WO2014071588A1 - Data replication method, storage controller and system - Google Patents

Abstract

Provided is a data replication method. The data replication method comprises: when receiving write I/O data delivered by a host, a label controller determining whether there is an existing receiving sub-cycle (11) that has started currently; if there is no existing receiving sub-cycle that has started currently, the label controller starting a receiving sub-cycle corresponding to a label generated currently and identifying the write I/O data (12) according to the label corresponding to the receiving sub-cycle; if there is an existing receiving sub-cycle that has started currently, identifying the I/O data (13) according to the label corresponding to the receiving sub-cycle; when the receiving sub-cycle ends, replicating the write I/O data identified by the label corresponding to the receiving sub-cycle to a destination end (14). When receiving the write I/O data delivered by the host, the label controller starts a new receiving sub-cycle if there is a newly generated label but no existing receiving sub-cycle that has started. Therefore, the data replication method solves a technical problem that the receiving sub-cycle of a next cache cycle can be started only when a sending sub-cycle of a previous cache cycle ends. Also provided is a storage controller and a system. The data replication method, storage controller and system can reduce a probability of write I/O data blocking.

Description

 Data replication method and storage controller and system

TECHNICAL FIELD Embodiments of the present invention relate to computer technologies, and in particular, to a data replication method and a storage controller and system. Background technique

 A storage device-based replication solution is a data replication solution between storage devices. It is mainly used for disaster recovery, data migration and data backup, and remote transmission of data based on user purposes. The end that sends data can be called: Production Center, and the end of receiving data is called: Disaster Recovery Center or Disaster Recovery Center. A storage device-based replication scheme is suitable for sending data from a single storage device to other nodes, and is particularly useful for transmitting data from a clustered storage system to other nodes. When the data arrives at the same time, the data of the receiving end is the same as the data of the transmitting end, and the data of the receiving end and the transmitting end are consistent.

 The cache-based replication scheme is based on a replication scheme of the storage device. The flow is as follows: The host sends a write input/output (Input/Output, hereinafter referred to as I/O) to the storage device A, and the storage device A opens the cache space to receive the write. I/O, the time that the buffer space receives write I/O is called the receive sub-period. During the receive sub-cycle, all write I/Os with the same access address are merged into the last received write I/O. When the receiving sub-cycle ends, the transmission sub-period is entered. The transmit sub-period is the period in which the write I/O data is received for the receive sub-cycle. A receiving sub-period and a transmitting sub-period form a complete buffering period for copying data to the storage device B. In order to keep the data in the storage device B and the data in the storage data A consistent, the storage device A does not receive the write I/O sent by the host, but directly discards the write I/O sent by the host. , causing write I/O blocking, only copying the write I/O received in the receive sub-cycle to storage device B. Only after the end of the transmission sub-cycle, that is, after the write I/O received in the receive sub-cycle is completed, it enters the next receive sub-cycle and continues to receive the write I/O sent by the host.

In order to improve the granularity of data consistency, the receiving sub-period can be shortened as much as possible, so that the transmission sub-period is also shortened, but the probability of writing I/O blocking is still relatively high. Summary of the invention

 Embodiments of the present invention provide a data replication method and a storage controller and system, which can reduce the probability of blocking I/O blocking.

 In a first aspect, an embodiment of the present invention provides a data replication method, including:

 When the tag controller receives the write I/O data sent by the host, it determines whether there is a currently received receive sub-period;

 If the receiving sub-period does not exist, the label controller starts the receiving sub-period corresponding to the currently generated label, and identifies the write I/O data according to the label corresponding to the receiving sub-period;

 If there is currently a received receiving sub-period, the tag controller identifies the write I/O data according to the tag corresponding to the receiving sub-period;

 At the end of the receiving sub-period, the tag controller copies the write I/O data of the tag identifier corresponding to the receiving sub-period to the destination end.

 In conjunction with the first aspect, in a first possible implementation, the method further includes: the label controller generates a label when the label generation time point arrives, and the labels generated by different label generation time points are different.

 With reference to the first aspect, or the first possible implementation manner of the first aspect, in a second possible implementation manner, the method further includes: the label controller broadcasting, to the non-label controllers in the cluster storage system, the label The label starts the message.

 In a second aspect, an embodiment of the present invention further provides a data replication method, including:

 When the non-label controller receives the write I/O data sent by the host, it determines whether there is a currently received receiving sub-period;

 If the receiving sub-period does not exist, the non-label controller starts the receiving sub-period corresponding to the currently received label, and identifies the write I/O data according to the label corresponding to the receiving sub-period;

 If there is currently a received receiving sub-period, the non-label controller identifies the write I/O data according to the label corresponding to the receiving sub-period;

 At the end of the receiving sub-period, the non-tag controller copies the write I/O data of the tag identifier corresponding to the receiving sub-period to the destination end.

With reference to the second aspect, in a first possible implementation manner, the non-label controller receives A label activation message broadcast by the label controller in the cluster storage system, the label activation message includes the label, and the labels broadcast by the label controller are different at different label generation time points.

 With reference to the second aspect, or the first possible implementation manner of the second aspect, in the second possible implementation manner, at the end of the receiving sub-period, or receiving the end of the label controller broadcast When receiving the identifier stop message of the sub-period corresponding to the label, the non-label controller stops identifying the write I/O data sent by the host according to the label corresponding to the end of the received sub-period.

 In a third aspect, an embodiment of the present invention provides a storage controller, including:

 The determining module is configured to: when receiving the write I/O data sent by the host, determine whether there is a currently received receiving sub-period;

 An identifier module, configured to start a receiving sub-period corresponding to the currently generated label if the currently-used receiving sub-period does not exist, and identify the writing according to the label corresponding to the receiving sub-period

I/O data;

 The identifier module is further configured to: if the receiving sub-period is currently activated, identify the write I/O data according to the label corresponding to the receiving sub-period;

 And a copying module, configured to: when the receiving sub-period ends, copy the write I/O data of the label corresponding to the receiving sub-period to the destination end.

 Combine the third aspect. In a first possible implementation manner, the method further includes: a label control module, configured to generate a label when a label generation time point arrives, and different labels generate time points to generate different labels.

 Combine the third aspect. Or the first possible implementation manner of the third aspect, in a second possible implementation manner, the label control module is further configured to: broadcast, to each non-label controller in the cluster storage system, a label that includes the label to be started. Message.

 In a fourth aspect, the embodiment of the present invention further provides a storage controller, including:

 The determining module is configured to: when receiving the write I/O data sent by the host, determine whether there is a currently received receiving sub-period;

An identifier module, configured to: if a receiving sub-period is not currently activated, start a receiving sub-period corresponding to the currently received label, and identify the write I/O data according to the label corresponding to the receiving sub-period; The identifier module is further configured to: if the receiving sub-period is currently activated, identify the write I/O data according to the label corresponding to the receiving sub-period;

 And a copying module, configured to: when the receiving sub-period ends, copy the write I/O data of the label corresponding to the receiving sub-period to the destination end.

 With reference to the fourth aspect, in a first possible implementation, the method further includes: a label control module, configured to generate a label when a label generation time point arrives, and different labels generate time points to generate different labels.

 With reference to the fourth aspect, or the first possible implementation manner of the fourth aspect, in a second possible implementation manner, the label control module is further configured to broadcast to each non-label controller in the cluster storage system, including the The tag's tag initiates the message.

 The fifth aspect of the present invention provides a data replication system, including: the storage controller provided by the third aspect and the storage controller provided by the fourth aspect.

 In a sixth aspect, an embodiment of the present invention further provides a storage controller, including: a memory, a processor, a bus, and a communication interface; wherein the processor, the communication interface, and the memory are mutually completed by the bus Communication

 The memory is used to store a program; the processor is configured to execute the program; and the program is configured to:

 When receiving the write I/O data sent by the host, it is determined whether there is a currently received receiving sub-period;

 If the received sub-period does not exist, the receiving sub-period corresponding to the currently generated label is started, and the write I/O data is identified according to the label corresponding to the receiving sub-period.

 If there is currently a received receiving sub-period, the write I/O data is identified according to the label corresponding to the receiving sub-period;

 At the end of the receiving sub-period, the write I/O data of the tag identifier corresponding to the receiving sub-period is copied to the destination through the communication interface.

 The seventh aspect of the present invention further provides a storage controller, including:

 a memory, a processor, a bus, and a communication interface; wherein the processor, the communication interface, and the memory complete communication with each other through the bus;

The memory is used to store a program; the processor is configured to execute the program; and the program is configured to: When receiving the write I/O data sent by the host, it is determined whether there is a currently received receiving sub-period;

 If the receiving sub-period does not exist, the receiving sub-period corresponding to the currently received label is started, and the write I/O data is identified according to the label corresponding to the receiving sub-period;

 If there is currently a received receiving sub-period, the write I/O data is identified according to the label corresponding to the receiving sub-period;

 At the end of the receiving sub-period, the write I/O data of the tag identifier corresponding to the receiving sub-period is copied to the destination through the communication interface.

 In the foregoing technical solution, when the storage controller receives the write I/O data sent by the host to the cache, if there is no initiated reception sub-period, the storage controller may start a new receiving sub-period corresponding to the currently generated label. And identifying the write I/O data sent by the current host according to the label corresponding to the receiving sub-period, and the storage controller only needs to determine whether there is a received receiving sub-period, and does not need to determine the status of the transmitting sub-period. If there is a received receiving sub-period, the write I/O data sent by the current host is directly identified according to the label corresponding to the receiving sub-period. After the end of any of the receive sub-cycles, the corresponding transmit sub-cycle begins, that is, the write I/O data identified by the tag is copied to the destination. When receiving write I/O data sent by the host, the conditions for starting a new receive sub-cycle are: There is a newly generated tag and there is no received receive sub-cycle. Instead of starting the receive sub-cycle of the next buffer cycle after the end of the transmit sub-cycle of the previous buffer cycle. Therefore, the transmission sub-period of the previous buffer cycle can be performed simultaneously with the reception sub-cycle of the next buffer cycle, thereby reducing the probability of blocking the writing of I/O data. DRAWINGS

 1A is a flowchart of a data replication method according to an embodiment of the present invention;

 FIG. 1B is a schematic diagram of a cache cycle superposition in a data replication process according to an embodiment of the present invention; FIG.

 2A is a flowchart of another data replication method according to an embodiment of the present invention;

 2B is a scenario diagram of a replica data of a cluster storage system according to an embodiment of the present invention; FIG. 2C is a schematic diagram of a startup label message sent by a label controller according to an embodiment of the present invention;

FIG. 3 is a flowchart of still another data replication method according to an embodiment of the present invention; 4 is a flowchart of still another data replication method according to an embodiment of the present invention; FIG. 5A is a schematic structural diagram of a storage controller according to an embodiment of the present invention;

 FIG. 5B is a schematic structural diagram of another storage controller according to the embodiment; FIG.

 6A is a schematic structural diagram of still another storage controller according to the embodiment;

 FIG. 6B is a schematic structural diagram of still another storage controller according to the embodiment; FIG.

 FIG. 7 is a schematic structural diagram of still another storage controller according to the embodiment;

 FIG. 8 is a schematic structural diagram of still another storage controller according to the embodiment. detailed description

 FIG. 1 is a flowchart of a data replication method according to an embodiment of the present invention. This embodiment is applicable to a scenario in which a source end having a single storage controller copies data to a destination end. As shown in FIG. 1A, the method provided in this embodiment includes:

 Step 11: When the tag controller receives the write I/O data sent by the host, it determines whether there is a received receive sub-cycle.

 Step 12: If there is no received receiving sub-period, the tag controller starts the receiving sub-period corresponding to the currently generated tag, and identifies the write I/O data according to the tag corresponding to the receiving sub-period. The receiving sub-period is a receiving sub-period corresponding to the currently generated label.

 Step 13: If there is currently a received receiving sub-cycle, the tag controller identifies the write I/O data according to the tag corresponding to the receiving sub-period.

 In this embodiment, the tag controller is a memory controller having a tag generation function. The tag controller can generate tags when each tag generation time point arrives, and the tags generated at different tag generation time points are different. The label controller periodically generates different labels. The interval between the two adjacent label generation time points may be the same or different. The interval between the two tag generation time points is called the tag generation cycle. A tag can be a time tag represented by time.

When the tag controller receives the write I/O data sent by the host, it first determines whether there is a currently received receive sub-cycle, for example, records each receive sub-cycle that is started, and clears at the end of a receive sub-period. The record of the receiving sub-period. If there is currently a received receiving sub-period, the label controller identifies the currently received write I/O data according to the label corresponding to the receiving sub-period; otherwise, the label controller starts the receiving sub-period corresponding to the currently generated label. A new receive sub-cycle is initiated and the currently received write I/O data is identified based on the currently generated tag. After the end of any one of the receiving sub-cycles, the transmission sub-period corresponding to the receiving sub-period is started.

 Normally, the host first sends the write I/O data to the storage controller's cache, and then the storage controller copies the write I/O data from the cache to the storage array. The start of the receive sub-cycle is to open a new buffer space in the memory for receiving write I / O data, the size of the cache space is set according to the duration of the receive sub-period, if the cache space is insufficient, the write I / O data will be in the receive sub-period Internal overflow, causing write I/O blocking. Different receiving sub-cycles, the cache space used does not affect each other independently.

 If the tag controller generates the tag, if the write I/O data sent by the host is not received, the currently generated tag is discarded. If the write I/O data sent by the host is received, it is determined whether the currently activated device exists. The sub-period is received to determine whether to start the receiving sub-period corresponding to the currently generated tag. If there is currently a received receive sub-cycle, the tag controller also discards the currently generated tag.

 It can be seen from the above process of starting a new receiving sub-period that whether or not there is a newly generated tag is one of the conditions for starting the receiving sub-period. The following three conditions must be met to initiate a new receive sub-cycle: Receive I/O data sent by the host, Newly generated tags, and no received receive sub-cycles.

 When the tag controller receives the write I/O data sent by the host, if there is no start receiving sub-period currently, and there is no newly generated tag, that is, the time point at which the tag is not currently generated, the new receiving sub-period cannot be started. Instead, directly discard the write I/O data sent by the host. After the storage controller discards the write I/O data sent by the host, the host resends the discarded write I/O data. When the tag generation frequency is increased, that is, the tag generation period is shortened and the host memory is large, when the tag controller receives the write I/O data sent by the host, the following phenomenon is rare: There is no received receive sub-cycle. , there are no newly generated tags. Therefore, the technical solution provided by the present invention has the probability of discarding write I/O, that is, the probability of blocking write I/O, is extremely low in the data copying process.

 Step 14: At the end of the receiving sub-period, the tag controller copies the write I/O data of the tag identifier corresponding to the receiving sub-period to the destination end.

At the end of the receiving sub-period, the tag controller copies the write I/O of the tag identifier corresponding to the receiving sub-period to the destination end. The tag controller will initiate multiple receive sub-cycles in sequence. At the end of any receiving sub-period, the tag controller will copy the write I/O of the tag identifier corresponding to the currently-received receiving sub-period to the destination. In the embodiment of the present invention, the synchronous copy mode may be used, or the asynchronous copy mode may be used to copy the write I/O data to the destination end. After the write I/O data received by the destination end, according to the label in each write I/O data, the receiving sub-period of each write I/O data is determined, according to the order in which the sending end sends the write I/O data, The received write I/O data is stored in the local storage device, thereby improving the effectiveness of receiving the write I/O data.

 At the end of the receiving sub-period corresponding to any label, the sending sub-period corresponding to the label is started, that is, the write I/O data identified by the label in the cache is copied to the destination end. The receiving sub-period corresponding to a tag and the transmitting sub-period corresponding to the tag constitute a complete buffering cycle. In the process of copying the write I/O data to the destination end, a new receiving sub-period can be started as long as the condition for starting the new receiving sub-period is satisfied, and therefore, the transmission sub-period of one buffer period can be synchronized with the next buffer period. The receiving sub-cycles are performed simultaneously, and multiple buffer periods are superimposed on each other, thereby reducing the probability of blocking writing of I/O data.

 As shown in FIG. 1B, in the first label period, the first buffer period corresponding to the first label is started; in the second label period, the second buffer period corresponding to the second label is activated; Within the third buffer period corresponding to the third label is started. When the first cache cycle has not ended, the second cache cycle has started. Similarly, when the second cache cycle has not ended, the third cache cycle has started, and multiple cache cycles are superimposed on each other. It should be noted that Figure 1B only shows a possible situation and does not mean that a new cache cycle will be initiated every tag cycle. In practical applications, it is possible to start a new buffer cycle after one or several tag cycles. There are two critical points for each receive sub-cycle: one is the time at which the receive sub-cycle starts, and the other is the time at which the receive sub-cycle ends. In the case where the write I/O is very dense, if a write I/O arrives at the end of the receive sub-cycle and there is a newly generated tag, a new receive sub-cycle is started.

 The amount of I/O data sent by the host is very large, so the number of cache cycles superimposed is higher. When the amount of I/O data written by the host is normal and the network replication bandwidth is normal, the data amount of the label controller's buffer period is fixed, because all the write I/O data received in one receiving sub-period is sent to the destination. After the end, the transmission sub-period ends, that is, the corresponding buffer period ends, so the number of superimposed buffer periods does not always increase.

It should be noted that the duration of the receiving sub-period cannot be too short or too long. Receiving sub-period The shorter the duration, the faster the switching frequency of the receiving sub-period and the transmitting sub-period, and the less the received write I/O data and the transmitted write I/O data. When the replication is abnormally interrupted, the source and The data consistency at the destination is higher. However, the duration of the receiving sub-period cannot be too short, because the receiving sub-period is too short, the less the amount of write I/O data sent, and the frequent switching of the receiving sub-period and the transmitting sub-period consumes the computing resources of the controller. Therefore, the duration of the receiving sub-period cannot be too short or too long, and it is necessary to find an appropriate value in the middle. This depends on design and actual experimentation to explore.

 In the method provided by the foregoing embodiment, when the tag controller receives the write I/O data sent by the host to the cache, if there is no initiated receive sub-period, the tag controller can start the new corresponding corresponding tag. Receiving a sub-period, and identifying the write I/O data sent by the current host according to the label corresponding to the receiving sub-period. In the embodiment of the present invention, the label controller only needs to determine whether there is a received receiving sub-period, and does not need to Determine the status of the transmission sub-cycle. If there is a received receiving sub-period, the write I/O data sent by the current host is directly identified according to the label corresponding to the receiving sub-period. After the end of any of the receiving sub-cycles, the corresponding transmission sub-cycle begins, that is, the write I/O data identified according to the tag is copied to the destination. In the method provided by the foregoing embodiment, when the write I/O data sent by the host is received, the condition for starting the new receiving sub-period is: there is a newly generated label and there is no activated receiving sub-period. Instead of starting the transmit sub-period of the next cache cycle, the receive sub-cycle of the next cache cycle can be initiated. Therefore, in the method provided by the foregoing embodiment, the sending sub-period of the previous buffering period can be performed simultaneously with the receiving sub-period of the next buffering period, thereby reducing the probability of blocking the writing of the I/O data.

 2A is a flowchart of another data replication method according to an embodiment of the present invention. This embodiment is applicable to a scenario in which a source end of a cluster storage system copies data to a destination end. The storage controller in the source cluster storage system includes a label controller and a non-label controller. In the source cluster storage system, one storage controller is selected as the label controller, and the others are non-label controllers. Typically, the tag controller is the primary storage controller for the clustered storage system.

 As shown in FIG. 2A, the method provided in this embodiment includes:

 Step 20: The tag controller generates a tag when the tag generation time point arrives, and broadcasts a tag start message including the tag to each non-tag controller in the cluster storage system. Among them, the labels generated by different label generation time points are different.

Step 21: When the tag controller receives the write I/O data sent by the host, it determines whether the current storage is saved. In the receive sub-cycle that has been started.

 Step 22: If there is no received receiving sub-cycle, the tag controller starts the receiving sub-period corresponding to the currently generated tag, and identifies the write I/O data according to the tag corresponding to the receiving sub-period.

 The label controller may generate a label when each label generation time point arrives, and broadcast a label start message including the label to each non-label controller in the cluster storage system when generating the label, and different labels generate labels generated at time points. Not the same.

 As shown in 2B, the source-side clustered storage system can include multiple storage controllers that manage different storage arrays. The source can include one host or multiple hosts, and the host can send different write I/O data to different storage controllers. Each storage controller at the source side copies the write I/O data received by itself to the storage system of the destination end. The storage system at the destination end can also be a cluster storage system. The source end may be a production center, and the destination end may be a disaster recovery center; the source end may also be a transmitting end in a data migration process, and the destination end may be a receiving end in a data migration process; the source end may also be a data end. Back up the source center of the system, and correspondingly, the destination is the backup center.

 After the tag controller generates a new tag, it broadcasts a tag initiation message including the tag to other storage controllers in the cluster storage system, that is, the non-label controller, that is, the tag startup message including the tag is broadcast in the local cluster storage system. The tag start message broadcast by the tag controller to each non-tag controller point at a tag generation time point is the same, that is, the tag controller sends the tag to each non-tag controller at a tag generation time point. identical. As shown in Figure 2C, the tag controller sends a tag initiation message including the tag to each non-tag controller at a fixed frequency F.

 Step 23: If there is currently a received receiving sub-period, the tag controller identifies the write I/O data according to the tag corresponding to the receiving sub-period.

 Step 24: At the end of the receiving sub-period, the label controller copies the write I/O of the label identifier corresponding to the receiving sub-period to the destination end.

 The method of processing the write I/O data sent by the host in the cluster storage system is the same. The difference is that: when the label controller generates the label in the cluster storage system, the label controller is also broadcast to each non-label controller. Tag startup message. Steps 21 to 24 are the same as steps 11 to 14 and will not be described here.

After the non-label controller receives the label start message, it processes the write I/O sent to the host. The method is similar to the tag controller, except that: The non-tag controller receives the write from the host.

In the case of I/O data, if it is determined that there is no receiving sub-period currently, the receiving sub-period corresponding to the label in the currently received label start message is started.

 The label controller broadcasts the label start message including the label to each non-label controller, so as to cause each storage controller in the cluster storage system to locally initiate a new receiving sub-period to enable each storage controller to identify any one of the labels. After the write I/O data is copied to the destination, the data in the cluster storage system at the source is consistent with the data on the destination. In order to simplify management, in different memory controllers, the duration of each receiving sub-period is set to a preset duration, and the duration of each receiving sub-period is equal. Generally, the write I/O sent by the host is dense. The time when the label start message sent by the label controller reaches different non-label controllers may have a time difference. Different storage controllers start the receive sub-period corresponding to the same label. The time points are similar. Since the receiving sub-periods of the same label have the same duration, correspondingly, the time points at which the different storage controllers end the receiving sub-periods corresponding to the same label are similar, that is, the time points at which the transmitting sub-cycles corresponding to the same label are started are similar. Therefore, each storage controller copies the write I/O data of the same tag identifier to the destination end at a similar time, so that after the write I/O data corresponding to one tag is copied to the destination end, the source end of the cluster storage system is The data is consistent with the data at the destination.

 Optionally, different storage controllers may receive a large difference in the amount of write I/O data sent by the host to the cache, so that different storage controllers start the transmission sub-periods corresponding to the same label at different time points. After the transmission sub-period corresponding to one label is completed, the data in the cluster storage system at the transmitting end and the data in the receiving end can be consistent. The present invention also provides the following technical solutions: The label controller at the end of a receiving sub-period, to each station The non-label controller broadcasts the stop identification message, and notifies each of the non-label controllers to stop writing I/O data sent by the host according to the label corresponding to the receiving sub-period, and starts to copy the cache identifier in the cache to the receiving end. Write I/O data.

In the method provided by the foregoing embodiment, the label controller periodically sends a label start message including a label to each non-label controller of the local cluster storage system. When the tag controller receives the write I/O data sent by the host to the cache, if there is no initiated receive sub-period, the new receive sub-period corresponding to the currently generated tag may be started, and corresponding to the receive sub-period according to the receive sub-period The tag identifies the write I/O data sent by the current host. If there is a receiving sub-cycle that has been started, then The write I/O data sent by the current host is identified according to the label corresponding to the receiving sub-period. Each non-label controller in the cluster storage system receives the write I/O data sent by the host to the cache. The difference from the processing method of the label controller is that when the receiving sub-cycle does not exist, the current receiving is started. The received sub-period corresponding to the tag in the tag start message. After the end of any of the receiving sub-cycles, the corresponding transmission sub-cycle begins, that is, the write I/O data identified by the tag is copied to the destination. In the method provided by the foregoing embodiment, when each storage controller in the cluster storage system receives the write I/O data sent by the host, the condition for starting the new receiving sub-period is: the newly generated label exists and the non-existent activated Receive sub-period. Instead of starting the receive sub-cycle of the next buffer cycle after the end of the transmit sub-cycle of the previous buffer cycle. Therefore, in the method provided by the foregoing embodiment, the sending sub-period of the previous buffering period can be performed simultaneously with the receiving sub-period of the next buffering period, thereby reducing the blocking write of the cluster storage system at the source end.

The probability of I/O data.

 FIG. 3 is a flowchart of still another data replication method that is commonly provided in the embodiment of the present invention. The execution body of this embodiment is a non-label controller in the cluster storage system at the source end. As shown in FIG. 3, the method provided in this embodiment includes:

 Step 31: When the non-tag controller receives the write I/O data sent by the host, it determines whether there is a currently received receive sub-period.

 Step 32: If there is no received receiving sub-period, the non-label controller starts the receiving sub-period corresponding to the currently received label, and identifies the write I/O data according to the label corresponding to the receiving the receiving sub-period. .

 The label controller may generate a label when each label generation time point arrives, and broadcast a label start message including the label to each non-label controller in the cluster storage system when generating the label, and different labels generate labels generated at time points. Not the same. After receiving the label start message, the non-label controller processes the write I/O sent by the host similarly to the label controller. The difference is that when the non-label controller receives the write I/O data sent by the host. If it is determined that there is currently no receiving sub-period, the receiving sub-period corresponding to the label in the currently received label start message is started.

 Step 33: If there is currently a received receiving sub-period, the non-tag controller identifies the write I/O data according to the label corresponding to the receiving sub-period.

When the non-label controller receives the label start message, if it does not receive the write issued by the host If the I/O data is sent, the label in the current label start message is discarded. If the write I/O data sent by the host is received, it is determined whether there is a currently received receiving sub-period, thereby determining whether to start the current label start message. The label corresponds to the receiving sub-period. If there is currently a received receive sub-cycle, the tag controller will also discard the currently generated tag.

 Step 34: At the end of the receiving sub-period, the non-tag controller copies the write I/O of the tag identifier corresponding to the receiving sub-period to the destination end.

 Optionally, different storage controllers may receive a large difference in the amount of write I/O data sent by the host to the cache, so that different storage controllers start the transmission sub-periods corresponding to the same label at different time points. After the transmission sub-period corresponding to any one of the labels is completed, the data in the cluster storage system at the source end is consistent with the data in the destination end, and the non-label controller receives the corresponding sub-cycle corresponding label broadcasted by the label controller. When the stop message is identified, the write I/O data sent by the host is determined according to the label corresponding to the end of the received sub-period. Alternatively, the non-tag controller stops identifying the write I/O data sent by the host according to the label corresponding to the end of the receiving sub-period at the end of any one of the receiving sub-periods.

 Optionally, each non-label controller receives the write I/O data in the currently received receiving sub-period at the end of any one of the receiving sub-cycles or when receiving the identifier stop message sent by the label controller. number. By counting the number of write I/O data, it can be determined whether all the write I/O data received in the currently received receive sub-cycle is copied to the destination. In addition, each non-label controller may also return a response message to the label controller after receiving the identifier stop message sent by the label controller, and the response message carries each receiving period in the receiving sub-period for the identifier stop message. To count the number of I/O data to be written, the tag controller counts the total number of write I/O data identified by all the storage controllers in the receiving sub-period for which the above-mentioned identification stop message is directed, thereby determining the cluster storage system. Whether to send all write I/O data in the receiving sub-period for which the above-mentioned identification stop message is directed to the destination.

 Optionally, in order to improve data replication efficiency, the storage controller accesses the same set of write I/O data with the same address and writes the last received write I in the write I/O data set in any one of the receiving sub-periods. /O data.

 FIG. 4 is a flowchart of still another data replication method according to an embodiment of the present invention. This embodiment is an execution process of each storage controller in a cache period. As shown in Figure 4, it includes:

Step 41: The tag controller generates a tag when each tag generation time point arrives, and The non-tag controller broadcasts a tag initiation message including the tag.

 Step 42A: When the tag controller receives the write I/O data sent by the host, it determines whether there is a currently received receiving sub-period. If there is no activated receiving sub-period, the receiving sub-period corresponding to the currently generated tag is started. And writing the I/O data according to the currently generated label, otherwise, identifying the write I/O data according to the label corresponding to the received receiving sub-cycle.

 Step 42B: When the non-tag controller receives the write I/O data sent by the host, it determines whether there is a received receiving sub-period currently. If there is no activated receiving sub-period, the currently received tag start message is started. The label corresponds to the receiving sub-period, and identifies the write I/O data according to the label in the currently-used label start message, otherwise, the write I/O data is identified according to the label corresponding to the activated receiving sub-period. .

 In this case, the write I/O data sent by the host is relatively dense, and the receive sub-periods initiated in step 42A and step 42B can be regarded as the receive sub-periods corresponding to the same label.

 Step 43: The tag controller broadcasts an identity stop message for the receiving sub-period to each non-tag controller at the end of a receiving sub-cycle initiated by step 42A.

 Step 44: When the non-label controller receives the identifier stop message, it stops the write I/O data sent by the label corresponding to the label corresponding to the receiving sub-period for the identifier stop message.

 Each storage controller receives the write I/O data sent by the host in the receiving sub-period. The amount of write I/O data may be different. Some write I/O data volume will be 4, and correspondingly, the I/O data is written with the identifier. The time will be longer; some write I/O data will be small, and accordingly, the time to write I/O data with the identifier will be shorter. Therefore, in a receiving sub-period corresponding to a label, some storage controllers may have identified all the received write I/O data according to the above label, and some storage controllers may not have received the received label according to the label identifier. Partially write I/O data. The data in the cluster storage system at the source end is consistent with the data at the destination end after the receiving sub-period corresponding to a label. After the labeling controller ends the receiving sub-period corresponding to the label, the label controller stops the message in the local cluster storage system. After receiving the identifier stop message, each storage controller stops the label identifier corresponding to the receiving sub-period for the identifier stop message. Write I/O data sent by the host. At the same time, the transmission sub-period corresponding to the receiving sub-period for the identification stop message is entered, that is, the write I/O data marked by the label corresponding to the receiving sub-period is simultaneously copied to the destination end.

Step 45A: The tag controller enters a sending sub-period corresponding to the receiving sub-period initiated by step 42A, and copies the write I/O data identified by the tag to the destination end. Step 45B: The non-tag controller enters the sending sub-period corresponding to the receiving sub-period initiated by step 42B, and copies the write I/O data identified by the tag to the destination end.

 FIG. 5A is a schematic structural diagram of a storage controller according to an embodiment of the present invention. As shown in FIG. 5A, the memory controller includes: a determining module 51, an identifying module 52, and a copying module 53.

 The determining module 51 is configured to: when receiving the write I/O data sent by the host, determine whether there is currently a received receiving sub-period.

 The identifier module 52 is configured to start a receiving sub-period corresponding to the currently generated label if the currently-used receiving sub-period does not exist, and identify the write I/O data according to the label corresponding to the receiving sub-period.

 The identifier module 52 is further configured to: if the currently received receiving sub-period exists, identify the write I/O data according to the label corresponding to the receiving sub-period.

 The copying module 53 is configured to: when the receiving sub-period ends, copy the write I/O data of the label identifier corresponding to the receiving sub-period to the destination end.

 Optionally, as shown in FIG. 5B, the above storage controller may further include a label control module 54. The label control module is configured to generate labels when the label generation time point arrives, and the labels generated by different label generation time points are different. When the identification module starts the receiving sub-cycle, the receiving sub-period corresponding to the tag generated by the tag control module is started.

 Optionally, the determining module is further configured to discard the generated label if the write I/O data sent by the host is not generated when the label is generated.

 Optionally, in order to make the foregoing storage controller applicable to a tag controller in the cluster storage system, the foregoing storage controller may further include: a tag control module.

 The tag control module is further configured to broadcast, to each non-tag controller in the cluster storage system, a tag initiation message including the tag. The identification module 52 initiates a reception sub-period based on the tag generated by the tag control module.

Optionally, in order to enable different storage controllers in the cluster storage system to start the transmission sub-periods corresponding to the same label at a similar time point, and then, after the corresponding transmission sub-period of one label ends, the data of the cluster storage system at the source end is The data at the destination can be consistent. The label control module is further configured to: when the receiving sub-period ends, broadcast, to the non-label controllers, an identifier stop message for the end receiving sub-period corresponding label, where the identifier stop message is used to notify the Each non-label controller stops the label label corresponding to the receiving sub-period according to the end Know the write I/O data sent by the host. The tag control module broadcasts an identification stop message for the end receiving sub-period corresponding tag to the non-tag controllers at the end of any one of the receiving sub-cycles.

 For the foregoing embodiment, reference may be made to the description of the corresponding embodiment of FIG. 1A, FIG. 2A and FIG. 4, and FIG. 6A is not a schematic structural diagram of another storage controller provided by this embodiment. As shown in FIG. 6A, the memory controller includes: a judging module 61, an identifying module 62, and a copying module 63.

 The determining module 61 is configured to: when receiving the write I/O data sent by the host, determine whether there is currently a received receiving sub-period;

 The identifier module 62 is configured to: if the receiving sub-period does not exist, start a receiving sub-period corresponding to the currently received label, and identify the write I/O data according to the label corresponding to the receiving sub-period;

 The identifier module 62 is further configured to: if the currently received receiving sub-period exists, identify the write I/O data according to the label corresponding to the receiving sub-period;

 The copying module 63 is configured to: when the receiving sub-period ends, copy the write I/O data of the label identifier corresponding to the receiving sub-period to the destination end.

 Optionally, as shown in FIG. 6B, the method further includes: a receiving module 64, configured to receive a label start message broadcast by the label controller in the cluster storage system, where the label start message includes the label, and different label generation time points The tags broadcast by the tag controller are different. When the identification module 62 starts the receiving sub-period, the receiving sub-period corresponding to the tag received by the receiving module is started.

 Optionally, the identifier module 62 is further configured to: when the receiving sub-period ends, or when receiving the identifier stop message broadcasted by the label controller for the end receiving sub-period corresponding label, stop the The label corresponding to the receiving sub-cycle of the end identifies the write I/O data sent by the host.

 Optionally, the determining module 61 is further configured to: when the receiving module receives the label initiation message, if there is no write I/O data sent by the host, discarding the label startup message label.

For the foregoing embodiment, reference may be made to the description of the corresponding embodiments in FIG. 3 and FIG. 4, and details are not described herein again. The embodiment of the present invention further provides a data replication system, including the storage controller shown in FIG. 5A and FIG. 6A, or a storage controller as shown in FIG. 5B and FIG. 6B. FIG. 7 is a schematic structural diagram of still another storage controller according to the embodiment. The storage controller provided in this embodiment is applicable to a single memory scenario, and is also applicable to a tag memory in a clustered presence system. The storage controller provided in this embodiment is applicable to a non-tag memory in a cluster presence system. The memory controller 700 shown in FIG. 包括 includes: a memory 730, a processor 710, a bus 740, and a communication interface 720; wherein the processor, the communication interface, and the memory Communication with each other is accomplished through the bus.

The processor 710 may be a central processing unit CPU, or a specific integrated circuit ASIC ted Circuit ), or configured to implement the present invention.

 The memory 730 is configured to store the program 732. Memory 730 may include high speed RAM memory and may also include non-volatile memory, such as at least one disk memory. In particular, program 732 can include program code, the program code including computer operating instructions.

 The processor is configured to execute the program 732.

 The program is for:

 When receiving the write I/O data sent by the host, it is determined whether there is a currently received receiving sub-period;

 If the received sub-period does not exist, the receiving sub-period corresponding to the currently generated label is started, and the write I/O data is identified according to the label corresponding to the receiving sub-period.

 If there is currently a received receiving sub-period, the write I/O data is identified according to the label corresponding to the receiving sub-period;

 At the end of the receiving sub-period, the write I/O data of the tag identifier corresponding to the receiving sub-period is copied to the destination through the communication interface.

 The program is also used to:

 Tags are generated when the tag generation time point arrives, and tags generated by different tag generation time points are different.

 Optionally, the program is further configured to: broadcast, by using the communication interface, a label start message including the label to each non-tag controller in the cluster storage system.

Optionally, the program is further configured to: pass the pass at the end of the receiving sub-period The information interface broadcasts an identifier stop message for the end of the receiving sub-period corresponding label to the non-label controllers, where the identifier stop message is used to notify the non-label controllers to stop according to the end of the receiving sub-period. The tag identifies the write I/O data sent by the host.

 Optionally, the program is further configured to: when the label is generated, if the write I/O data sent by the host is not generated, the generated label is discarded.

 For the foregoing embodiment, reference may be made to the description of the corresponding embodiment of FIG. 1A, FIG. 2A and FIG. 4, and details are not described herein again.

 FIG. 8 is a schematic structural diagram of still another storage controller according to the embodiment. The storage controller provided in this embodiment is applicable to a non-tag memory in a clustered presence system. The memory controller 800 shown in FIG. 8 includes: a memory 830, a processor 810, a bus 840, and a communication interface 820; wherein the processor, the communication interface, the memory Communication with each other is accomplished through the bus.

 Processor 810 may be a central processing unit CPU, or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

 The memory 830 is configured to store the program 832. Memory 830 may include high speed RAM memory and may also include non-volatile memory, such as at least one disk memory. In particular, program 832 can include program code, the program code including computer operating instructions.

 The processor is configured to execute the program 832.

 The program is for:

 When receiving the write I/O data sent by the host, it is determined whether there is a currently received receiving sub-period;

 If the receiving sub-period does not exist, the receiving sub-period corresponding to the currently received label is started, and the write I/O data is identified according to the label corresponding to the receiving sub-period;

 If there is currently a received receiving sub-period, the write I/O data is identified according to the label corresponding to the receiving sub-period;

 At the end of the receiving sub-period, the write I/O data of the tag identifier corresponding to the receiving sub-period is copied to the destination through the communication interface.

Optionally, the program is further configured to: Receiving, by the communication interface, a label activation message broadcast by the label controller in the cluster storage system, the label activation message includes the label, and the label broadcast by the label controller is different at different label generation time points.

 Optionally, the program is further configured to: when the receiving sub-period ends, or when receiving, by the communication interface, the identifier stop message broadcasted by the label controller for the end receiving sub-period corresponding label And stopping to write the write I/O data sent by the host according to the label corresponding to the end of the receiving sub-period.

 Optionally, the program is further configured to: when the label initiation message is received by the communication interface, if there is no write I/O data sent by the host, discarding the label startup message label.

 For the foregoing embodiment, reference may be made to the description of the corresponding embodiments in FIG. 3 and FIG. 4, and details are not described herein again. A person skilled in the art can understand that all or part of the steps of implementing the above method embodiments may be completed by using hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, and the program is executed when executed. The method includes the steps of the foregoing method embodiments; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.

 Finally, it should be noted that the above embodiments are only for explaining the technical solutions of the present invention, and are not intended to be limiting thereof; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that The technical solutions described in the foregoing embodiments may be modified, or some or all of the technical features may be equivalently replaced; and the modifications or substitutions do not deviate from the technical solutions of the embodiments of the present invention. range.

Claims

claims 1. A data replication method, characterized by including: When the tag controller receives the write I/O data sent by the host, it determines whether there is currently a started receiving sub-cycle; If there is currently no activated receiving sub-period, the tag controller starts the receiving sub-period corresponding to the currently generated tag, and identifies the write I/O data according to the tag corresponding to the receiving sub-period; If there is currently a started receiving sub-cycle, the tag controller identifies the write I/O data according to the tag corresponding to the receiving sub-cycle; At the end of the receiving sub-period, the tag controller copies the write I/O data identified by the tag corresponding to the receiving sub-period to the destination. 2. The method according to claim 1, characterized in that the method further includes: the tag controller generates a tag when the tag generation time point arrives, and the tags generated at different tag generation time points are different. 3. The method according to claim 2, characterized in that the method further includes: the tag controller broadcasting a tag startup message including the tag to each non-tag controller in the cluster storage system. 4. The method according to claim 3, characterized in that, the method further includes: at the end of the receiving sub-period, the tag controller broadcasts the ending receiving sub-period to each non-tag controller. The identification stop message corresponding to the tag, the identification stop message, is used to notify each non-tag controller to stop writing I/O data issued by the host according to the tag identification corresponding to the completed receiving sub-period. 5. The method according to claim 2, 3 or 4, characterized in that the method further includes: When generating a tag, if there is no write I/O data issued by the host, the tag controller discards the generated tag. 6. A data copying method, characterized by including: When the non-tag controller receives the write I/O data sent by the host, it determines whether there is currently a started receiving sub-cycle; If there is currently no started receiving sub-period, the non-tag controller starts the current receiving The tag corresponding to the receiving sub-period, and the write is identified according to the tag corresponding to the receiving sub-period I/O data; If there is currently a started receiving sub-cycle, the non-tag controller identifies the write I/O data according to the tag corresponding to the receiving sub-cycle; At the end of the receiving sub-period, the non-tag controller copies the write I/O data identified by the tag corresponding to the receiving sub-period to the destination. 7. The method according to claim 6, wherein the method further includes: the non-tag controller receiving a tag startup message broadcast by a tag controller in the cluster storage system, the tag startup message including the tag , and the tags broadcast by the tag controller are different at different tag generation time points. 8. The method according to claim 7, characterized in that, the method further includes: at the end of the receiving sub-period, or, receiving the tag corresponding to the ended receiving sub-period broadcast by the tag controller. When identifying a stop message, the non-tag controller stops identifying the write I/O data issued by the host according to the tag corresponding to the completed receiving sub-period. 9. The method according to claim 7 or 8, characterized in that, the method further includes: when receiving the tag start message, if there is no write I/O data issued by the host, the non- The tag controller discards the tag in the tag start message. 10. A storage controller, characterized in that it includes: The judgment module is used to judge whether there is currently a started receiving sub-cycle when receiving the write I/O data sent by the host; An identification module, configured to start the receiving sub-period corresponding to the currently generated tag if there is currently no activated receiving sub-period, and identify the write I/O data according to the tag corresponding to the receiving sub-period; The identification module is also used to identify the write I/O data according to the tag corresponding to the reception sub-period if there is currently a started receiving sub-period; A copy module, configured to copy the write I/O data identified by the tag corresponding to the receive sub-cycle to the destination end at the end of the receive sub-cycle. 11. The controller according to claim 10, further comprising: a label control module, configured to generate labels when the label generation time point arrives, and different labels are generated at different label generation time points. 12. The controller according to claim 11, wherein the tag control module is further configured to broadcast a tag startup message including the tag to each non-tag controller in the cluster storage system. 13. The controller according to claim 12, wherein the tag control module is further configured to broadcast to each non-tag controller a response to the end of the receiving sub-period when the receiving sub-period ends. The identification stop message of the tag is used to notify each non-tag controller to stop writing I/O data issued by the host according to the tag identification corresponding to the completed receiving sub-period. 14. The controller according to claim 11, 12 or 13, wherein the judgment module is also configured to discard all write I/O data if there is no write I/O data issued by the host when generating a label. Describe the generated tags. 15. A storage controller, characterized in that it includes: The judgment module is used to judge whether there is currently a started receiving sub-cycle when receiving the write I/O data sent by the host; An identification module, configured to start a receiving sub-period corresponding to the currently received tag if there is currently no activated receiving sub-period, and identify the write I/O data according to the tag corresponding to the receiving sub-period; The identification module is also used to identify the write I/O data according to the tag corresponding to the reception sub-period if there is currently a started receiving sub-period; A copy module, configured to copy the write I/O data identified by the tag corresponding to the receive sub-cycle to the destination end at the end of the receive sub-cycle. 16. The controller according to claim 15, further comprising: a receiving module, configured to receive the tag startup broadcast by the tag controller in the cluster storage system; the tag startup message includes the tag, And the tags broadcast by the tag controller are different at different tag generation time points. 17. The controller according to claim 16, wherein the identification module is further configured to receive the end of the receiving sub-period broadcast by the tag controller when the receiving sub-period ends, or When the identification of the corresponding tag stops the message, the write I/O data issued by the host according to the identification of the tag corresponding to the completed receiving sub-period is stopped. 18. The controller according to claim 16 or 17, characterized in that the judgment mode The block is also configured to discard the tag in the tag startup message if there is no write I/O data sent by the host when the receiving module receives the tag startup message. 19. A storage controller, characterized in that it includes: a memory, a processor, a bus, and a communication interface; wherein the processor, the communication interface, and the memory complete communication with each other through the bus; The memory is used to store programs; the processor is used to execute the program; the program is used to: When receiving the write I/O data sent by the host, determine whether there is currently a started receiving sub-cycle; If there is currently no started receiving sub-cycle, start the receiving sub-cycle corresponding to the currently generated tag, and write I/O data according to the tag identification corresponding to the receiving sub-period. If there is currently a started receiving sub-cycle, identify the write I/O data according to the tag corresponding to the receiving sub-cycle; At the end of the receiving sub-period, the write I/O data identified by the tag corresponding to the receiving sub-period is copied to the destination end through the communication interface. 20. The controller according to claim 19, characterized in that the program is also used to: The label is generated when the label generation time point arrives. The labels generated at different label generation time points are different. 21. The controller according to claim 20, wherein the program is further configured to: broadcast a tag startup message including the tag to each non-tag controller in the cluster storage system through the communication interface. 22. The controller according to claim 21, characterized in that the program is also used to: At the end of the receiving sub-period, broadcast an identification stop message for the tag corresponding to the ended receiving sub-period to each non-tag controller through the communication interface, and the identification stop message is used to notify each non-tag The controller stops identifying the write I/O data sent by the host according to the tag corresponding to the completed receiving sub-cycle. 23. The controller according to claim 20, 21 or 22, characterized in that the program is also used to: when generating a tag, if there is no write I/O data issued by the host, discard it The generated tags. 24. A storage controller, characterized in that it includes: a memory, a processor, a bus and a communication interface; wherein the processor, the communication interface and the memory complete communication with each other through the bus; The memory is used to store programs; the processor is used to execute the program; the program is used to: When receiving the write I/O data sent by the host, determine whether there is currently a started receiving sub-cycle; If there is currently no activated receiving sub-period, start the receiving sub-period corresponding to the currently received tag, and write the I/O data according to the tag identification corresponding to the receiving sub-period; If there is currently a started receiving sub-cycle, identify the write I/O data according to the tag corresponding to the receiving sub-cycle; At the end of the receiving sub-period, the write I/O data identified by the tag corresponding to the receiving sub-period is copied to the destination end through the communication interface. 25. The controller according to claim 24, characterized in that the program is also used to: A tag startup message broadcast by a tag controller in the cluster storage system is received through the communication interface. The tag startup message includes the tag, and the tags broadcast by the tag controller are different at different tag generation time points. 26. The controller according to claim 25, characterized in that the program is also used to: When the receiving sub-period ends, or when the identification stop message broadcast by the tag controller for the tag corresponding to the ending receiving sub-period is received through the communication interface, stop corresponding to the receiving sub-period according to the ending. The label identifies the write I/O data issued by the host. 27. The controller according to claim 25 or 26, characterized in that the program is also used for: When the tag startup message is received through the communication interface, if there is no write I/O data sent by the host, the tag in the tag startup message is discarded. 28. A data replication system, characterized by comprising: the storage controller according to any one of claims 10 to 14 and the storage controller according to any one of claims 15 to 18.