JP2019168733A - Information processing system, cache capacity distribution method, storage control apparatus, and method and program thereof - Google Patents

Abstract

To optimize a capacity to be allocated to a partition of a cache corresponding to each processing without measuring and improve an access performance to a storage medium.SOLUTION: A storage control apparatus includes a reduction candidate setting unit that sets a partition having a page which is not accessed over predetermined time or a page in which an access frequency is less than a predetermined value as a partition capable of reducing a capacity thereof, and a capacity control unit that reduces the capacity of the partition capable of reducing the capacity and adds it to the capacity of the partition where cache miss occurs when the page where the cache miss occurs is included in a history of pages evicted from the cache memory to the lower storage medium.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an information processing system, a cache capacity distribution method, a storage control device, a method thereof, and a program.

  In the above technical field, Patent Document 1 discloses a technique for adjusting the capacity to be allocated based on the cache hit frequency. Patent Document 2 discloses a technique for calculating the capacity of each partition when the cache capacity is divided into partitions and handled as an independent cache. In Non-Patent Document 3, newly stored data and once hit data are managed by different LRU caches, and data stored in the cache is accessed when data previously stored in each cache unit is accessed. A technique of moving a certain capacity from the other cache so as to hit at the same time is disclosed.

International Publication WO2014 / 142337 JP 2007-041904 A

H. S. Stone, J. Turek and J. L. Wolf, "Optimal partitioning of cache memory," in IEEE Transactions on Computers, vol. 41, no. 9, pp. 1054-1068, Sep 1992. G. E. Suh, L. Rudolph, and S. Devadas. 2004. Dynamic Partitioning of Shared Cache Memory. J. Supercomput. 28, 1 (April 2004), 7-26. Nimrod Megiddo and Dharmendra S. Modha. 2003. ARC: A Self-Tuning, Low Overhead Replacement Cache.In Proceedings of the 2nd USENIX Conference on File and Storage Technologies (FAST '03) .USENIX Association, Berkeley, CA, USA, 115 -130.

  However, in the technique described in the above-mentioned document, optimization of the capacity allocated to each partition of the cache is performed at the time of distribution with a measurement time such as a cache hit rate, and there is no measurement time when capacity is changed during operation. There is a limit to improving access performance because it cannot cope with the optimization of network access.

  The objective of this invention is providing the technique which solves the above-mentioned subject.

In order to achieve the above object, a storage control device according to the present invention provides:
In the cache memory, a reduction candidate setting means for setting a partition including a page that has not been accessed for a predetermined time or a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity Capacity control means to be added to the capacity of
Is provided.

In order to achieve the above object, a storage control method according to the present invention includes:
In a cache memory, a reduction candidate setting step for setting a partition that includes a page that has not been accessed for a predetermined time or a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity A capacity control step to add to the capacity of
including.

In order to achieve the above object, a storage control program according to the present invention provides:
In a cache memory, a reduction candidate setting step for setting a partition that includes a page that has not been accessed for a predetermined time or a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity A capacity control step to add to the capacity of
Is executed on the computer.

In order to achieve the above object, a system according to the present invention provides:
A storage medium;
Access means for instructing access to the storage medium based on access requests from a plurality of processes;
Storage control means for controlling access to the storage medium of the access means, having a cache memory;
An information processing system comprising:
The storage control means
In the cache memory, a reduction candidate setting means for setting a page that has not been accessed for a predetermined time or a partition including a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which a cache miss has occurred is included in the history of pages that have been evicted from the cache memory to the lower storage medium, the cache miss has occurred by reducing the capacity of the partition capable of reducing the capacity. Capacity control means to add to the capacity of the partition;
Is provided.

In order to achieve the above object, a cache capacity allocation method according to the present invention includes:
An assigning step of assigning a partition including at least one page in the cache memory to each of the plurality of processes;
In the cache memory, a reduction candidate setting step for setting a page that has not been accessed for a predetermined time or a partition including a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity A capacity control step to add to the capacity of
While executing the reduction candidate setting step in the background, the capacity control step is repeated, and a distribution management step for managing the distribution of the cache capacity for each process;
including.

  ADVANTAGE OF THE INVENTION According to this invention, it can optimize during a process, without measuring the capacity | capacitance which should be allocated to the partition of the cache corresponding to each process, and can improve the access performance with respect to a storage medium.

1 is a block diagram illustrating a configuration of a storage control device according to a first embodiment of the present invention. It is a block diagram which shows the structure of the information processing system containing the storage control apparatus which concerns on 2nd Embodiment of this invention. It is a block diagram which shows the function structure of the storage control apparatus which concerns on 2nd Embodiment of this invention. It is a figure which shows the structure of the page state management table which concerns on 2nd Embodiment of this invention. It is a figure which shows the structure of the partition capacity | capacitance setting table which concerns on 2nd Embodiment of this invention. It is a figure which shows the structure of the replacement page log | history table which concerns on 2nd Embodiment of this invention. It is the block diagram which showed the hardware constitutions of the storage control apparatus in 2nd Embodiment of this invention. It is the flowchart which showed the process sequence of the storage control apparatus in 2nd Embodiment of this invention. It is the flowchart which showed the procedure of the access process in 2nd Embodiment of this invention. It is the flowchart which showed the procedure of the access processing at the time of the cache hit in 2nd Embodiment of this invention. It is the flowchart which showed the procedure of the access processing at the time of the cache miss in 2nd Embodiment of this invention. It is the flowchart which showed the procedure of the access processing at the time of the cache miss in 2nd Embodiment of this invention. It is the flowchart which showed the procedure of the setting capacity | capacitance adjustment process of the partition in 2nd Embodiment of this invention. It is the flowchart which showed the procedure of the partition releasable capacity update processing in 2nd Embodiment of this invention. It is a block diagram which shows the function structure of the storage control apparatus which concerns on 3rd Embodiment of this invention. It is a block diagram which shows the function structure of the storage control apparatus which concerns on 4th Embodiment of this invention. It is the flowchart which showed the procedure of the processing speed adjustment process in 4th Embodiment of this invention. It is a block diagram which shows the structure of the information processing system containing the storage control apparatus which concerns on 5th Embodiment of this invention.

  Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings. However, the constituent elements described in the following embodiments are merely examples, and are not intended to limit the technical scope of the present invention only to them. The “page” is a unit for accessing the cache in the system, and the unit for accessing the storage medium is often a unit finer than the page. In this specification, in order to make the explanation easy to understand, it is described that all accesses are performed in units of pages. However, in actuality, accesses are performed in units smaller than pages.

[First Embodiment]
A storage control device 100 as a first embodiment of the present invention will be described with reference to FIG. The storage control device 100 is a device that controls access to a storage medium including a cache memory.

  As illustrated in FIG. 1, the storage control device 100 includes a reduction candidate setting unit 101 and a capacity control unit 102. The reduction candidate setting unit 101 sets, in the cache memory 130, a partition 111 that includes a page that has not been accessed for a predetermined time or a page whose access frequency is lower than a predetermined value, as a partition 121 that can be reduced in capacity. The capacity control unit 102 reduces the capacity of the partition 121 that can reduce the capacity when the page 122 in which the cache miss has occurred is included in the history 123 of the page that has been evicted from the cache memory 130 to the lower storage medium 140. Thus, it is added to the capacity of the partition 124 in which the cache miss has occurred.

  According to the present embodiment, a partition including a page with few accesses is set as a partition capable of reducing the capacity, and the capacity is moved to a partition in which a cache miss has occurred in the evicted page. Thereby, it is possible to optimize during processing without measuring the capacity to be allocated to the cache partition corresponding to each processing, and to improve the access performance to the storage medium.

[Second Embodiment]
Next, a storage controller that controls the cache memory according to the second embodiment of the present invention will be described. The storage control apparatus of this embodiment sets a partition including a page with few accesses to a partition capable of reducing the capacity, and repeatedly moves the capacity to the partition where a cache miss has occurred in the evicted page.

  Here, with respect to the setting of partitions that can be reduced, the capacity of partitions containing unnecessary pages is reduced based on unnecessary page detection that detects pages that are not accessed as unnecessary pages in a specified time period. Set as possible partition. This unnecessary page detection is performed using, for example, a clock algorithm. In addition, when there are a plurality of partitions whose capacity can be reduced in capacity transfer, a partition including a page most likely to be evicted to a lower storage medium is set as a partition whose capacity is reduced. Then, in the partition whose capacity is to be reduced, when data storage is newly required in the partition to which the capacity is moved, the page that is most likely to be evicted to the lower storage medium is reduced. In addition, the cache capacity distribution method according to the present embodiment performs an allocation process for allocating a partition including at least one page in the cache memory and a reduction candidate setting process in the background for each of a plurality of processes. And a distribution management process for managing the distribution of the cache capacity allocated to each process by repeating the capacity control process.

《Explanation of prerequisite technology》
In recent years, the capacity of data processed in a computer has increased with the improvement of the processing capability of the computer and the increase in the density of storage elements and storage media. Along with this, the amount of access to lower storage hierarchies composed of low-speed storage media having a larger capacity in computer processing is increasing. For example, if the entire data used in the process does not fit in the memory of the computer, the data is stored in a storage medium such as HDD (Hard Disk Drive) or SSD (Solid State Drive) using NAND flash. Do. In many computers, a processor does not have a large storage area, and uses a lower-order memory composed of a DRAM (Dynamic Random Access Memory) or the like in order to process a large amount of data. Since the lower storage hierarchy is usually slow in a computer, processing performance is improved by caching data in an upper storage medium. In order to obtain a sufficient cache effect, a cache having a sufficient capacity corresponding to the increased access amount is required.

  On the other hand, in processors adopted in recent computers, the number of instruction execution units called cores and threads is increasing in order to execute a plurality of processes in parallel for the purpose of improving processing capability. In order to improve the performance by utilizing the processing capability of these processors, it is required to increase the number of processes executed in parallel on each computer. The plurality of processes executed on the computer usually execute processes with different procedures for different data. Therefore, the cache capacity that can be used by each process executed in parallel decreases in inverse proportion to the number of processes.

  That is, it is necessary to allocate the capacity so as to maximize the cache effect of the entire processes executed in parallel without allocating an extra cache capacity for each process. As the number of processes executed in parallel increases, it is difficult to allocate a capacity with a margin for the amount of data used in each process so that sufficient performance can be obtained. In addition, an increase in the number of parallel executions increases the possibility of contention between data cached by each process. For example, when one of the processes refers to a large amount of data, data (working set) that is frequently referred to by other processes executed in parallel is expelled by consuming a large amount of cache capacity. Since the data included in the working set is mutually purged between the plurality of processes, the access to the lower storage hierarchy is increased in each process, leading to a significant performance degradation.

  To deal with this problem, for example, in the technique disclosed in Non-Patent Document 1, a cache is divided into areas called partitions for each process to prevent contention between processes. Each partition operates as an independent cache of the allocated capacity. Therefore, by using partitions, it is possible to avoid contention between processes in which cached data drives out data of other processes. Moreover, the number of hits for the cache capacity is improved by setting an optimum partition capacity for each process. The optimum partition capacity of each process included in the cache depends on not only the memory capacity of the computer or storage system but also the combination of processes executed in parallel. The working set changes as the execution of each process proceeds.

  Therefore, for example, in the technique disclosed in Non-Patent Document 2, the capacity distribution of each partition is dynamically optimized based on the access status in the processor cache. The number of hits when the partition capacity changes is predicted from the cache usage status, and the capacity allocation that maximizes the number of hits for the partition is calculated. In the technique disclosed in Non-Patent Document 2, in order to predict the number of hits when the capacity changes, the priority of replacement in the partition when each access hits is referred to. However, in an environment having a large capacity and containing a large number of data, such as a cache for a storage medium such as an HDD or SSD, it is difficult to check the priority of replacement for an access hit in each partition. For example, when LRU (Least Recently Used) is used as a cache replacement algorithm, since the data structure representing the priority is in the form of a list in many cases, it is necessary to check the replacement priority of the data hit by the cache from the top. You need to follow the list. Therefore, the response time at the time of a cache hit increases, leading to a decrease in access performance.

  Non-Patent Document 3 discloses a cache replacement algorithm that dynamically adjusts the capacity of each part of a cache according to access. In this algorithm, newly stored data and once hit data are managed by different LRU caches, and when data previously stored in each cache unit is accessed, the data hits in the cache To move a certain capacity from the other cache. Therefore, it is not necessary to predict the number of hits when the capacity changes as in the technique disclosed in Non-Patent Document 2. However, it does not support capacity adjustment between three or more caches, and it is not assumed that the capacity between caches or partitions assigned to different processes is optimized.

  The operation itself for detecting a page that has not been accessed for a long period of time is used in a general technique for determining data to be replaced in cache and memory management. Usually, in order to directly detect a detected page as a target for replacement. Used.

(Task)
In order to achieve high performance in multiple processes executed in parallel, it is possible to allocate a capacity so as to maximize the effect of the cache for each process while preventing cache contention between processes in the shared cache. Desired. Therefore, it is necessary to obtain an optimum partition capacity based on the working set of each process. In order to dynamically measure the working set of each process, it is necessary to monitor the data access for each corresponding partition and check the total capacity of the data that is repeatedly accessed.

  However, in general, it takes time to examine the working set of each process and find out the result. It can be said that the higher the access frequency of each data included in the working set, the higher the possibility that access will be detected earlier from the start of measurement. Conversely, data that is relatively infrequently accessed in the working set is likely to take time from the start of measurement until access to the data is detected, and as a result, to measure the entire working set without omission It takes a lot of time.

  In other words, it is a problem that it takes time to determine the capacity allocated to the cache partition of each process. For example, when the execution of a new process is started or ended on the computer, the optimum capacity distribution for each process partition changes. In many cases, the working set of each process changes with time. However, since it takes a lot of time to determine the working set, it is difficult to cope with these capacity distribution changes. In particular, in order to calculate the surplus capacity in each partition, it is necessary to detect data out of the working set. For this reason, for example, it is not possible to deal with a case where the capacity of the partition needs to be immediately added as in a process that has been newly started, and the performance is greatly reduced.

  That is, in the base technology, even when the capacity of the partition allocated to each operation during the cache operation is changed, a process for detecting the occurrence of a cache miss and requesting an additional capacity, and a capacity with less access Is associated with the process of finding partitions that can be reduced. For this reason, the time for changing the capacity is limited, and the data access performance is degraded.

<< Description of this embodiment >>
In this embodiment, in a plurality of processes executed in parallel, a method is provided for dynamically adjusting the capacity of each partition so that the cache hit rate of the cache partition capacity of each process on the storage medium is maximized. To do. For this purpose, it is determined whether or not the capacity of the partition can be reduced on the basis of access information to the storage medium through the cache by the process assigned to each partition. Then, when the capacity of each partition needs to be added, the capacity is moved from the partition capable of capacity reduction prepared in advance to the partition determined to require the capacity addition.

  In the present embodiment, as a criterion for determining whether or not the capacity of the partition can be reduced, data stored in the cache but out of the working set is detected. In addition, from the information on access to the storage medium through the cache partition corresponding to each process, as a criterion for adding the capacity of the partition, data that has been repeatedly accessed but detected due to insufficient cache capacity is detected. To do.

  First, in this embodiment, the presence or absence of access to each page included in the cache is monitored to detect a page that has not been accessed for a long time. A page that has not been accessed for a long period of time is likely to have a low access frequency and is likely to be a page that is not included in the working set. Therefore, the capacity of the detected page is set to a capacity that can be released from the partition including the page.

  Next, in the present embodiment, the history of the most recent page evicted from the cache is stored, and collated with the history stored when a cache miss occurs in each partition. When the cache missed page is included in the history, it is considered that the cache missed page was repeatedly accessed but was missed because it was evicted from the cache. For this reason, a request is made to add a capacity of a partition necessary for storing in the cache a page that has been missed by the accessed process.

  Then, the requested capacity is additionally allocated to the partition for which the capacity addition request is made. On the other hand, one partition is selected from the partitions whose capacity can be released, and the capacity required to be added is subtracted from the capacity allocated to the selected partition.

  In each partition whose capacity has increased or decreased, page replacement is performed based on the capacity after the increase or decrease. When storing new data due to a cache miss, if the number of pages included in the partition where the data is stored is less than the capacity, replace the page of the other partition with more pages than the capacity, and change the partition Increase the number of pages. By repeating the above processing, the capacity of each partition is dynamically adjusted so as to maximize the cache hit rate.

  This embodiment includes a page status management function, a partition capacity setting function, a replacement target page selection function, a replacement page history management function, an unnecessary page capacity release function, a capacity addition determination function, and an additional capacity acquisition source selection function. And comprising.

  The page state management function manages information about each page in the cache. In other words, in order to maximize the cache hit rate by determining the capacity reduction and capacity addition for the cache partition corresponding to each process and moving the capacity between partitions, the last access time for each page, or Manages the presence / absence and number of accesses for a certain period of time. The information for managing each page includes the last access to each page in addition to the general cache management information such as identifiers such as storage destination addresses in the storage medium of the data stored in each page and the priority of replacement. It has at least one information regarding time, presence / absence of access within a certain time, and number of accesses.

  The partition capacity setting function can set the capacity of each partition during operation, and manages designation of the capacity allocated to each partition defined in the cache and the releasable capacity.

  The replacement target page selection function determines the page to be replaced for storing data, or controls the priority order of pages to be replaced. That is, for each partition that operates as a virtual independent cache defined in the cache, the cache is based on the priority for replacing each page in the partition and the partition capacity obtained from the partition capacity setting function. A page to store new data is determined by mistake. Further, the priority order for replacement of each page of the partition in which new data is stored is updated based on the replacement result.

  The replacement page history management function manages the history of pages that have been stored in the cache in the past and replaced by other data. That is, for pages replaced by data newly stored in a certain past period in each partition, information on identifiers that specify storage destinations such as addresses in the storage medium of data stored in the replaced pages is managed. To do.

  The unnecessary page capacity release function detects an unnecessary page that has not been accessed for a certain period of time from the access information corresponding to each page, or detects an infrequently accessed unnecessary page, and the capacity allocated to the partition to which the unnecessary page belongs Calculate the releasable capacity. That is, the information of each page managed by the page state management function is circulated, and access is not performed for a certain period of time from the last access time, the presence / absence of access within a certain period of time, the number of accesses, or the access frequency is low Detect unnecessary pages. Then, for the partition including the unnecessary page, the capacity of the unnecessary page is added to the releasable capacity managed by the partition capacity setting function.

  The capacity addition determination function determines whether or not capacity addition to a partition is necessary from the history of pages replaced in the cache and cache miss information of processing corresponding to each partition. That is, for an access to a storage medium that has caused a cache miss, an identifier such as an access destination address is compared with an identifier managed by the replacement page history management function. If there is an identifier that matches the access destination, it is determined that the capacity of the partition needs to be increased to increase the cache hit rate, the partition corresponding to the process that requested the access is specified, and the capacity for the specified partition Make additional requests.

  The additional capacity acquisition source selection function refers to the history information maintained by the history management function, and selects a partition having a releasable capacity as a capacity acquisition source and a capacity acquisition source. That is, in response to a capacity addition request made by the capacity addition determination function, a partition that provides an additional capacity to satisfy the request is selected. The partition selection condition is a partition having a releasable capacity that is larger than the capacity requested for addition in the partition capacity setting function. When there are a plurality of partitions that satisfy the selection condition, for example, an arbitrary partition may be selected at random, or an optimal partition may be selected by any other algorithm.

《Information processing system》
FIG. 2 is a block diagram showing a configuration of an information processing system 200 including a storage control unit 220 as a storage control device according to the present embodiment.

  Referring to FIG. 2, the information processing system 200 includes one or more host computers 210, a storage medium 240, and a network 250 that connects the host computer 210 and the storage medium 240. Note that the storage medium 240 may be distributed over a plurality of storage media and connected to the network 250.

  The host computer 210 includes an OS (Operating System) 211, a plurality of cores operating on the OS 211 and a thread 212 that is software, and the storage control unit 220 of the present embodiment. The storage control unit 220 includes a cache 221 and controls access processing including reading and writing to the storage medium 240 in the processing of the OS 211 and the plurality of cores and threads 212.

  The storage medium 240 is a write-once storage medium such as an HDD employing SSD or SMR (Shingled Magnetic Recording) using a NAND flash as a storage element. Note that the storage medium 240 may include a medium body 241 and a buffer. Further, the form of the network 250 is not particularly limited, and the host computer 210 and the storage medium 240 may be directly connected. The host computer 210 may be configured by a plurality of devices or systems connected by the network 250.

<< Functional configuration of storage controller >>
FIG. 3 is a block diagram showing a functional configuration of the storage control unit 220 as the storage control device according to the present embodiment. In FIG. 3, the same components as those in FIG. 2 are denoted by the same reference numerals.

  Referring to FIG. 3, the storage control unit 220 includes a cache 221 as a page pool, an access control unit 301, a cache hit / miss determination unit 302, a page state management unit 308 having a page state management table 381, and a partition. A partition capacity management unit 307 having a capacity setting table 371. The storage control unit 220 also includes a replacement target page selection unit 303, a replacement history management unit 304 having a replacement page history table 341, a capacity addition determination unit 305, an additional capacity acquisition source selection unit 306, and an unnecessary page capacity release. Unit 309.

  In FIG. 3, a cache hit / miss determination unit 302, a page state management unit 308, a partition capacity management unit 307, a replacement target page selection unit 303, and an unnecessary page capacity release unit 309 are included in a reduction candidate setting unit. Function as. In addition, the cache hit / miss determination unit 302, the page state management unit 308, the partition capacity management unit 307, the replacement target page selection unit 303, the replacement history management unit 304, the capacity addition determination unit 305, and the additional capacity acquisition The original selection unit 306 functions as a capacity control unit.

"cache"
A cache 221 serving as a page pool stores cached data in units of pages for access to the storage medium 240 of the core or thread of the host computer 210.

《Access control unit》
The access control unit 301 receives an access request from the software of the host computer 210, and performs necessary access processing for the storage medium 240 and a response to the result of the access request. In addition, the access control unit 301 accesses a specified page of the cache 221 instead of accessing the storage medium 240 according to an instruction from the replacement target page selection unit 303, or stores data of a page included in the cache 221 as the storage medium 240. Write to.

<< Cash hit / miss judgment part >>
The cache hit / miss determination unit 302 receives information on the access destination address of the storage medium 240 in the access request received by the access control unit 301, refers to the page state management table 381, and the access destination data is in the cache 221. It is determined whether or not. Further, the cache hit / miss determination unit 302 notifies the replacement target page selection unit 303 of the determination result of cache hit or cache miss.

《Page status management unit》
The page state management unit 308 uses the page state management table 381 to obtain management information necessary for realizing the cache function for the storage medium 240 and information necessary for adjusting the capacity of the partition for each page included in the cache 221. to manage.

(Page status management table)
FIG. 4 is a diagram showing a configuration of the page state management table 381 according to the present embodiment.

  The entry of each page of the page state management table 381 includes a use flag 401 indicating that the page corresponding to the entry is used, an address 402 indicating a storage location of the data of the page in the storage medium 240, and the page Partition number 403 indicating the partition to which the file belongs. Each page entry has a replacement priority 404 indicating the priority of replacement in each partition, and an access flag 405 indicating that each page has a cache hit. Further, each page entry indicates that each page is updated from data recorded in the storage destination of the storage medium 240 and an unnecessary candidate flag 406 indicating that each page is determined to be a cache-unnecessary page. An update flag 407.

  The format of data stored in each entry of the page state management table 381 is not limited to the format shown in FIG. 4. For example, instead of having the use flag 401, the value of the address 402 is stored in the storage area of the storage medium 240. It may be substituted by setting to a specific value that does not correspond to. Further, the replacement priority order 404 may be expressed in the form of a link list when, for example, LRU (Least Recently Used) is used as a replacement algorithm for pages in a partition instead of having numerical order information. In addition, for example, when the unnecessary page capacity releasing unit 309 determines whether the page stores unnecessary data using the information on the access time of each page, the last access time is used instead of the access flag 405 for each entry. You may make it have the information of.

<Partition capacity management section>
The partition capacity management unit 307 uses the partition capacity setting table 371 to operate each partition as an independent cache defined in the cache, and manages information on the capacity set in the partition.

(Partition capacity setting table)
FIG. 5 is a diagram showing a configuration of the partition capacity setting table 371 according to the present embodiment.

  Each entry of the partition capacity setting table 371 corresponds to a defined individual partition, a partition number 501 for identifying the partition, and a set capacity 502 indicating the number of pages allocated to the partition corresponding to the entry. Have. Further, each entry of the partition capacity setting table 371 has a releasable capacity 503 indicating a capacity releasable from other partitions in the partition, and a real capacity 504 indicating the number of pages actually included in the partition.

  The format of the data stored in each entry of the partition capacity setting table 371 is not limited to the format shown in FIG. 5. For example, instead of having the partition number 501, the order of entries counted from the top is used as the partition number. May be used. In addition, the set capacity 502, the releasable capacity 503, and the real capacity 504 may be indicated by a capacity in bytes instead of the number of pages. Further, in the case where a plurality of requests are processed together without processing the capacity addition request for each partition in the capacity addition determination unit 305 once, the number of capacity addition requests is set in the entry corresponding to each partition. It may be stored.

《Replacement target page selection part》
The replacement target page selection unit 303 instructs the access control unit 301 to access the cache 221 or the storage medium 240 based on the cache hit / cache miss determination result by the cache hit / miss determination unit 302. Then, the replacement target page selection unit 303 updates the replacement priority order 404 of each page stored in the page state management table 381 and the address information corresponding to the data. Further, the replacement target page selection unit 303 instructs the access control unit 301 to access and respond to the page corresponding to the address specified in the access request.

  Further, when a cache miss occurs, the replacement target page selection unit 303 refers to the set capacity 502 of the partition capacity setting table 371 and determines a partition that provides a page to be replaced in order to store new data. To do. In addition, the replacement target page selection unit 303 adds the address 402 of data already stored in the page to be replaced to the replacement page history table 341 of the replacement history management unit 304. Then, when the update flag 407 of the page to be replaced is set, the replacement target page selection unit 303 sets the storage target in the storage medium 240 before replacing the data stored in the page with new data. The access control unit 301 is instructed to write.

《Replacement history management unit》
Using the replacement page history table 341, the replacement history management unit 304 uses the replacement page history table 341 to store the storage medium 240 regarding the data stored in the page with respect to the page replaced by the data newly stored in the past certain period. To manage the storage destination address. Further, the replacement history management unit 304 deletes an address that has passed a certain period from the page address stored in the replacement page history table 341.

(Replacement page history table)
FIG. 6 is a diagram showing a configuration of the replacement page history table 341 according to the present embodiment.

  The replacement page history table 341 has replacement histories 610 to 630 that are the history of pages replaced for each partition, and the replacement histories 610 to 630 each include replacement addresses 611 to 631 representing individual replaced pages. Note that the number of replacement addresses 611 to 631 stored in each of the replacement histories 610 to 630 is a factor that determines whether or not the capacity of the partition is frequently changed. For example, as the number of replacement addresses 611 to 631 increases, the number of times that satisfies the condition for adding the capacity of the partition at the time of a cache miss increases.

  Note that the format of data stored in each entry of the replacement page history table 341 is not limited to the format shown in FIG. For example, instead of storing replacement addresses 611 to 631, the replacement histories 610 to 630 may use a data format for determining whether a specific address is included in the set of replacement histories 610 to 630, such as a Bloom filter. Good. In this case, the replacement history 610 to 630 is formed from a plurality of Bloom filters, and a new Bloom filter and an old Bloom filter are replaced every time a certain period elapses, thereby determining a page replaced within a past certain period. It is possible.

《Capacity addition judgment unit》
Is the capacity addition determination unit 305 included in the replacement history 610 to 630 of the partition corresponding to the accessed process in which the storage destination address is stored in the replacement page history table 341 for the cache missed access? judge. The capacity addition determination unit 305 determines that the capacity needs to be added to the partition when the storage destination address is included in the replacement histories 610 to 630 and no capacity addition request is made. A capacity addition source selection unit 306 is instructed to add capacity for one page.

  In this embodiment, addition of a capacity for one page is instructed, but a capacity for a plurality of pages may be added. As the number of pages to be added increases, the capacity change of one partition increases and the capacity optimization progresses quickly, but conversely, the capacity change tends to become severe and unstable, the cache memory capacity, partition A suitable number may be set in consideration of conditions such as capacity, access frequency, and the like.

<< Additional capacity acquisition source selection part >>
The additional capacity acquisition source selection unit 306 receives the capacity addition instruction for the partition from the capacity addition determination unit 305, refers to the releasable capacity 503 for each partition in the partition capacity setting table 371, and selects a partition having a releasable capacity. Extract. Then, a partition that provides capacity in response to the capacity addition instruction is selected from the extracted partitions. The additional capacity acquisition source selection unit 306 subtracts one page from the releasable capacity 503 of the partition capacity setting table 371 for the selected partition, and sets the set capacity of the partition capacity setting table 371 for the target partition for the capacity addition instruction. 502 is added for one page.

  When the capacity addition determination unit 305 makes a request for adding a plurality of pages, the additional capacity acquisition source selection unit 306 performs subtraction from a plurality of pages or a plurality of partitions.

<< Unnecessary page capacity release part >>
The unnecessary page capacity releasing unit 309 circulates each entry in the page state management table 381 and confirms the access flag 405 and the unnecessary candidate flag 406. When the access flag 405 is set when the entry is visited, it is determined that the page is included in the working set in order to indicate that the corresponding page has been accessed after the previous entry was visited. Then, the access flag 405 is reset. On the other hand, when the access flag 405 is not set at the time of entry circulation and the unnecessary candidate flag 406 is not set, the page is not accessed for a certain period of time, that is, a page not included in the working set. And the unnecessary candidate flag 406 is set. Further, for a partition including the page, one page is added to the releasable capacity 503 of the partition capacity setting table 371.

  In the present embodiment, for example, unnecessary page capacity is detected using a clock algorithm, but other algorithms may be used.

<< Hardware configuration of storage controller >>
FIG. 7 is a block diagram showing a hardware configuration of the storage control unit 220 as the storage control device in the present embodiment. In FIG. 7, the storage control unit 220 is illustrated as software processing of the host computer 210, but the storage control unit 220 may be realized by a one-chip computer independent of the host computer 210.

  In FIG. 7, a CPU (Central Processing Unit) 710 is a processor for arithmetic control, and implements a functional component of the storage controller 220 of FIG. 3 by executing a program. A ROM (Read Only Memory) 720 stores fixed data and programs such as initial data and programs. The cache 221 holds data that is frequently accessed when accessing the external storage medium 240. In the present embodiment, an area is allocated to each process in units of partitions including at least one page. The network interface 730 communicates with the storage medium 240 and other devices via the network 250. Note that the number of CPUs 710 is not limited to one, and may be a plurality of CPUs (cores). The network interface 730 preferably includes a CPU independent of the CPU 710 and writes or reads transmission / reception data in a RAM (Random Access Memory) 740 area. It is desirable to provide a DMAC (Direct Memory Access Controller) that transfers data between the RAM 740 and the storage 750 (not shown). Further, the input / output interface 760 preferably has a CPU independent of the CPU 710 and writes or reads input / output data to / from the area of the RAM 740. Therefore, the CPU 710 recognizes that the data has been received or transferred to the RAM 740 and processes the data. Further, the CPU 710 prepares the processing result in the RAM 740 and leaves the subsequent transmission or transfer to the network interface 730, the DMAC, or the input / output interface 760.

  The RAM 740 is a random access memory that the CPU 710 uses as a work area for temporary storage. In the RAM 740, an area for storing data necessary for realizing the present embodiment is secured. The access address 741 is an address accessed from the OS 211 or the core or thread 212 to the storage medium 240. The cache hit address 742 is an address that has a cache hit among the access addresses 741. The cache miss address 743 is an address in the access address 741 that has a cache miss. The capacity addition partition 744 is a partition that requests addition of capacity when the cache missed address matches the replaced address (extruded to the storage medium). The capacity reduction partition 745 is a partition that includes a page with no access or low frequency and is a candidate for reduction, and is a partition selected to move the capacity to the capacity addition partition 744. The access information 746 is access (including Read and Write) contents and data to the storage medium 240. Input / output data 747 is data input / output via the input / output interface 760. The transmission / reception data 748 is data transmitted / received to / from the storage medium 240 via the network interface 730. The application use area 749 is an area used by the application in processing other than storage control.

  The storage 750 stores a database, various parameters, or the following data or programs necessary for realizing the present embodiment. The page state management table 381 is a table for managing each page of the cache 221 shown in FIG. The partition capacity setting table 371 is a table for managing the partition of the cache 221 allocated to each process shown in FIG. The replacement page history table 341 is a table for storing, for each partition, the history of page addresses pushed out from the cache 221 to the storage medium 240 shown in FIG. The released capacity calculation algorithm 751 is an algorithm for selecting a partition that can reduce the capacity by detecting a page in the cache that is not used or used less frequently, and includes a clock algorithm in this example. The page cyclic cycle (cyclic rate) 752 stores a cyclic cycle based on a clock algorithm.

  The storage 750 stores the following programs. The OS 211 is a basic program that controls the entire host computer 210. The plurality of threads 212 are processing units in a program that is currently being executed by the host computer 210. The storage control program 753 is a program that realizes access control of the present embodiment in response to access to the storage medium 240 from the OS 211 or the plurality of threads 212. The access processing module 754 is a module that processes access to the storage medium 240 from the OS 211 and the plurality of threads 212, and includes a cache hit control module 755 that performs control when a cache hit occurs, and a cache miss that performs control when a cache miss occurs. A control module 756. The set capacity control module 757 is a module that controls the movement of the set capacity from the partition whose capacity can be reduced to the partition that requests capacity addition. The free capacity calculation module 758 is a module for selecting a partition whose capacity can be reduced by detecting a page in the cache that is not used or used infrequently according to the free capacity calculation algorithm 751.

  The input / output interface 760 interfaces input / output data with input / output devices. Although the display unit 761 and the operation unit 762 are connected to the input / output interface 760, the present invention is not limited to this.

  Note that the RAM 740 and the storage 750 in FIG. 7 do not show programs and data related to general-purpose functions and other realizable functions that the host computer 210 has.

<< Processing procedure of storage controller >>
FIG. 8 is a flowchart showing a processing procedure of the storage control unit 220 as the storage control device in the present embodiment. This flowchart is executed by the CPU (core) 710 in FIG. 7 using the RAM 740, and implements the functional configuration unit in FIG.

  In step S801, the storage control unit 220 performs initialization processing including processing of each table. In step S <b> 803, the storage control unit 220 starts an access process for controlling access to the storage medium 240 in the host computer 210. In step S805, the storage control unit 220 starts the set capacity control process for reducing and adding the set capacity of the partition allocated to the cache 221 corresponding to each process. In step S807, the storage control unit 220 starts a released capacity calculation process for calculating a capacity (number of pages) that can be released from the partition allocated to the cache 221 corresponding to each process.

  In step S809, the storage control unit 220 determines whether the storage control process is finished. If it is determined that the storage control process is finished, the process is finished. In FIG. 8, it has been described that the access process, the set capacity control process, and the released capacity calculation process are performed independently and in parallel, but they may be executed in association with each other.

(Access processing)
FIG. 9 is a flowchart showing a procedure of access processing in the present embodiment. This flowchart is a procedure for processing an access request from the software of the host computer 210.

  When the access control unit 301 of the storage control unit 220 receives an access request from the software of the host computer 210 in step S <b> 901, the access control unit 301 notifies the cache hit / miss determination unit 302 of access request information. In step S903, the cache hit / miss determination unit 302 of the storage control unit 220 refers to the page state management table 381 and checks whether there is a page whose address 402 matches the storage destination of the access request storage medium 240. In step S905, the cache hit / miss determination unit 302 branches depending on whether a cache hit or a cache miss occurs.

  When there is a page that stores data that matches the storage location of the access request storage medium 240 (Yes in step S905), in step S907, the cache hit / miss determination unit 302 determines the access request address and the access Is transmitted to the replacement target page selection unit 303 as to whether it is Read or Write. In step S907, the replacement target page selection unit 303 of the storage control unit 220 executes processing at the time of a cache hit, and executes processing from step S911 onward after the execution is completed.

  If there is no page storing data matching the storage location of the access request storage medium 240 (No in step S905), in step S909, the cache hit / miss determination unit 302 determines the access request address and the access Is transmitted to the replacement target page selection unit 303 as to whether it is Read or Write. In step S909, the replacement target page selection unit 303 executes processing at the time of a cache miss.

  When the access to the storage medium 240 and / or the cache 221 instructed from the replacement target page selection unit 303 in step S907 or step S909 is completed, the access control unit 301 in the host computer 210 in step S911. The result is notified to the software that made the access request, and the processing of the access request is completed.

(Access processing at the time of cache hit)
FIG. 10 is a flowchart showing the procedure of the access process (S907) at the time of cache hit in this embodiment. In this embodiment, LRU is assumed as the cache replacement algorithm, but other replacement algorithms may be used.

  In step S1001, the replacement target page selection unit 303 of the storage control unit 220 compares the cache hit address with the address 402 of the page state management table 381 to identify the page corresponding to the cache hit address. In step S1003, the replacement target page selection unit 303 instructs the access control unit 301 to execute an access request for the page of the cache 221 identified in step S1001.

  In step S1005, the replacement target page selection unit 303 updates the replacement priority 404 of the cache hit page in the page state management table 381 to the lowest value. Further, the replacement target page selection unit 303 sets the access flag 405 corresponding to the page and resets the unnecessary candidate flag 406 in step S1005. In step S1007, the replacement target page selection unit 303 increments the replacement priority 404 of the cache hit page in the page state management table 381 by 1 for each of the other pages included in the same partition as the cache hit page.

  In step S1009, the replacement target page selection unit 303 determines whether the cache hit access is Read or Write. If the cache hit access is Read (Yes in step S1009), the replacement target page selection unit 303 resets the cache hit page update flag 407 of the page state management table 381 and ends the processing in step 1011. To do. When the cache hit access is Write (No determination in step S1009), the replacement target page selection unit 303 sets the cache hit page update flag 407 in the page state management table 381 in step S1013 and ends the process. To do.

  The above processing from step S1001 to S1013 is performed, and in step S907 in FIG. 9, the operation for the replacement target page selection unit 303 of the storage control unit 220 to perform processing at the time of a cache hit is completed. The processing order of steps S1001 to S1013 is not necessarily the order shown in FIG. For example, the processing of steps S1003 to S1013 can be executed at the same time or executed with the order changed. In step S 1001, the process for specifying the page corresponding to the address hit by the cache hit compared to the address 402 of the page state management table 381 is specified when the cache hit / miss determination unit 302 determines the cache hit. It may be a process of directly receiving the processed page information. Further, the processing for the replacement priority 404 in step S1005 and step S1007 is such that, when the replacement priority 404 is represented in a linked list format, the position of the cache hit page in the linked list is the lowest in the replacement priority. It is realized by the movement operation.

(Access processing at the time of cache miss)
FIG. 11A and FIG. 11B are flowcharts showing the procedure of the access processing (S909) at the time of a cache miss in this embodiment.

  In step S1101, the replacement target page selection unit 303 of the storage control unit 220 accesses the partition capacity setting table 371 and sets the partition setting capacity 502 and the actual capacity 504 corresponding to the process that made the access request in which the cache miss occurred. Get and compare. The replacement target page selection unit 303 branches in accordance with the comparison result in step S1103.

  As a result of the comparison, if the actual capacity 504 of the partition is less than the set capacity 502 (Yes in step S1103), the replacement target page selection unit 303 accesses the partition capacity setting table 371 in step S1105, and the actual capacity 504 is set. Partitions exceeding capacity 502 are extracted. In step S1107, the replacement target page selection unit 303 selects one arbitrary partition that provides a page necessary for storing new data associated with a cache miss from the partitions extracted in step S1105. In step S1109, the replacement target page selection unit 303 refers to the page state management table 381 and selects a page having the highest replacement priority 404 among the pages included in the partition selected in step S1107 as a cache miss. It selects as a storage location of the accompanying new data, and executes the processing after step S1113.

  As a result of the comparison in step 1101, if the actual capacity 504 of the partition is greater than or equal to the set capacity 502 (No determination in step S1103), the replacement target page selection unit 303 refers to the page state management table 381 in step S1111. In step S1111, the replacement target page selection unit 303 sets the page having the highest replacement priority 404 among the pages included in the partition corresponding to the process in which the cache miss has occurred as the storage destination of new data associated with the cache miss. select.

  In step S1113, the replacement target page selection unit 303 refers to the access flag 401 and the update flag 407 of the page state management table 381 for the page selected as the storage destination of new data due to a cache miss to the access control unit 301. Then, it is confirmed whether the data of the selected page should be reflected on the storage medium 240. In step S1115, the replacement target page selection unit 303 branches depending on whether or not it should be reflected in the storage medium 240. When at least one of the use flag 401 and the update flag 407 is not set for the page selected as the storage destination of the new data (No determination in step S1115), the replacement target page selection unit 303 performs the processing after step S1119 in FIG. 11B. Execute.

  On the other hand, when both the use flag 401 and the update flag 407 are set for the page selected as the storage destination of the new data (Yes determination in step S1115), the replacement target page selection unit 303 determines the page status in step S1117. The address 402 corresponding to the cache missed page in the management table 381 is acquired. Then, the access control unit 301 is instructed to write the data currently stored in the cache missed page into the area indicated by the address 402 of the storage medium 240.

  In step S1119, the replacement target page selection unit 303 determines whether the access request with a cache miss is Read or Write. When the access request from the software of the host computer 210 is Read (Yes in step S1119), the replacement target page selection unit 303 requests the storage medium 240 to read the access control unit 301 in step S1121. Perform access. Then, an instruction is given to store the read result in the page of the cache 221 selected in step S1109 or step S1111. In step S1123, the replacement target page selection unit 303 resets the update flag 407 of the page state management table 381 for the page in which the result is stored in step S1121, and executes step S1129 and the subsequent steps.

  When the access request from the software of the host computer 210 is Write (No determination in step S1119), the replacement target page selection unit 303 sends the data requested to be written to the access control unit 301 in step S1125. An instruction is given to store in the page of the cache 221 selected in step S1109 or step S1111. In step S1127, the replacement target page selection unit 303 sets the update flag 407 of the cache hit page in the page state management table 381 for the page in which the write result is stored in step S1125.

  In step S1129, the replacement target page selection unit 303 sets the partition number 403 of the page state management table 381 for the page selected in step S1109 or step S1111, and the partition corresponding to the process that performed the access request in which the cache miss occurred. The replacement priority 404 is updated to the lowest value. In step S1131, the replacement target page selection unit 303 sets the access flag 405 of the page status management table 381 and resets the unnecessary candidate flag 406 for the page selected in step S1109 or step S1111. In step S1133, the replacement target page selection unit 303 sets the replacement priority 404 of the page state management table 381 to 1 for each of the other pages included in the partition corresponding to the process that performed the access request in which the cache miss occurred. increase.

  In step S1135, the replacement target page selection unit 303 notifies the capacity addition determination unit 305 of the address of the data stored immediately before in the page where the new data is stored, and replaces the corresponding partition replacement history 610-610. In 630, new replacement addresses 611 to 631 are set.

  In step S1137, the replacement target page selection unit 303 branches based on the comparison result in step S1101. If the actual capacity 504 of the partition is greater than or equal to the set capacity 502 (Yes determination in step S1137), the process ends. On the other hand, if the actual capacity 504 of the partition is less than the set capacity 502 (No determination in step S1137), the replacement target page selection unit 303 accesses the partition capacity setting table 371 in step S1139. Next, 1 is added to the actual capacity 504 of the partition corresponding to the process that performed the access request in which the cache miss occurred. Then, 1 is subtracted from the actual capacity 504 of the partition in which the page selected in step S1109 or step S1111 was included, and the process ends.

  The above processing from step S1101 to S1139 is performed, and in step S909 in FIG. 9, the operation for the replacement target page selection unit 303 to perform processing at the time of cache miss is completed. The processing order of steps S1101 to S1139 is not necessarily the order shown in FIGS. 11A and 11B. For example, the processing of steps S1129 to S1139 can be executed at the same time, or can be executed with the order changed. Further, in step S1107, an unnecessary candidate flag 406 for a partition that provides a page necessary for storing new data due to a cache miss, for example, a page having the highest replacement priority 404 of each partition with reference to the page state management table 381. Check out. Then, a partition in which the unnecessary candidate flag 406 is set may be preferentially selected.

(Set capacity adjustment processing)
FIG. 12 is a flowchart showing the procedure of the set capacity adjustment processing of the partition in this embodiment. In this flowchart, the capacity addition determination unit 305 determines the necessity of adding a partition capacity, the additional capacity acquisition source selection unit 306 determines a partition that provides capacity to the partition that needs capacity addition, and the partition Adjust the set capacity.

  The capacity addition determination unit 305 of the storage control unit 220 receives the address of the access request that caused the cache miss from the replacement target page selection unit 303. In step S1201, the replacement history 610 to 630 of the partition corresponding to the process that made the access request in the replacement page history table 341 of the replacement history management unit 304 is replaced with replacement addresses 611 to 631 that match the access request address. Is included.

  In step S1203, the capacity addition determination unit 305 branches depending on whether replacement addresses 611 to 631 that match the address of the access request are included. If the cache miss address is not included in the replacement history 610 to 630 of the partition corresponding to the process (No determination in step S1203), the process ends.

  When the cache miss address is included in the replacement history 610 to 630 of the partition corresponding to the process (Yes determination in step S1203), the capacity addition determination unit 305, in step S1205, adds the additional capacity acquisition source selection unit. An instruction to add capacity for one page to the partition that issued the access request in which a cache miss has occurred is issued to 306. In step S1207, the additional capacity acquisition source selection unit 306 of the storage control unit 220 refers to the releasable capacity 503 for each partition in the partition capacity setting table 371 and extracts a partition having a releasable capacity.

  In step S1209, the additional capacity acquisition source selection unit 306 branches depending on whether or not a partition having a releasable capacity has been extracted. If there is no extracted partition (No determination in step S1209), the additional capacity acquisition source selection unit 306 ends the process without adding the capacity to the partition instructed by the capacity addition determination unit 305.

  If there is an extracted partition (Yes in step S1209), the additional capacity acquisition source selection unit 306, in step S1211, sets the capacity for one page to the partition for which capacity addition is instructed from the extracted partitions. Select one of the partitions to provide. In step S1211, the additional capacity acquisition source selection unit 306 subtracts the set capacity 502 and the releasable capacity 503 of the selected partition in the partition capacity setting table 371 by one page. In step S1213, the additional capacity acquisition source selection unit 306 adds one page of the set capacity 502 of the partition for which the capacity addition determination unit 305 has been instructed to add capacity in the partition capacity setting table 371. finish.

  The processing from the above steps S1201 to S1213 is performed, the capacity addition determination unit 305 determines the necessity of adding the capacity of the partition, and the additional capacity acquisition source selection unit 306 provides the capacity to the partition that needs the capacity addition. Determine the partition to be completed, and complete the operation to adjust the set capacity of the partition. The processing order from steps S1201 to S1213 is not necessarily the procedure shown in FIG. For example, the processes in steps S1211 and S1213 can be executed at the same time or executed in a different order. Further, in step S1211, the partition providing the capacity may be selected from the partition having the largest releasable capacity 503 with reference to the partition capacity setting table 371, for example.

  As described above, the setting capacity of the partition may be adjusted not in units of one page but in units of a plurality of pages.

(Releasable capacity calculation processing)
FIG. 13 is a flowchart showing the procedure of partition releasable capacity calculation processing in the present embodiment. In this flowchart, an unnecessary page capacity releasing unit 309 extracts pages estimated to be not included in the working set from pages included in each partition defined on the cache, and sets the releasable capacity 503 of each partition. Update.

  The unnecessary page capacity releasing unit 309 of the storage control unit 220 pays attention to the first entry of the page state management table 381 in step S1301.

  In step S1303, the unnecessary page capacity releasing unit 309 refers to the access flag 405 and the unnecessary candidate flag 406 in the entry of interest. Then, the unnecessary page capacity releasing unit 309 determines whether or not the access flag 405 is set in step S1305.

  If the access flag 405 has been set (Yes in step S1305), the unnecessary page capacity releasing unit 309 resets the access flag 405 of the entry of interest in step S1307, and executes the processing from step S1315 onward. . On the other hand, when the access flag 405 is not set (No determination in step S1305), the unnecessary page capacity releasing unit 309 determines whether or not the unnecessary candidate flag 406 is set in step S1309.

  When the unnecessary candidate flag 406 is set (Yes determination in step S1309), the processing after step S1315 is executed. If the unnecessary candidate flag 406 is not set (No determination in step S1309), the unnecessary page capacity releasing unit 309 sets the unnecessary candidate flag 406 of the entry of interest in step S1311. In step S1313, the unnecessary page capacity releasing unit 309 adds the capacity of one page to the releasable capacity 503 of the partition including the page corresponding to the entry of interest in the partition capacity setting table 371.

  In step S <b> 1315, the unnecessary page capacity releasing unit 309 determines whether the entry under attention is the last entry in the page state management table 381. If the currently focused entry is the last entry in the page state management table 381 (Yes in step S1315), the process returns to step S1301 to repeat the process. When the currently focused entry is not the last entry in the page state management table 381 (No determination in step S1317), the unnecessary page capacity releasing unit 309 determines that the focused entry is the next entry in the page state management table 381 in step S1317. And return to step S1303 to repeat the process.

  The processing from the above steps S1301 to S1317 is performed, and the unnecessary page capacity release unit 309 extracts pages estimated to be not included in the working set from the pages included in each partition defined on the cache. An operation of updating the releasable capacity 503 of the partition is executed. The processing order from step S1301 to S1317 is not necessarily the order shown in FIG. For example, the processes in steps S1311 and S1313 can be executed at the same time, or can be executed with the order changed. Further, the procedure for sequentially performing the processing on each entry in the page state management table 381 from steps S1301 to S1317 may be a procedure for paying attention to each entry in an arbitrary order. In addition, as a method of determining that the page corresponding to each entry is not included in the working set and has not been accessed for a long period of time, instead of checking whether the access flag 405 is set, for example, each page The last access time of each page is recorded in the entry of the page state management table 381 corresponding to. Then, it may be a procedure for setting the unnecessary candidate flag 406 by determining that the page is not included in the working set when a certain time has elapsed from the current time.

  According to the present embodiment, it is possible to optimize during processing without measuring the capacity to be allocated to the cache partition corresponding to each processing, and to improve the access performance to the storage medium.

  In other words, based on the access information to the storage medium through the cache, it is determined whether or not the capacity of the partition can be reduced. repeat. As a result, the capacity to be allocated to each partition is optimized without directly measuring the capacity to be allocated to the cache partition corresponding to each process, and the access performance to the storage medium of the entire process executed by the host computer is realized. It is possible to increase.

  Specifically, the detected page is not directly replaced, but only the capacity that can be released from the partition is calculated, and the page to be replaced when the capacity is actually insufficient in another partition is determined. . By this operation, a combination of this embodiment and an arbitrary cache replacement algorithm can be realized. In addition, it is possible to reduce the occurrence of cache misses due to erroneous detection of unnecessary pages by delaying page replacement until it is necessary to change the capacity allocation of each partition.

[Third Embodiment]
Next, a storage control apparatus according to the third embodiment of the present invention will be described. The storage control device according to the present embodiment is different from the second embodiment in that it has a function of adjusting the page cycle period (cyclic rate) in the unnecessary page capacity releasing unit 309. In the present embodiment, unnecessary page detection includes a detection cycle adjustment function that adjusts a cycle for detecting an unnecessary page. Since other configurations and operations are the same as those in the second embodiment, the same configurations and operations are used. The same reference numerals are assigned and detailed description thereof is omitted.

<< Functional configuration of storage controller >>
FIG. 14 is a block diagram illustrating a functional configuration of the storage control unit 1420 as the storage control device according to the present embodiment. In FIG. 14, the same functional components as those in FIG. 3 are denoted by the same reference numerals, and redundant description is omitted.

  Referring to FIG. 14, a page patrol speed adjustment unit 1410 is newly provided as compared with the second embodiment shown in FIG. 3.

  The page circulation speed adjustment unit 1410 instructs the speed of processing in which the unnecessary page capacity release unit 309 refers to each entry in the page state management table 381 to determine whether a page corresponding to the entry is included in the working set. The processing speed is related to the threshold value for determining whether each page is included in the working set. Setting the speed faster shortens the time it takes for each entry to go around once and go around again. The access frequency of pages determined to be included in the set is increased. That is, the unnecessary page capacity releasing unit 309 extracts pages that are not actively included in the working set and that are likely to be unnecessary. Conversely, by setting the processing speed slower, the time until each entry is recirculated becomes longer, and the unnecessary page capacity releasing unit 309 can be set so that unnecessary pages are extracted more strictly. is there.

  The instruction of the processing speed to the unnecessary page capacity releasing unit 309 by the page circulation speed adjusting unit 1410 can be used in the following case. For example, when the releasable capacity 503 is insufficient in the partition defined in the cache, and the capacity addition determining unit 305 cannot satisfy the instruction for adding the capacity of the partition, the processing speed is increased and the unnecessary pages are actively added. Extraction can solve the shortage of the releasable capacity 503.

  According to the present embodiment, it is possible to further adjust the frequency and strictness of the unnecessary page extraction by the unnecessary page capacity releasing unit 309 in consideration of the cache capacity and the access frequency.

[Fourth Embodiment]
Next, a storage control apparatus according to the fourth embodiment of the present invention will be described. The storage control device according to the present embodiment measures the detection time of the working set for adjusting the page cycle period (cyclic rate) in the unnecessary page capacity release unit 309, compared with the second embodiment and the third embodiment. It differs in that it has the function to do. In the present embodiment, a cache hit detection function that detects a cache hit that occurs in access to the cache memory in association with the partition to be accessed, and a page that is frequently accessed to the partition based on the detection result of the cache hit. A detection time measuring function for measuring the time required for detection. Then, the unnecessary page detection cycle is adjusted based on the measured time. Since other configurations and operations are the same as those of the second embodiment, the same configurations and operations are denoted by the same reference numerals, and detailed description thereof is omitted.

<< Functional configuration of storage controller >>
FIG. 15 is a block diagram illustrating a functional configuration of the storage control unit 1520 as the storage control device according to the present embodiment. In FIG. 15, the same functional components as those in FIG. 3 or FIG. 14 are denoted by the same reference numerals, and redundant description is omitted.

  Referring to FIG. 15, a working set detection time measurement unit 1511 is newly provided as compared with the third embodiment shown in FIG.

  The working set detection time measurement unit 1511 obtains the cache hit or cache miss determination result of each access request from the cache hit / miss determination unit 302 and calculates the cache hit rate per fixed time. Further, the working set detection time measurement unit 1511 obtains the frequency of cache misses for which it is determined from the capacity addition determination unit 305 that the capacity addition of the partition is necessary, and calculates the rate at which cache misses occur in all accesses. In addition, the working set detection time measurement unit 1511 compares the calculated cache hit rate with the ratio of cache misses determined to require the addition of the partition capacity, and determines the cache misses determined to require the addition of the partition capacity. Measure the time required for the ratio to reach a certain level compared to the cache hit rate. Then, the working set detection time measurement unit 1511 includes a page rotation speed adjustment unit so that the time interval for which the unnecessary page capacity release unit 309 cycles through the entries in the page state management table 381 corresponding to each page becomes the measurement result time. Adjust speed through 1410.

<Processing speed adjustment processing>
FIG. 16 is a flowchart showing a procedure of processing speed adjustment processing in the fourth embodiment of the present invention. In this flowchart, the processing speed of the unnecessary page capacity releasing unit 309 is optimally maintained through the page circulation speed adjusting unit 1410.

  In step S1601, the working set detection time measurement unit 1511 of the storage control unit 220 sets a counter that measures the number of cache hits, the number of cache misses, and the number of cache misses determined to require additional partition capacity to 0 (zero). Reset to. In step S1603, the working set detection time measurement unit 1511 acquires the cache hit / cache miss determination results of all access requests from the cache hit / miss determination unit 302 until a predetermined time has elapsed. The frequency is measured with a counter. Similarly, in step S1603, the working set detection time measurement unit 1511 acquires information on all cache misses determined to require the addition of a partition capacity from the capacity addition determination unit 305, and measures the frequency using a counter. .

  In step S1605, the working set detection time measurement unit 1511 calculates the cache hit rate a from the cache hit and cache miss frequencies measured in step S1603. In step S1607, the working set detection time measurement unit 1511 calculates the ratio b of the number of cache misses determined in step S1603 that is determined to require additional partition capacity to the total access requests.

  In step S1609, the working set detection time measurement unit 1511 determines whether the ratio b of all access requests exceeds a certain ratio as compared to the cache hit ratio a. When the ratio b of all access requests exceeds a certain ratio as compared to the cache hit ratio a (Yes in step S1609), the working set detection time measurement unit 1511 has insufficient detection of the working set in the cache. And the processing from step S1601 is re-executed.

  When the ratio b in the total access requests is equal to or less than a certain ratio compared to the cache hit ratio a (No in step S1609), the working set detection time measurement unit 1511 detects the working set in the cache in step S1611. Is calculated, and the elapsed time from the start of the process of step S1601 is calculated.

  In step S1613, the working set detection time measurement unit 1511 determines that the time required for the unnecessary page capacity release unit 309 to go through all the entries in the page state management table 381 corresponding to the pages included in the cache is the time calculated in step S1611. Calculate page patrol speed to be equal. Specifically, the number of entries in the page state management table 381 to be circulated per certain time is calculated by dividing the number of entries in the page state management table 381 by the time calculated in step S1611.

  In step S <b> 1615, the working set detection time measuring unit 1511 sets the number of entries calculated in step S <b> 1613 as the processing speed of the unnecessary page capacity releasing unit 309 and ends the processing.

  The processes from step S1601 to S1615 are performed, and the operation of maintaining the processing speed of the unnecessary page capacity releasing unit 309 optimally through the page circulation speed adjusting unit 1410 is completed. The processing order from steps S1601 to S1615 is not necessarily the order shown in FIG. For example, the processes in steps S1605 and S1607 can be executed at the same time, or can be executed with the order changed.

  According to the present embodiment, since the page patrol speed is adjusted by referring to the status of the cache hit in the actual operation, it can be optimized according to the usage status of the cache and the storage medium, and the access performance to the storage medium is further Can be increased.

[Fifth Embodiment]
Next, a storage control apparatus according to the fifth embodiment of the present invention will be described. The storage control device according to this embodiment is different from the second to fourth embodiments in that the storage control unit as the storage control device is not in the host computer, but outside the network, or However, they are different in that they are included in the storage medium. Since other configurations and operations are the same as those of the second embodiment, the same configurations and operations are denoted by the same reference numerals, and detailed description thereof is omitted.

《Information processing system》
FIG. 17 is a block diagram illustrating a configuration of an information processing system 1700 including storage control units 1720 and 1742 as storage control devices according to the present embodiment. In FIG. 17, the same components as those in FIG.

  In the information processing system 1700, a host computer 1710 having no storage control unit, an independent storage control unit 1720 having a cache 221, and a storage medium 1740 including a storage control unit 1742 having a cache 221 are connected via a network 250. It is connected.

  According to the present embodiment, the processing is not limited to processing by the host computer as in the second to fourth embodiments, and the cache is controlled outside the host computer, thereby reducing the load on the host computer and improving the access performance to the storage medium. Improvements can be made.

[Other Embodiments]
The present invention has been described above with reference to the embodiments, but the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. In addition, a system or an apparatus in which different features included in each embodiment are combined in any way is also included in the scope of the present invention.

  In addition, the present invention may be applied to a system composed of a plurality of devices, or may be applied to a single device. Furthermore, the present invention can also be applied to a case where an information processing program that implements the functions of the embodiments is supplied directly or remotely to a system or apparatus. Therefore, in order to realize the functions of the present invention on a computer, a program installed in the computer, a medium storing the program, and a WWW (World Wide Web) server that downloads the program are also included in the scope of the present invention. . In particular, at least a non-transitory computer readable medium storing a program for causing a computer to execute the processing steps included in the above-described embodiments is included in the scope of the present invention.

Claims (11)

In the cache memory, a reduction candidate setting means for setting a partition including a page that has not been accessed for a predetermined time or a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity Capacity control means to be added to the capacity of
A storage control device comprising: The reduction candidate setting means includes:
An unnecessary page detection unit that detects a page that has not been accessed as an unnecessary page with the predetermined time as a cycle,
The storage control device according to claim 1, wherein a partition including the unnecessary page is set as a partition capable of reducing the capacity.   The storage control device according to claim 2, wherein the unnecessary page detection unit detects a page that has not been accessed for the predetermined time by using a clock algorithm.   The storage control device according to claim 2, wherein the unnecessary page detection unit includes a detection cycle adjustment unit that adjusts the cycle for detecting the unnecessary page. A cache hit detecting means for detecting a cache hit generated in accessing the cache memory in association with a partition to be accessed;
A detection time measuring means for measuring a time required for detecting a page having a high access frequency to the partition from the detection result of the cache hit, and further comprising:
The storage control device according to claim 4, wherein the detection cycle adjustment unit adjusts a detection cycle of the unnecessary page based on the measured time.   6. The capacity control unit according to claim 1, wherein when there are a plurality of partitions capable of reducing the capacity, a partition including a page most likely to be evicted to a lower storage medium is set as a partition whose capacity is reduced. The storage control device according to any one of the above.   The storage control device according to any one of claims 1 to 6, wherein the capacity control unit reduces a page most likely to be evicted to a lower storage medium in a partition whose capacity is to be reduced. In a cache memory, a reduction candidate setting step for setting a partition that includes a page that has not been accessed for a predetermined time or a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity A capacity control step to add to the capacity of
A storage control method. In a cache memory, a reduction candidate setting step for setting a partition that includes a page that has not been accessed for a predetermined time or a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity A capacity control step to add to the capacity of
Storage control program for causing a computer to execute. A storage medium;
Access means for instructing access to the storage medium based on access requests from a plurality of processes;
Storage control means for controlling access to the storage medium of the access means, having a cache memory;
An information processing system comprising:
The storage control means
In the cache memory, a reduction candidate setting means for setting a page that has not been accessed for a predetermined time or a partition including a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which a cache miss has occurred is included in the history of pages that have been evicted from the cache memory to the lower storage medium, the cache miss has occurred by reducing the capacity of the partition capable of reducing the capacity. Capacity control means to add to the capacity of the partition;
An information processing system comprising: An assigning step of assigning a partition including at least one page in the cache memory to each of the plurality of processes;
In the cache memory, a reduction candidate setting step for setting a page that has not been accessed for a predetermined time or a partition including a page whose access frequency is lower than a predetermined value as a partition whose capacity can be reduced;
When the page in which the cache miss has occurred is included in the history of the page that has been evicted from the cache memory to a lower storage medium, the partition in which the cache miss has occurred by reducing the capacity of the partition that can reduce the capacity A capacity control step to add to the capacity of
While executing the reduction candidate setting step in the background, the capacity control step is repeated, and a distribution management step for managing the distribution of the cache capacity for each process;
Cache capacity allocation method including

Images