JP2005258789A - Storage device, storage controller, and write back cache control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To inhibit the frequent occurrence of competition between a host I/O and destage on a cache block. <P>SOLUTION: A cache management module 33 receives a writing request accompanied with start of commit processing or check point processing from an NAS (Network Attached Storage) controller 10 and stores requested data on the cache block of a cache memory 32. The cache management module 33 inhibits the cache block (namely, the cache block equivalent to a log region) storing data forming the log region 311 to an inherent writing destination so that the cache block becomes the object of a destage operation when implementing the destage operation for writing data on the updated cache block on the cache memory 32 into an inherent writing destination on a disk 31 keeping the log region 311. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a storage device that includes a secondary storage area in which a log area is secured and a write-back cache that temporarily stores data to be written to the secondary storage area, and is stored in the write-back cache. The present invention relates to a storage apparatus, a storage controller, and a write-back cache control method suitable for cache control including a destage operation for writing a data block to the secondary storage area.

  2. Description of the Related Art A network attached storage controller (hereinafter referred to as NAS controller) is known as a host (higher level apparatus) of a storage apparatus having a secondary storage area. A device composed of this NAS controller and storage device is called a NAS device. The NAS controller functions as a server represented by a file server or a database server. The NAS controller is connected to a client machine (client computer) that uses the storage device via the controller via a network such as a LAN (local area network).

  A NAS controller generally has a write-back cache. However, the NAS controller cannot perform a simple write-back operation when accessing the storage apparatus by a network file system operation from a client machine, such as a so-called NFS operation. That is, when the NAS controller receives data transmitted from the client machine, it needs to return reception completion to the client machine after ensuring that the data has been written to the storage device (secondary storage area). The reason is that if the data received by the NAS controller from the client machine is temporarily stored in the write-back cache in the controller and the reception completion is returned to the client machine, the following problems may occur. Because. In other words, if the system goes down before the NAS controller actually writes data to the storage device, the data that should have been received is lost.

  Incidentally, a journal file system is known as a technique for ensuring consistency at the file system level of a system and improving input / output (I / O) performance with respect to a storage apparatus. The journal file system performs the following operation when it is necessary to perform an operation to change the data image in the secondary storage area of the storage device in response to a data write (data update) request from the client machine. First, the journal file system temporarily stores the requested data in a specific area in the secondary storage area called a log area via the write-back cache of the NAS controller, and returns a write completion to the client machine. At an appropriate timing thereafter, the journal file system actually writes the data requested by the write request in the location that should be updated in the secondary storage area. The data at the location that should be updated is called actual data, and the data writing at this location is called actual data writing or actual data updating. For this update, data stored in the write-back cache of the NAS controller is used instead of data stored in the log area in order to speed up processing.

  The log area is allocated to areas of continuous logical addresses, and is used as a ring buffer for storing data (data update history) in chronological order. A collection of data stored in the log area is called a journal file. Saving data in the log area is called “commit processing”. Also, updating the actual data is called “checkpoint processing”.

  In an operation that has undergone a commit process, it is guaranteed that the corresponding actual data will be updated. On the other hand, actual data is not updated in an operation that has not yet been committed. In particular, commit processing is collected as a set of operations that are meaningful in the file system, and processing involving a plurality of I / Os is often handled as one commit unit. Therefore, even if an unexpected system failure occurs, it is possible to guarantee the consistency between the operation and the data image in the secondary storage area in units of commits.

  During the commit process, the actual data is not updated for the corresponding operation, and the completion of writing to the host device is not returned. For this reason, if the system goes down before the commit process is completed, it is equivalent to the above operation not being performed. If the commit process is completed, the completion of writing can be returned to the host device without passing through the checkpoint. However, in order to make this possible, it must be ensured that the data once the commit process is completed will be reflected in the actual data. Therefore, the journal file system has a function that guarantees that when there is data that has been committed but the checkpoint process is incomplete due to a system down, the data is reflected in the actual data when the system is rebooted. Required.

  Recent NAS controllers apply a journal file system having the above-described functions. This type of NAS controller can return a reception completion to the client machine when the data is completely stored in the log area of the storage apparatus. Therefore, this type of NAS controller can perform a write-back operation that is advantageous in terms of performance after the commit process, and can guarantee consistency at the file system level.

  An important point that determines the I / O performance to the storage apparatus in response to a data write request from a client machine is the responsiveness of commit processing, that is, writing to a log area (log writing). If the commit process is slow, the process completion notification to the client machine is delayed accordingly.

  Incidentally, recent storage apparatuses generally have a write-back cache (a disk cache to which a write-back method is applied) (see, for example, Patent Document 1). The write-back cache of this storage device is designed so that the data on the cache is protected even when the system is down unexpectedly by battery backup or the like. Therefore, it is guaranteed that data once stored in the write-back cache of the storage apparatus will be written to the secondary storage area of the storage apparatus.

Therefore, recently, at the time of log writing, the data to be written to the log area is temporarily stored in the location in the write-back cache of the storage device corresponding to the write destination, and the commit process is completed. The performance is improved. When this storage device write-back cache is used, the checkpoint processing also uses the storage device write-back cache location corresponding to the location in the secondary storage area where the data on the write-back cache of the NAS controller should be updated. You can complete it by simply saving it in
JP-A-10-293717 (paragraphs 0003 to 0005)

  In the above-described conventional technology, a host (for example, NAS controller) that uses a storage device performs commit processing and checkpoint processing by using the write-back cache of the storage device, thereby performing the commit processing and checkpoint processing. Speeding up.

  The data stored in the write-back cache of the storage device is stored in the secondary storage area of the storage device by the storage device itself asynchronously with the host I / O operation (hereinafter referred to as host I / O). Is written to. This writing is called destage or destage operation.

  Now, the data block (that is, cache block) on the write-back cache that is the target of the destage operation is in a reserved state, and data cannot be updated by the host I / O. Therefore, when a cache block assigned to a certain logical address is being destaged, the host I / O for the same logical address is waited until the destage operation is completed and the reserved state of the cache block is released. That is, in a storage apparatus equipped with a write-back cache, contention on the cache block between the host I / O and the destage occurs. The occurrence of this competition greatly affects the performance of the storage apparatus. In particular, when the host employs a journal file system, this contention occurs frequently, which is a problem. The inventor of the present invention repeatedly causes log writing by the host I / O (writing to a logical address corresponding to the log area secured in the secondary storage area of the storage device) as a cause of the frequent occurrence of this contention. It came to recognize that there is.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a storage apparatus, a storage controller, and a storage device capable of suppressing frequent occurrence of competition between the host I / O and the destage on the cache block. The object is to provide a write-back cache control method.

  According to one aspect of the present invention, a secondary storage area in which a log area with continuous logical addresses is secured for storing a history of data updates from a host in time series, and writing to the secondary storage area is performed. There is provided a storage apparatus including a write-back cache managed in units of cache blocks for temporarily storing data to be stored. The storage device receives a write request accompanying the start of commit processing or checkpoint processing of the host, and stores data requested in the cache block on the write-back cache, and the write-back cache Destage control means for controlling a destage operation for writing the updated cache block data to the original write destination on the secondary storage, wherein the log area is the original write destination. Destaging control means for suppressing a cache block corresponding to a log area in which data on the write-back cache is stored as a target of the destage operation.

  In the above configuration, when a write request accompanying the start of commit processing or checkpoint processing is sent from the host to the storage device, the storage device writes data to the secondary storage area according to the write request. Instead, an operation (host I / O) for writing (storing) data in a cache block on the write-back cache corresponding to the original write destination in the secondary storage area is performed. Here, the write destination of the secondary storage area specified by the write request from the host is either in the log area or outside the log area (that is, in the actual data area).

  The destage control means performs a destage operation for writing the cache block data on the write-back cache updated by the data write (host I / O) to the original write destination in the secondary storage area. During execution, the cache block storing the data on the write-back cache whose original write destination is in the log area is prevented from being the target of the destage operation.

  In general, the ratio of the log area to the secondary storage area is very small. The log area is used in a ring shape. Therefore, the frequency at which each block in the log area is used is significantly higher than the frequency at which each block outside the log area (that is, in the actual data area) is used. Therefore, in the above configuration, the cache block storing the data on the write-back cache whose original write destination is the log area, that is, the cache block corresponding to the log area, is prevented from being the target of the destage operation. By doing so, it is possible to avoid as much as possible that the cache block is frequently reserved. As a result, it is possible to suppress frequent occurrence of contention between the host I / O and the destage on the cache block, and it is possible to suppress delay in response time of log writing and improve the performance as a storage device.

  Here, in order to prevent the cache block corresponding to the log area from being the target of the destage operation, the destage control means uses each cache block on the write-back cache as a cache block corresponding to the log area. It is only necessary to be able to determine or estimate whether there is a cache block other than that (a cache block corresponding to the actual data area). For example, if the host notifies the storage apparatus of the logical address range to which the log area is allocated, the above determination is possible. However, if the frequency that each block in the log area is used is significantly higher than the frequency that each block in the actual data area is used, the cache block that is used frequently is used Even if the cache block is estimated to be a considerable cache block, the probability that it is erroneously estimated is extremely low. Therefore, it is possible to effectively suppress frequent occurrence of competition between the host I / O and the destage on the cache block.

  According to the present invention, the host I / O and the destage are suppressed by suppressing the cache block storing the data on the write-back cache whose original write destination is the log area from being subject to the destage operation. Thus, frequent occurrence of contention on the cache block can be suppressed, and a delay in response time of log writing can be suppressed to improve performance as a storage apparatus.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a storage system according to an embodiment of the present invention. In FIG. 1, the NAS controller 10 is connected to the storage device 30 via, for example, a SCSI (Small Computer System Interface) bus 20. It is also possible to use an FC (Fibre Channel) bus instead of the SCSI bus 20. The NAS controller 10 is also connected to client machines (hereinafter referred to as clients) 41-1 to 41-N via the network 40.

  The NAS controller 10 functions as a file server for realizing a journal file system, for example, and provides the storage area of the storage device 30 to the clients 41-1 to 41-N via the network 40. The NAS controller 10 includes a storage management module 11 for managing the storage device 30 and a cache memory 12 used as a cache (that is, a write-back cache) to which a write-back method is applied. The storage management module 11 includes a file system (not shown) that manages files including journal files.

  The storage device 30 includes a disk 31 for realizing a secondary storage area and a storage controller SC for controlling access to the disk 31. The storage controller SC includes a cache memory 32 that is used as a write-back cache (disk cache to which a write-back method is applied), and a cache management module 33 that manages the cache memory 32. The cache management module 33 is realized by the storage device 30 reading and executing a specific software program installed in the storage device 30. This program is stored in a computer-readable storage medium such as a magnetic disk represented by a floppy (registered trademark) disk, an optical disk represented by a CD-ROM or a DVD, or a semiconductor memory represented by a flash memory. It can be stored in advance and distributed. Further, this program may be downloaded (distributed) via a network.

  The disk 31 may be a physical disk or a logical disk as an aggregate of the physical disks. In the area (secondary storage area) of the disk 31, a log area 311 and an actual data area 312 are secured. The areas on the disk 31 allocated to the log area 311 and the actual data area 312 are specified by the storage management module 11 of the NAS controller 10. The log area 311 is assigned to a continuous logical address area. The log area 311 is used as a ring buffer in which logical addresses are continuous for storing a history of data update in chronological order.

  The cache memory 32 is configured such that data on the cache is protected even when the system is down due to battery backup or the like. Therefore, even if a volatile memory is used for the cache memory 32 itself, the cache memory 32 is functionally a kind of rewritable nonvolatile memory. The cache memory 32 is, for example, an m-way set associative cache (m is an integer of 2 or more).

  FIG. 2 shows a configuration (cache configuration) of the cache memory 32. As shown in the figure, the area of the cache memory 32 is divided into m rows (rows) and n digits (columns), that is, divided into n × m cache blocks. In this embodiment, a partial area of the cache memory 32, for example, k lines (k sets) of the 0th to k−1th lines (that is, the set 0 to k−1) correspond to the log area 311 of the disk 31. Assigned to a specific area (hereinafter referred to as a log equivalent area) 321. The remaining area of the cache memory 32, that is, mk lines (mk sets) of the kth to m−1th lines (that is, the sets k to m−1) correspond to the actual data area 312 of the disk 31. It is assigned to a specific area (hereinafter referred to as actual data equivalent area) 322.

Next, the operation of the system of FIG. 1 will be described.
First, the operation at the time of system boot of the NAS controller 10 will be described with reference to the sequence chart of FIG.

  The storage management module 11 of the NAS controller 10 notifies the logical address of the area on the disk 31 used as the log area 311 on the disk 31 to the cache management module 33 of the storage apparatus 30 at the time of system boot (step S1). . Then, the cache management module 33 functions as a log equivalent area allocation unit, and allocates an area on the cache memory 32 corresponding to the logical address of the log area 311 notified by the storage management module 11 as a log equivalent area 321 (step S2). In this case, an area other than the log equivalent area 321 on the cache memory 32 can be used as a cache area for actual data corresponding to the actual data area 312 on the disk 31, that is, an actual data equivalent area 322.

Next, an operation when a data write request from the client 41-1 to the NAS controller 10 occurs will be described with reference to the sequence chart of FIG.
Assume that a data write request is transmitted from the client 41-1 to the NAS controller 10 (step S11). The storage management module 11 of the NAS controller 10 receives the requested write data in response to a data write request from the client 41-1, and starts a commit process (step S12).

  The storage management module 11 first stores the received data in the cache memory 12 and sends a write request (here, log write request) to the storage device 30 (step S13). In this case, the logical address of the write destination (log write destination) specified by the write request (log write request) is within the range of the logical address of the log area notified from the storage management module 11 of the NAS controller 10 at the time of system boot. In other words, the log area 311 secured on the disk 31 falls within the logical address range.

  When the cache management module 33 of the storage device 30 receives a write request from the storage management module 11 of the NAS controller 10, whether or not the logical address of the write destination specified by the request is within the logical address range of the log area 311. Determine whether. As in this example, when the logical address of the write destination requested from the storage management module 11 is within the logical address range of the log area 311, the cache management module 33 determines that the write request is a log write request. To do. In this case, the cache management module 33 functions as a data storage unit and stores the requested data in the log equivalent area 321 on the cache memory 32 (step S14). Here, the requested data is stored in the cache block of the corresponding column in the i-th (i is one of 0 to k−1) row (set i) in the log equivalent area 321 in units of blocks. The After step S14, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S15).

  When the write completion (here, log write completion) is notified from the cache management module 33 of the storage apparatus 30, the storage management module 11 completes the commit process and notifies the client 41-1 of the write completion (step S16, S17).

  Thereafter, the storage management module 11 of the NAS controller 10 starts checkpoint processing (step S18). First, the storage management module 11 stores the data updated in the cache memory 12 (that is, dirty data in which the update on the cache memory 12 is not reflected in the storage device 30) in the actual data area 312 of the storage device 30. In order to write data, a write request (actual data write request) is issued to the cache management module 33 of the storage device 30 (step S19). In this case, the logical address of the write destination (actual data write destination) specified by the write request is out of the logical address range of the log area 311.

  When the cache management module 33 of the storage device 30 receives a write request from the storage management module 11 of the NAS controller 10, whether or not the logical address of the write destination specified by the request is within the logical address range of the log area 311. Determine whether. When the logical address of the write destination requested from the storage management module 11 is out of the logical address range of the log area 311 as in this example, the cache management module 33 indicates that the write request is not a log write request, that is, real data It is determined that it is a write request. In this case, the cache management module 33 functions as a data storage unit and stores the requested data in the real data equivalent area 322 on the cache memory 32 (step S20). Here, the requested data is stored in the cache block of the corresponding column in the jth (j is any one of k to m−1) row (set j) in the actual data equivalent area 322 in units of blocks. Is done. After step S20, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S21).

  When the storage management module 11 is notified of the writing completion (actual data writing completion in this case) from the cache management module 33 of the storage device 30, the storage management module 11 corresponds to the actual data for which writing has been completed. The data is discarded and the checkpoint process is completed (step S22).

  The cache management module 33 of the storage device 30 functions as a destage control means asynchronously with the I / O (host I / O) from the NAS controller 10 to the storage device 30, and performs a destage operation at a predetermined timing. I do. The feature of this embodiment is that this destage operation is reflected in the storage device 30 by the data updated in the real data equivalent area 322 of the cache memory 32 (that is, the update in the real data equivalent area 322 of the cache memory 32 is reflected in the storage device 30). (Dirty data that has not been performed) is performed only on the target (step S23). In other words, the feature of this embodiment is that the data block in the log equivalent area 321 of the cache memory 32 is not subject to destage. The predetermined timing is assumed to be a regular timing. In addition, when attempting to store data in the actual data equivalent area 322 of the cache memory 12 in step S20, all the cache blocks of the corresponding columns in the actual data equivalent area 322 are dirty blocks. That is, the case where there is no empty block in the corresponding column may be set as the above-described predetermined timing.

  As a result of the above processing, a destage for the log equivalent area 321 of the cache memory 32 of the storage device 30 occurs during the data transfer between the normal NAS controller 10 and the clients 41-1 to 41-N. Can be suppressed. That is, it is possible to suppress frequent occurrence of contention between the host I / O and the destage on the cache block. As a result, even if log writing is frequently performed, it is possible to suppress the corresponding cache block on the cache memory 32 from being reserved, and to reduce the log writing delay. Performance can be improved.

  In the case of the shutdown process of the NAS controller 10 or the cache sync process requested from the NAS controller 10 (that is, the process for reflecting the contents of the cache memory 32 on the disk 31), the cache management module 33 stores the log equivalent area. All the updated data on the cache memory 32 is written to the disk 31 regardless of the actual data equivalent area 322 and 321.

[First Modification]
Next, a first modification of the above embodiment will be described. In the first modification, unlike the above embodiment, the area on the disk 31 used as the log area 311 on the disk 31 from the storage management module 11 of the NAS controller 10 to the cache management module 33 of the storage apparatus 30. The logical address of is not notified. For this reason, the cache management module 33 cannot divide and use the cache memory area of the cache memory 32 into the log equivalent area 321 and the actual data equivalent area 322. Therefore, in the first modified example, among the cache blocks of the cache memory 32, data of a cache block (hereinafter referred to as a log area equivalent block or a log area equivalent cache block) to which a block in the log area 311 is allocated (hereinafter referred to as a log area equivalent block or a log area equivalent cache block). That is, log data) is frequently updated, so that it is possible to suppress the occurrence of destage for the log data on the cache memory 32. That is, the feature of the first modification is that a block to be destaged among the cache blocks of the cache memory 32 is selected according to the access frequency.

  Therefore, the cache management module 33 of the storage apparatus 30 has an access frequency management table 331 shown in FIG. 5 in order to manage the access frequency of each cache block in the cache memory 32. Each entry of the access frequency management table 331 is associated with each cache block Bij (i = 0 to m−1, j = 0 to n−1) of the cache memory 32, and the access frequency of the block Bij is set. Access frequency information Fij for management is held. The access frequency information Fij indicates the number of times the corresponding block has been accessed (updated) in a certain time. However, when the cache block data is destaged, the access frequency information Fij is initialized, and the access frequency indicated by the access frequency information Fij is zero.

  Next, referring to the sequence chart of FIG. 6 for the operation of the first modification of the above embodiment, taking the case where a data write request from the client 41-1 to the NAS controller 10 is generated as in the above embodiment. To explain. .

  In response to the data write request from the client 41-1 (step S31), the storage management module 11 of the NAS controller 10 receives the requested write data and starts the commit process (step S32). The storage management module 11 stores the received data in the cache memory 12 and sends a write request (here, a log write request) to the storage device 30 (step S33).

  When the cache management module 33 of the storage device 30 receives a write request from the storage management module 11 of the NAS controller 10, it functions as a data storage unit and stores the requested data in a cache block on the cache memory 32 ( Step S34a). At the same time, the cache management module 33 functions as an access frequency management means, and updates the access frequency information held in the entry of the access frequency management table 331 corresponding to the cache block in which the requested data is stored (step S34b). . Thereafter, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S35).

  When the write completion (here, log write completion) is notified from the cache management module 33 of the storage apparatus 30, the storage management module 11 completes the commit process and notifies the client 41-1 of the write completion (step S36, S37).

  Thereafter, the storage management module 11 of the NAS controller 10 starts checkpoint processing (step S38). The storage management module 11 first issues a write request (actual data write request) to the cache management module 33 of the storage apparatus 30 in order to write the data updated on the cache memory 12 to the real data area 312 of the disk 31. (Step S39).

  When the cache management module 33 of the storage device 30 receives a write request (actual data write request) from the storage management module 11 of the NAS controller 10, it functions as a data storage unit, and the requested data is stored in the cache memory 32. The data is stored in the corresponding cache block (step S40a). At the same time, the cache management module 33 functions as an access frequency management means to update the access frequency information held in the entry of the access frequency management table 331 corresponding to the cache block in which the requested data is stored (step S40b). . Thereafter, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S41).

  When the storage management module 11 is notified from the cache management module 33 of the storage device 30 that the writing is completed (here, the actual data writing is completed), the storage management module 11 discards the data on the log that has been written and completes the checkpoint process. (Step S42).

  The cache management module 33 of the storage device 30 functions as a destage control means asynchronously with the I / O (host I / O) from the NAS controller 10 to the storage device 30, and performs de-decoding at the same timing as in the above embodiment. Stage processing (step S43) is performed. A specific procedure of the destage processing (step S43) will be described with reference to the flowchart of FIG.

  First, the cache management module 33 refers to the access frequency information Fij held in each entry of the access frequency management table 331 (step S43a). Next, based on the access frequency of each cache block Bij in the cache memory 32 indicated by each access frequency information Fij, the cache management module 33 selects a cache block whose access frequency is lower than a certain level among the cache blocks. Is selected on the assumption that it is a cache block to which a block in the actual data area 312 is allocated (hereinafter referred to as a block corresponding to the actual data area or a cache block corresponding to the actual data area) (step S43b). The cache management module 33 performs a destage operation only on the selected cache block, that is, a cache block whose access frequency is lower than a certain level and estimated to be a cache block corresponding to the actual data area (step S43c). ). The access frequency information of the destaged cache block is initialized. Here, a cache block whose access frequency exceeds a certain level is estimated as a cache block corresponding to the log area, and is excluded from the destage target.

  As a result of the above processing, a cache block whose access frequency exceeds a certain level is not a destage target. Here, the proportion of the log area 311 in the storage area (secondary storage area) of the disk 31 is significantly smaller than the actual data area 312. However, the log write request and the actual data write request are generated the same number of times. This means that if the access frequency of each block in the log area 311 is represented by F1, and the access frequency of each block in the actual data area 312 is represented by F2, F1 >> F2. Therefore, a block having a high access frequency among the cache blocks of the cache memory 32 of the storage apparatus 30 has a high probability of being a cache block corresponding to the log area 311 (a cache block corresponding to the log area). Similarly, a block with low access frequency among the cache blocks of the cache memory 32 has a high probability of being a cache block corresponding to the real data area 312 (a cache block corresponding to the real data area). Therefore, in the first modified example, among the cache blocks of the cache memory 32, there is a high probability that a cache block corresponding to the actual data area is a destage target. Therefore, according to the first modification, the data corresponding to the block corresponding to the log area on the cache memory 32 during the data transfer between the normal NAS controller 10 and the clients 41-1 to 41-N is processed. It is possible to suppress the occurrence of a stage (that is, log data destage). As a result, even if log writing is frequently performed, it is possible to suppress the corresponding cache block on the cache memory 32 from being reserved and to reduce the log writing delay.

  In the case of the shutdown process of the NAS controller 10 or the cache sync process requested by the NAS controller 10, the cache management module 33 stores all the updated data on the cache memory 32 regardless of the access frequency. Export to

[Second Modification]
Next, a second modification of the above embodiment will be described. The feature of the second modification is that instead of managing the access frequency of each cache block of the cache memory 32, the order of the cache blocks to be destaged is managed according to the LRU (Least Recently Used) rule. In the point.

  Therefore, the cache management module 33 of the storage device 30 is shown in FIG. 8 as a destage waiting block management table for managing the order of cache blocks (that is, dirty blocks) to be destaged on the cache memory 32. It has a destage queue (destage queue) 332. The destage queue 332 is a column of identifiers (block IDs) of each dirty block Bij on the cache memory 32, and represents a column of dirty blocks to be destaged. The arrangement order of block IDs in the destage queue 332, that is, the arrangement order of dirty blocks is managed according to the LRU rule.

  Next, an outline of the operation of the second modification will be described. First, the storage management module 11 of the NAS controller 10 determines the size (that is, log size) of the area on the disk 31 to be used as the log area 311 on the disk 31 with respect to the cache management module 33 of the storage apparatus 30 at the time of system boot. Notice. Here, the cache memory area of the cache memory 32 is not divided into a log equivalent area 321 and a real data equivalent area 322.

  The cache management module 33 of the storage apparatus 30 manages the order of dirty blocks that are targets of the destage operation in the cache memory 32 using the destage queue 332 that is updated according to the LRU rule. That is, the cache management module 33 lowers the priority as a destage target for a cache block that is frequently used even if it is a dirty block. For this reason, a cache block with high use frequency, such as a cache block corresponding to a log area, has a lower destaging priority than other dirty blocks. Further, when the total capacity of dirty blocks in the cache memory 32 is equal to or smaller than the log size notified from the storage management module 11 of the NAS controller 10, the cache management module 33 is equivalent to all log blocks in the cache memory 32. Destage is not performed because there is a high possibility that this is a cache block. With the above mechanism, it is possible to avoid selecting a cache block corresponding to a log area with high use frequency as a destage target.

  Next, details of the operation of the second modification will be described with reference to the sequence chart of FIG. 9, taking as an example the case where a data write request from the client 41-1 to the NAS controller 10 is generated.

  In response to a data write request from the client 41-1 (step S51), the storage management module 11 of the NAS controller 10 receives the requested write data and starts commit processing (step S52). The storage management module 11 stores the received data in the cache memory 12 and sends a write request (here, a log write request) to the storage device 30 (step S53).

  When the cache management module 33 of the storage device 30 receives a write request from the storage management module 11 of the NAS controller 10, it functions as a data storage unit and stores the requested data in a cache block on the cache memory 32 ( Step S54a). At the same time, the cache management module 33 functions as a destage wait block management means, and connects the ID of the cache block (that is, dirty block) in which the requested data is stored to the end of the destage queue 332 (step S54b). . In the following description, in order to simplify the description, the destage queue 332 is treated as a column of cache blocks (dirty blocks). Thereafter, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S55).

  When the write completion (here, log write completion) is notified from the cache management module 33 of the storage device 30, the storage management module 11 completes the commit process and notifies the client 41-1 of the write completion (step S56, S57).

  Thereafter, the storage management module 11 of the NAS controller 10 starts checkpoint processing (step S58). The storage management module 11 first issues a write request (actual data write request) to the cache management module 33 of the storage apparatus 30 in order to write the data updated on the cache memory 12 to the real data area 312 of the disk 31. (Step S59).

  When the cache management module 33 of the storage device 30 receives a write request (actual data write request) from the storage management module 11 of the NAS controller 10, it functions as a data storage unit, and the requested data is stored in the cache memory 32. The data is stored in the corresponding cache block (step S60a). At the same time, the cache management module 33 functions as a destage wait block management means, and connects the cache block storing the requested data to the end of the destage queue 332 (step S60b). Thereafter, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S61).

  When the storage management module 11 is notified from the cache management module 33 of the storage device 30 that the writing is completed (here, the actual data writing is completed), the storage management module 11 discards the data on the log that has been written and completes the checkpoint process. (Step S62).

  The cache management module 33 of the storage device 30 functions as a destage control means asynchronously with the I / O (host I / O) from the NAS controller 10 to the storage device 30, and at a predetermined timing (for example, periodically) The destage processing (step S63) is performed. A specific procedure of this destage processing (step S63) will be described with reference to the flowchart of FIG.

  First, the cache management module 33 counts the number of dirty blocks in the cache memory 32, and calculates the total capacity of the dirty blocks from the number of dirty blocks (steps S63a and S63b). Next, the cache management module 33 determines whether the total capacity of the dirty block exceeds the log size notified from the storage management module 11 of the NAS controller 10 (step S63c). The cache management module 33 determines that the dirty block in the cache memory 32 includes a cache block equivalent to the actual data area only when the total capacity of the dirty block exceeds the log size. A cache block is selected (step S63d). Here, among the dirty blocks managed by the stage queue 332, the first dirty block that exceeds the log size, that is, the dirty block whose time has elapsed since the last access is the actual data The cache block corresponding to the area is estimated and selected. On the other hand, the dirty block on the tail side of the stage queue 332, that is, the dirty block that is frequently accessed, is estimated to be a cache block corresponding to the log area and is excluded from the destage target. The cache management module 33 performs a destage operation on the dirty block selected in step S63c (step S63e). The destaged block is removed from the destage queue 332.

  With the above processing, it is possible to suppress selection of a cache block corresponding to a log area with high use frequency on the cache memory 32 as a destage target. Therefore, according to the second modification, it is possible to suppress the occurrence of log data destage during the data transfer between the normal NAS controller 10 and the clients 41-1 to 41-N. . As a result, even if log writing is frequently performed, it is possible to suppress the corresponding cache block on the cache memory 32 from being reserved and to reduce the log writing delay.

  In the case of the shutdown process of the NAS controller 10 or the cache sync process requested from the NAS controller 10, the cache management module 33 updates all the cache memory 32 regardless of the state of the destage queue 332. The recorded data is written to the disk 31.

[Third Modification]
Next, a third modification of the above embodiment will be described. The feature of the third modification is that the order of cache blocks (dirty blocks) to be destaged is managed according to the LRU rule for each column (cache column) of the cache memory 32.

  Therefore, the cache management module 33 of the storage apparatus 30 corresponds to the columns 0 to n−1 of the cache memory 32 and corresponds to the destage queues (destage queues) 332-0 to 332- (n−1) shown in FIG. ). The destage queue 332-j (j = 0 to n-1) is a column of the block IDs of the dirty blocks Bij belonging to the column j of the cache memory 32, and the dirty block to be destaged in the column j Represents a column. The arrangement order of block IDs in the destage queue 332-j, that is, the arrangement order of dirty blocks is managed according to the LRU rule.

  Next, an outline of the operation of the third modification will be described. First, the storage management module 11 of the NAS controller 10 sends the cache memory of the storage device 30 in the area on the disk 31 used as the log area 311 on the disk 31 to the cache management module 33 of the storage apparatus 30 at the time of system boot. The ratio R to the total capacity of 32 is notified. Here, the cache memory area of the cache memory 32 is not divided into a log equivalent area 321 and a real data equivalent area 322.

  As shown in FIG. 2, the cache memory 32 is an m-way set associative cache. A cache block having the same hash value for a logical address is assigned to each column (cache column) j of the cache memory 32. For this reason, cache blocks corresponding to areas continuous on logical addresses, such as the log area 311, are uniformly distributed and allocated to each column. In the above embodiment, the first modification, and the second modification, the cache memory 32 does not necessarily have to be an m-way set associative cache.

  The cache management module 33 of the storage apparatus 30 manages the order of cache blocks to be destaged in the cache memory 32 by using a destage queue 332-j that is updated according to the LRU rule for each column j. In other words, the cache management module 33 lowers the priority as a destage target even if a cache block having a high use frequency among the cache blocks belonging to the column j is a dirty block. For this reason, a cache block with high use frequency, such as a cache block corresponding to a log area, has a lower destaging priority than other dirty blocks. Furthermore, the cache management module 33 determines that the dirty block belonging to column j is in a state where the ratio Rj of the number of dirty blocks to the number of cache blocks belonging to column j is equal to or less than the ratio R notified from the storage management module 11 of the NAS controller 10. Are not likely to be cache blocks corresponding to the log area, and destage is not performed. With the above mechanism, it is possible to avoid selecting a cache block corresponding to a frequently used log area as a destage target.

  Next, details of the operation of the third modification will be described with reference to the sequence chart of FIG. 12, taking as an example a case where a data write request from the client 41-1 to the NAS controller 10 is generated.

  In response to the data write request from the client 41-1 (step S71), the storage management module 11 of the NAS controller 10 receives the requested write data and starts the commit process (step S72). The storage management module 11 stores the received data in the cache memory 12 and sends a write request (here, a log write request) to the storage device 30 (step S73).

  When the cache management module 33 of the storage device 30 receives a write request from the storage management module 11 of the NAS controller 10, it functions as a data storage unit and stores the requested data in a cache block on the cache memory 32 ( Step S74a). This cache block is selected from clean cache blocks belonging to the column j on the cache memory 32 determined by the logical address specified by the write request (that is, the cache block whose data update has already been reflected in the disk 31). The At the same time, the cache management module 33 functions as a destage wait block management means, and connects the cache block (ID) in which the requested data is stored to the end of the destage queue 332-j (step S74b). Thereafter, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S75).

  When the write completion (here, log write completion) is notified from the cache management module 33 of the storage device 30, the storage management module 11 completes the commit process and notifies the client 41-1 of the write completion (Step S76, S77).

  Thereafter, the storage management module 11 of the NAS controller 10 starts checkpoint processing (step S78). The storage management module 11 first issues a write request (actual data write request) to the cache management module 33 of the storage apparatus 30 in order to write the data updated on the cache memory 12 to the real data area 312 of the disk 31. (Step S79).

  When the cache management module 33 of the storage device 30 receives a write request (actual data write request) from the storage management module 11 of the NAS controller 10, it functions as a data storage unit, and the requested data is stored in the cache memory 32. It is stored in the cache block belonging to the corresponding column (step S80a). At the same time, the cache management module 33 functions as a destage wait block management means, and stores the cache block storing the requested data into the corresponding destage queues 332-0 to 332- (n-1). Connect to the end of the stage queue (step S80b). Thereafter, the cache management module 33 notifies the storage management module 11 of the NAS controller 10 of the completion of writing (step S81).

  When the storage management module 11 is notified from the cache management module 33 of the storage device 30 that the writing is completed (here, the actual data writing is completed), the storage management module 11 discards the data on the log that has been written and completes the checkpoint process. (Step S82).

  The cache management module 33 of the storage device 30 functions as a destage control means asynchronously with the I / O (host I / O) from the NAS controller 10 to the storage device 30, and at a predetermined timing (for example, periodically) Destage processing (step S83) is performed. A specific procedure of this destage processing (step S83) will be described with reference to the flowchart of FIG.

  First, the cache management module 33 counts the number of dirty blocks for each column j of the cache memory 32, and calculates the ratio Rj of the number of dirty blocks to the number of cache blocks belonging to the column j (steps S83a, S83b, S83c). , S83g, S83h). Next, the cache management module 33 determines, for each column j, whether the ratio Rj of the number of dirty blocks exceeds the ratio R notified from the storage management module 11 of the NAS controller 10 (step S83d). The cache management module 33 selects a cache block to be destaged on the assumption that the dirty block in the column j includes a cache block corresponding to the actual data area only for the column j in which Rj exceeds R. (Step S83e). Here, of the dirty blocks managed by the destage queue 332-j, the first dirty block that exceeds the ratio R, that is, the dirty block whose time has elapsed since the last access, The cache block corresponding to the actual data area is estimated and selected. On the other hand, the dirty block on the tail side of the stage queue 332-j, that is, the frequently accessed dirty block, is estimated to be a cache block corresponding to the log area and is excluded from the destage target. The cache management module 33 performs a destage operation on the dirty block selected in step S83c (step S83f). The destaged cache block is removed from the destage queue 332-j.

  By the above processing, it is possible to suppress the cache block corresponding to the log area with high use frequency on the cache memory 32 from being selected as a destage target in units of columns. Therefore, according to the second modification, it is possible to suppress the occurrence of log data destage during the data transfer between the normal NAS controller 10 and the clients 41-1 to 41-N. . As a result, even if log writing is frequently performed, it is possible to suppress the corresponding cache block on the cache memory 32 from being reserved and to reduce the log writing delay.

  In the case of the shutdown process of the storage apparatus 30 or the cache sync process requested from the NAS controller 10, the cache management module 33 does not depend on the state of the destage queues 332-0 to 332- (n-1). All the updated data on the cache memory 32 is written to the disk 31.

  The present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment or a modification.

1 is a block diagram showing a configuration of a storage system according to an embodiment of the present invention. The block diagram which shows the structure of the cache memory 32 in FIG. The sequence chart which shows the operation | movement procedure at the time of system boot of the NAS controller 10 in the embodiment. 6 is a sequence chart showing an operation procedure when a data write request from the client 41-1 to the NAS controller 10 is generated in the embodiment. The figure which shows the data structure example of the access frequency management table 331 applied in the 1st modification of the embodiment. 9 is a sequence chart showing an operation procedure when a data write request is issued from the client 41-1 to the NAS controller 10 in the first modification. The flowchart which shows the specific procedure of the destage process (step S43) in FIG. The figure which shows the example of a data structure of the destage queue 332 applied in the 2nd modification of the embodiment. 11 is a sequence chart showing an operation procedure when a data write request is issued from the client 41-1 to the NAS controller 10 in the second modified example. 10 is a flowchart showing a specific procedure of destage processing (step S63) in FIG. The figure which shows the data structure example of the destage queue 332-0-332- (n-1) applied in the 3rd modification of the embodiment. 11 is a sequence chart showing an operation procedure when a data write request is issued from the client 41-1 to the NAS controller 10 in the third modified example. The flowchart which shows the specific procedure of the destage process (step S83) in FIG.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... NAS controller (host), 11 ... Storage management module, 12 ... Cache memory, 30 ... Storage apparatus, 31 ... Disk (secondary storage area), 32 ... Cache memory (write-back cache), 33 ... Cache management module, 40 ... Network, 41-1 to 41-N ... Client, 311 ... Log area, 312 ... Real data area, 321 ... Log equivalent area, 322 ... Real data equivalent area, 331 ... Access frequency management table, 332, 332-0 332- (n-1) ... Destage queue, SC ... Storage controller.

Claims (9)

A secondary storage area that secures a log area with continuous logical addresses for storing the history of data update from the host in time series order, and a cache for temporarily storing data to be written to the secondary storage area In a storage device with a write-back cache managed in block units,
Data storage means for receiving a write request accompanying the start of the commit process or checkpoint process of the host and storing the requested data in the cache block on the write-back cache;
Destage control means for controlling a destage operation for writing the updated cache block data on the write-back cache to the original write destination on the secondary storage, and the original write in the log area And a destage control unit that suppresses a cache block corresponding to a log area in which data on the write-back cache is stored as a target of the destage operation. The write-back cache is a set associative cache managed by rows and columns,
The storage device manages the order of dirty cache blocks that are the target of the destage operation on the write-back cache by using a destage queue for each column that is updated according to an LRU rule for each column. A waiting block management means;
The destage control means estimates that the cache block on the tail side of the destage queue for each column is a cache block corresponding to the log area, and the closer to the tail of the destage queue, the destage target The storage apparatus according to claim 1, wherein the priority is reduced. The destaging control means is configured so that the proportion of dirty cache blocks to be destaged in each column of the write-back cache is notified in advance by the host of the write-back cache in the log area. The storage apparatus according to claim 2, wherein when the ratio is larger than the ratio to the capacity, the destage operation target is selected from the cache block on the head side of the destage queue corresponding to the column. Further comprising destage wait block management means for managing the order of dirty cache blocks to be subjected to the destage operation on the write-back cache by a destage queue updated according to LRU rules;
The destage control means estimates that the cache block on the tail side of the destage queue is a cache block corresponding to the log area, and the closer to the end of the destage queue, the priority as the destage target The storage device according to claim 1, wherein the storage device is lowered. The destage control means, in the write back cache, when the total capacity occupied by the dirty cache block to be destaged is larger than the capacity of the log area notified in advance from the host, The storage apparatus according to claim 4, wherein the destage operation target is selected from a cache block on a head side of the destage queue. Log equivalent area allocating means for allocating an area corresponding to the logical address of the log area on the write-back cache as a log equivalent area based on the logical address of the log area notified in advance from the host; And
The data storage means stores the requested data in the log-corresponding area on the write-back cache or on the write-back cache depending on whether the logical address specified in the write request is within the logical address range of the log area. Store in a cache block outside the log equivalent area,
The destage control means estimates that a cache block included in the log-corresponding area on the write-back cache is a cache block corresponding to the log area, and excludes it from the destage operation target. The storage apparatus according to claim 1. An access frequency management means for managing the access frequency of the block for each cache block of the write-back cache;
The destage control unit estimates a cache block whose access frequency exceeds a certain level as a cache block corresponding to the log area based on the access frequency for each cache block managed by the access frequency management unit. The storage apparatus according to claim 1, wherein the storage apparatus is excluded from the destage operation target. In a storage controller for controlling access to a secondary storage area in which a log area with continuous logical addresses is secured for storing a history of data update from a host in chronological order,
A write-back cache managed in units of cache blocks for temporarily storing data to be written to the secondary storage area;
Data storage means for receiving a write request accompanying the start of the commit process or checkpoint process of the host and storing the requested data in the cache block on the write-back cache;
Destage control means for controlling a destage operation for writing the updated cache block data on the write-back cache to the original write destination on the secondary storage, and the original write in the log area A storage controller comprising: a destage control unit that suppresses a cache block corresponding to a log area in which data on the write-back cache is stored as a target of the destage operation. A secondary storage area in which a log area with continuous logical addresses is secured to store a history of data updates from the host in chronological order, and a cache for temporarily storing data to be written in the secondary storage area A write back cache control method for controlling the write back cache, which is applied to a storage apparatus including a write back cache managed in units of blocks,
Receiving a write request accompanying the start of commit processing of the host and storing the requested data in the cache block on the write-back cache;
Receiving a write request associated with the start of the host checkpoint process and storing the requested data in the cache block on the write-back cache;
A step of controlling a destaging operation for writing the updated cache block data on the write-back cache to the original write destination in the secondary storage area, wherein the log area includes the original write destination A cache block corresponding to a log area in which data on the write-back cache is stored is suppressed from being a target of the destage operation.

Images