific embodiment is shown in
Step 401: A host sends a write data request to an input/output manager A.
The host sends a write data request to the input/output manager A, where the write data request carries an address of to-be-written data.
Step 402: Send the write data request to a controller A.
In this embodiment of the present invention, the controller A is a home controller of the input/output manager A. The input/output manager A receives the write data request, and sends the write data request to the controller A via a PCIe switching device A or a PCIe switching device B. In this embodiment of the present invention, that the input/output manager A receives a write data request and sends the write data request to the controller A via the PCIe switching device A is used as an example.
Step 403: Determine that the controller A is not a home controller of a target LU.
The controller A receives the write data request sent by the input/output manager A, and determines, according to an identifier of the target LU of to-be-written data that is carried in the write data request, that the controller A is not a home controller of the target LU. The controller A queries a correspondence between a controller and an LU, and determines that a controller B is a home controller of the target LU.
Step 404: Send the write data request to a controller B.
The controller A sends the write data request to the controller B via the PCIe switching device A or the PCIe switching device B. In this embodiment, that the PCIe switching device B forwards the write data request to the controller B is used as an example.
Step 405: Obtain a cache address of the to-be-written data.
The controller B receives the write data request sent by the controller A, and determines a cache device used to cache the to-be-written data, which is the cache device M in this embodiment of the present invention. For a specific embodiment, refer to the manner in which the controller A obtains the cache address of the to-be-written data from the cache device M.
Step 406: Send an identifier of a cache device M and a cache address M to the controller A.
The controller B obtains the cache address M, and sends the identifier of the cache device M and the cache address M to the controller A via the PCIe switching device B. In another embodiment, the identifier of the cache device M and the cache address M may also be directly sent to the controller A via the PCIe switching device A or the PCIe switching device B.
Step 407: Send the identifier of the cache device M and the cache address M to the input/output manager A.
The controller A receives the identifier of the cache device M and the cache address M that are sent by the controller B, and sends the cache address M of the to-be-written data via the PCIe switching device.
Step 408: The host sends the to-be-written data to the input/output manager A.
The input/output manager A receives the identifier of the cache device M and the cache address M, and responds to the write data request sent by the host. The host sends the to-be-written data to the input/output manager A.
Step 409: Write the to-be-written data to the cache address M.
The input/output manager A receives the to-be-written data sent by the host, and directly writes the to-be-written data to the cache address M via the PCIe switching device A according to the identifier of the cache device M and the cache address M. The input/output manager A receives, via the PCIe switching device A, a response indicating that the to-be-written data is written successfully that is sent by the cache device M. The input/output manager A sends a response indicating that the write data request is completed to the host, and notifies the host that a write request operation is complete.
Step 410: Notify the controller B that the to-be-written data is written to the cache address M.
The input/output manager A writes the to-be-written data to the cache address M successfully, and notifies the controller B that the to-be-written data is written to the cache address M. This specifically includes that the input/output manager A forwards the notification to the controller A via the PCIe switching device A, and that the controller A forwards the notification to the controller B via the PCIe switching device B; or, the input/output manager A directly sends the notification to the controller B via the PCIe switching device A or the PCIe switching device B.
Step 411: The controller B establishes a correspondence among an address of to-be-written data, the cache device M, and the cache address M.
The controller B receives the notification sent by the input/output manager A, and establishes a correspondence among the address of the to-be-written data, the cache device M, and the cache address M.
For how the cache device M establishes the correspondence between the address of the to-be-written data and the cache address M, refer to the description in the foregoing embodiment, and details are not described herein again.
A cache device N allocates a cache address N to the to-be-written data, thereby establishing a correspondence between the address of the to-be-written data and the cache address N. The cache device N may obtain the address of the to-be-written data from a cache address allocation instruction sent by the controller A, and after allocating the cache address N, the cache device N establishes a correspondence between the address of the to-be-written data and the cache address N.
To prevent loss of the to-be-written data cached in the cache device M, when the to-be-written data needs multiple cache devices to serve as caches, in a scenario in which the controller A is not a home controller of the target LU of the to-be-written data, the input/output manager A sends a write data request to the controller B. For a process thereof, refer to the description in the foregoing embodiment. For a process of obtaining the cache address of the to-be-written data by the controller B, refer to the scenario in which the controller A is a home controller of the target LU of the to-be-written data and the controller A obtains cache addresses of multiple cache devices. For other steps, also refer to the description in the foregoing embodiment, and details are not described herein again.
After the host writes the data into the storage array, the host accesses the written data, that is, sends a read data request. A specific process is shown in
Step 501: Send a read data request.
A host sends a read data request to an input/output manager A, where the read data request carries an address of to-be-read data. The address of the to-be-read data includes an identifier of a logical unit LU in which the to-be-read data is located, an LBA of the to-be-read data, and a length of the to-be-read data. Specifically, the host may send the read data request to the input/output manager A by using the SCSI protocol, which is not limited by the present invention. For ease of description, the to-be-read data here is the to-be-written data described above.
Step 502: Send the read data request to a controller A.
The input/output manager A receives the read data request sent by the host, and sends the read data request to the controller A via a PCIe switching device A.
Step 503: The controller A sends an identifier of a cache device M and a cache address M to the input/output manager A.
When the controller A is a home controller of the LU in which the to-be-read data is located and the to-be-read data is cached in a cache device such as the cache device M, a correspondence among the address of the to-be-read data, an identifier of the cache device, and the cache address is queried according to the read data request, and the cache address M used to cache the to-be-read data in the cache device M is determined. When the to-be-read data is still cached in the cache device M, the cache address of the to-be-read data in the cache device M is the cache address M. The controller A sends an identifier of the cache device M and the cache address M to the input/output manager A via the PCIe switching device A.
Step 504: Read the to-be-read data from the cache address M.
The input/output manager A directly reads the to-be-read data from the cache address M via the PCIe switching device A according to the identifier of the cache device M and the cache address M.
Step 505: Return the to-be-read data.
The input/output manager A reads the to-be-read data from the cache address M, and returns the to-be-read data to the host.
When the input/output manager A sends a to-be-read data query request to the controller A via the PCIe switching device A according to the read data request, and the controller A is not a home controller of the LU in which the to-be-read data is located, the controller A queries a correspondence between the LU in which the to-be-read data is located and the home controller, and determines that a controller B is a home controller of the LU in which the to-be-read data is located. The controller A sends the to-be-read data query request to the controller B via a PCIe switching device B. That the foregoing to-be-written data is still the to-be-read data mentioned here is used as an example. Therefore, the address of the to-be-read data is the address of the to-be-written data described above. When the to-be-read data is still cached in the cache device M, the cache address of the to-be-read data in the cache device M is the cache address M. The controller B queries the correspondence among the address of the to-be-written data, the identifier of the cache device M, and the cache address M, determines the identifier of the cache device M that caches the to-be-read data and the cache address M, and sends the identifier of the cache device M and the cache address M to the controller A via the PCIe switching device B. The controller A sends the identifier of the cache device M and the cache address M to the input/output manager A via the PCIe switching device A. The controller B may also directly send the identifier of the cache device M and the cache address M to the input/output manager A via the PCIe switching device A or the PCIe switching device B. For a subsequent read operation, refer to the read operation in the foregoing embodiment, and details are not described herein again.
That the foregoing to-be-written data is still the to-be-read data mentioned here is used as an example. Therefore, an address of to-be-read data is the address of the to-be-written data described above. When the to-be-read data is already stored in a target SSD, a home controller of the LU in which the to-be-read data is located queries a correspondence between the address of the to-be-read data (the address of the to-be-written data) and a storage address of the to-be-read data in the target SSD, obtains the storage address of the to-be-read data in the target SSD, and sends the storage address of the to-be-read data in the target SSD to the input/output manager A via the PCIe switching device A or the PCIe switching device B. The storage address of the to-be-read data in the target SSD includes an identifier of a controller of the target SSD and a logical storage address of the to-be-read data in the target SSD. The input/output manager A directly reads the to-be-read data from the logical storage address of the to-be-read data in the target SSD via the PCIe switching device A or the PCIe switching device B according to the storage address of the to-be-read data in the target SSD.
In the foregoing embodiment, when the to-be-read data is partly saved in the target SSD and partly cached in the cache device M in this embodiment of the present invention, as described above, the input/output manager A directly reads data from the cache address via the PCIe switching device A or the PCIe switching device B according to the cache address of the to-be-read data in the cache device; and the input/output manager A directly reads data from the logical storage address in the target SSD via the PCIe switching device A or the PCIe switching device B according to the identifier of the controller of the target SSD and the logical storage address of the to-be-read data in the target SSD, which is not described in detail herein.
When multiple cache devices perform an operation of caching the to-be-read data, generally the home controller of the LU in which the to-be-read data is located returns, to the input/output manager A, an identifier of the primary cache device M that caches the to-be-read data and the cache address M. For other procedural operations, refer to the read operation in the foregoing embodiment, and details are not described herein again.
In the storage array, data deduplication is performed, which can save storage space and reduce storage costs. In the storage array shown in
Before the to-be-written data cached in the cache device M is stored into the SSD of the storage array, data deduplication is performed, which can save storage space effectively and improve a utilization rate of the storage space. Using the storage array shown in
In specific implementation, the comparing of eigenvalues of data blocks is implemented by a controller. Because data deduplication is performed in the storage array, and each unique data block has an eigenvalue, many eigenvalues are generated. To implement a balance between controllers in the storage array, each controller is responsible for comparing of eigenvalues of some data blocks according to a data block eigenvalue distribution algorithm such as a Hash distribution algorithm. In this way, each controller maintains only eigenvalue indexes of some unique data stored in the storage array according to the data block eigenvalue distribution algorithm, where the eigenvalue indexes of some unique data are referred to as an eigenvalue index set. The controller queries the eigenvalue index set for an eigenvalue of a data block that is to be written into the SSD, and determines whether the eigenvalue is the same as an eigenvalue in the eigenvalue index set. For example, the controller A needs to maintain an eigenvalue index set A according to the eigenvalue distribution algorithm, and therefore, the controller A is a home controller of every eigenvalue in the eigenvalue index set A; or, a controller in which an eigenvalue from the eigenvalue index set A is the same as an eigenvalue of a data block X is both a home controller of the eigenvalue of the data block X and a home controller of every eigenvalue in the eigenvalue index set A.
Specifically, the eigenvalue index set is formed by eigenvalue indexes, as shown in
In the storage array shown in
Step 701: A cache device M calculates an eigenvalue of a data block X.
A controller A sends to the cache device M an instruction to obtain the eigenvalue of the data block X, where the instruction carries an address of the data block X. The cache device M receives the instruction to obtain the eigenvalue of the data block X that is sent by the controller A. In a case, the cache device M stores a correspondence between the address of the data block X and a cache address B, and determines the data block X according to the address of the data block X that is carried in the instruction to obtain the eigenvalue of the data block X, calculates the eigenvalue of the data block X, and caches the eigenvalue of the data block X to a cache address X.
Step 702: Send the eigenvalue of the data block X to the controller A.
The cache device M obtains the eigenvalue of the data block X, and sends a response message of the eigenvalue of the data block X to a home controller A of an LU in which the data block X is located, where the response message of the eigenvalue of the data block X carries the eigenvalue of the data block X. In addition, the response message of the eigenvalue of the data block X further carries an identifier of the cache device M that caches the eigenvalue of the data block X and the cache address X of the eigenvalue of the data block X in the cache device M.
Step 703: Determine a home controller of the eigenvalue of the data block X according to an eigenvalue distribution algorithm.
Step 704: The controller A queries a local eigenvalue index set A.
When the controller A is the home controller of the eigenvalue of the data block X, the controller A queries the local eigenvalue index set A, and determines whether an eigenvalue same as the eigenvalue of the data block X exists in the eigenvalue index set A.
When an eigenvalue which is the same as the eigenvalue of the data block X exists in the eigenvalue index set A, steps 705a and 706a are performed. As shown in
Step 705a: The controller A updates a reference count in an index of an eigenvalue 1.
The reference count in the index of the eigenvalue 1 is 1, that is, only the data block A exists in the storage array. It is found that the eigenvalue of the data block X is the same as the eigenvalue 1, and therefore, the reference count is updated to2.
Step 706a: The controller A instructs the cache device M to delete the data block X.
The controller A instructs the cache device M to delete the data block X. The controller A establishes a correspondence between the address of the data block X and the eigenvalue of the data block X, or the controller A establishes a correspondence among the address of the data block X, the eigenvalue of the data block X, and a storage address of the data block A.
It is determined in step 704 that the data block X is a duplicate data block. Therefore, the data block X does not need to be saved into an SSD, and the cache device M is instructed to delete the data block X.
When no eigenvalue same as the eigenvalue of the data block X exists in the eigenvalue index set A, steps 705b, 706b, 707, 708, 709, and 710 are performed.
Step 705b: Obtain a cache address B of the data block X cached in the cache device M.
The controller A obtains the cache address B of the data block X from the cache device M via the PCIe switching device A according to the cache address X of the eigenvalue of the data block X in the cache device M.
Step 706b: Send an identifier of the cache device M and the cache address B to a controller of a target SSD.
The controller A obtains the identifier of the cache device M and the cache address B, and sends the identifier of the cache device M and the cache address B to the controller of the target SSD via the PCIe switching device A or the PCIe switching device B.
Step 707: The controller of the target SSD reads the data block X from the cache address B.
The controller of the target SSD receives the identifier of the cache device M and the cache address B, and directly reads, according to the identifier of the cache device M and the cache address B, the data block X from the cache address B via the PCIe switching device A or the PCIe switching device B.
Step 708: The controller of the target SSD sends a storage address of the data block X in the target SSD to the controller A.
The controller of the target SSD reads the data block X from the cache address B, and stores the data block X into the target SSD. The controller of the target SSD sends a storage address of the data block X in the target SSD to the controller A via the PCIe switching device A. The storage address of the data block X in the target SSD includes an identifier of the controller of the target SSD and a logical storage address for storing the data block X in the target SSD.
Step 709: The controller A establishes an eigenvalue index of the data block X.
The controller A receives the storage address of the data block X in the target SSD, establishes the eigenvalue index of the data block X, and sets the reference count to 1. The controller A establishes a correspondence among the address of the data block X, the eigenvalue of the data block X, and the storage address of the data block X in the target SSD. The controller A also needs to record the cache address X of the eigenvalue of the data block X. When the eigenvalue of the data block X is stored into the SSD, the controller A also needs to record the storage address of the eigenvalue of the data block X in the target SSD.
In another case, the controller A is not the home controller of the eigenvalue of the data block X but only a home controller of an LU in which the data block X is located. In this embodiment of the present invention, that a controller B is the home controller of the eigenvalue of the data block X is used as an example, and the controller A sends the eigenvalue of the data block X to the controller B via the PCIe switching device A or the PCIe switching device B. The controller B receives the eigenvalue of the data block X that is sent by the controller A, and queries an eigenvalue index set B of the controller B. When the controller B finds that an eigenvalue same as the eigenvalue of the data block X exists in the eigenvalue index set A, for example, an eigenvalue of a data block R is same as the eigenvalue of the data block X, the controller B instructs the cache device M to delete the data block X. This specifically includes that the controller B sends a delete instruction to the controller A via the PCIe switching device B. The controller A sends the delete instruction to the cache device M via the PCIe switching device A, and the cache device M deletes the data block X. The controller B updates the reference count of the index of the eigenvalue same as the eigenvalue of the data block X, that is, increases the reference count by 1. When the data block R is already stored in the SSD, the storage address of the data block R in an index of the data block R includes an identifier of a controller of the SSD that stores the data block R and a logical storage address for storing the data block R in the SSD. When the data block R is in the cache device, the storage address of the data block R in the index of the data block R includes the identifier of the cache device and a cache address. The controller A establishes a correspondence among the address of the data block X, the eigenvalue of the data block X, and the address of the home controller B of the eigenvalue of the data block X, and therefore, the controller A does not require a correspondence among the address of each data block, the eigenvalue of the data block, and the storage address of the data block, and an amount of data stored by the controller A is reduced effectively. Alternatively, the controller A establishes a correspondence among the address of the data block X, the eigenvalue of the data block X, and the storage address of the data block R. When reading the data block X subsequently, the controller A can directly determine the storage address of the data block R by querying the correspondence among the address of the data block X, the eigenvalue of the data block X, and the storage address of the data block R, and the input/output manager A directly reads the data block X from the storage address of the data block R via the PCIe switching device A or the PCIe switching device B, thereby improving data reading efficiency.
When the controller A is only the home controller of the LU in which the data block X is located, but is not the home controller of the eigenvalue of the data block X, the controller B finds that no eigenvalue same as the eigenvalue of the data block X exists in the eigenvalue index set B, the controller B obtains the cache address B of the data block X in the cache device M by sending a request to the controller A via the PCIe switching device B. The controller A sends the request to the cache device M via the PCIe switching device A. The cache device M sends the identifier of the cache device M and the cache address B to the controller B. The controller B sends the identifier of the cache device M and the cache address B to the controller of the target SSD via the PCIe switching device A or the PCIe switching device B (here the PCIe switching device A is used as an example). The controller of the target SSD directly reads the data block X from the cache address B via the PCIe switching device A or the PCIe switching device B according to the identifier of the cache device M and the cache address B, and stores the data block X into the target SSD. The controller of the target SSD sends the storage address of the data block X in the target SSD to the controller B via the PCIe switching device A or the PCIe switching device B. The controller B receives the storage address of the data block X in the target SSD, establishes the eigenvalue index of the data block X, and sets a reference count in the index to 1. The controller B also needs to record the cache address X of the eigenvalue of the data block X. When the eigenvalue of the data block X is stored into the SSD, the controller B also needs to record the storage address of the eigenvalue of the data block X in the SSD.
The controller B receives the storage address of the data block X in the target SSD, and sends a notification to the controller A. The notification carries the storage address of the data block X in the target SSD. The controller A establishes a correspondence among the address of the data block X, the eigenvalue, and the storage address of the data block X in the target SSD according to the notification sent by the controller B. In another embodiment, when the controller A is only the home controller of the LU in which the data block X is located, but not the home controller of the eigenvalue of the data block X, the controller A establishes a correspondence among the address of the data block X, the eigenvalue of the data block X, and the address of the controller B.
According to the storage array in this embodiment of the present invention, the cache device implements calculation of a fingerprint of the data block X, which saves computing resources of the controller. During a process of storing the data block X into the target SSD, the controller provides only the identifier of the cache device M and the cache address B, and the controller of the target SSD directly reads the data block X from the cache address B, which saves computing resources and memory resources of the controller and improves performance of the storage array.
Based on the storage array shown in
When the data written in the storage array shown in
In this embodiment of the present invention, in another implementation case, the concept of homing does not necessarily exist between an input/output manager and a controller. That is, the controller A is not a home controller of the input/output manager A. Each input/output manager saves a correspondence between an LU and a controller to which the LU is homed. The input/output manager queries, according to an identifier of a target LU that is carried in a data operation request, a correspondence between the identifier of the target LU and a home controller, to determine a home controller of the target LU, and directly sends the request to the home controller of the target LU via the PCIe switching device A or the PCIe switching device B. In addition, communication may be performed, via any PCIe switching device, between controllers, or between a controller and an SSD, or between an input/output manager and a controller, or between an input/output manager and an SSD, or between a cache device and a controller, or between a cache device and an SSD. In this embodiment of the present invention, a logical storage address for storing a data block X in a storage address in a target hard disk refers to a logical block address for storing the data block X in the target hard disk, and specifically refers to a logical block address for storing the data block X in the target SSD in this embodiment of the present invention.
Any cache device is connected to any hard disk via a switching device. Bidirectional communication is implemented between any two devices connected via any switching device. Any two switching devices are directly connected. In a storage array architecture provided in this embodiment of the present invention, logically, controllers are collectively referred to as a controller plane, switching devices are collectively referred to as a switching plane, hard disks are collectively referred to as a storage plane, input/output managers are collectively referred to as an input/output management plane, and cache devices are collectively referred to as a cache plane. In the architecture provided in this embodiment of the present invention, data reading and writing control is separated from data reading and writing. A controller implements data reading and writing control, but data reading and writing (or in other words, read and written data) does not flow through the controller, which saves CPU computing resources of the controller and memory resources of the controller, improves data writing efficiency, and improves data processing efficiency of the storage array. The storage array architecture in this embodiment of the present invention can implement expansion of devices such as controllers and hard disks, and controllers, switching devices, hard disks, and the like may be added flexibly according to performance requirements of the storage array.
Alternatively, the technical solution in this embodiment of the present invention is also applicable to a scenario in which a storage array includes one input/output manager, one controller, one switching device, one cache device, and several hard disks. For a manner of writing data into the storage array in this scenario, refer to the description in the foregoing embodiment. For a scenario in which data deduplication is performed in the storage array, refer to the description in the foregoing embodiment. For a data reading operation performed in a storage array, refer to the description in the foregoing embodiment. Alternatively, a storage array may also include two controllers and one switching device, where the two controllers are connected to the switching device. For operations of data writing, data deduplication and data reading in such a scenario, refer to the description in the foregoing embodiment, and details are not described herein again. In this embodiment of the present invention, a device A reads data from a cache address A (or in other words, directly reads data from the cache address A) or writes data to the cache address A (or in other words, directly writes data to the cache address A) via the PCIe switching device A or the PCIe switching device B according to an identifier of a device B and the cache address A. Such an embodiment may be implemented by using a direct memory access (DMA) technology, where the device A and the device B represent devices that specifically perform DMA access in this embodiment of the present invention.
The controller obtains a cache address of the device B, and sends the identifier of the device B and the cache address of the device B to a device C via the PCIe switching device A or the PCIe switching device B. Because the controller communicates with the device B to obtain the cache address via the PCIe switching device A or the PCIe switching device B and already learns the identifier of the device B, the cache address is obtained, and the identifier of the device B and the cache address of the device B may be sent to the device C. Alternatively, the controller may also obtain the identifier and cache address of the device B. The identifier of the device B may be the address of the device B or another identifier that uniquely identifies the device.
A person of ordinary skill in the art may be aware that, the exemplary units and algorithm steps described with reference to the embodiments disclosed in the specification may be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed by hardware or software depends on particular applications and design constraint conditions of the technical solutions. A person skilled in the art may use different methods to implement the described functions for each particular application, but it should not be considered that the implementation goes beyond the scope of the present invention.
It may be clearly understood by a person skilled in the art that, for the purpose of convenient and brief description, for a detailed working process of the foregoing system, apparatus, and unit, refer to a corresponding process in the foregoing method embodiments, and details are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system and method may be implemented in other manners. For example, the described apparatus embodiment is merely exemplary. For example, the unit division is merely logical function division and may be other division in actual implementation. For example, a plurality of units or components may be combined or integrated into another system, or some features may be ignored or not performed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections may be implemented through some interfaces. The indirect couplings or communication connections between the apparatuses or units may be implemented in electronic, mechanical, or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each of the units may exist alone physically, or two or more units are integrated into one unit.
When the functions are implemented in the form of a software functional unit and sold or used as an independent product, the functions may be stored in a computer-readable non-volatile storage medium. Based on such an understanding, the technical solutions of the present invention essentially, or the part contributing to the prior art, or some of the technical solutions may be implemented in a form of a software product. The software product is stored in a non-volatile storage medium, and includes several instructions for instructing a computer device (which may be a personal computer, a server, or a network device) to perform all or some of the steps of the methods described in the embodiments of the present invention. The foregoing non-volatile storage medium includes: any medium that can store program code, such as a USB flash drive, a removable hard disk, a read-only memory (ROM), a magnetic disk, or an optical disc.
This application is a continuation of International Application No. PCT/CN2014/086530, filed on Sep. 15, 2014, which is hereby incorporated by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| Parent | PCT/CN2014/086530 | Sep 2014 | US |
| Child | 15449083 | US |
