JP2017058904A - Cache device and control method for cache device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a cache device capable of holding data in a cache so as to correspond to variation in access density, and a control method thereof.SOLUTION: A cache device includes: a cache 1; access density variation analysis means 2; cache holding priority setting means 3; and cache data control means 4. The cache temporarily holds the data of a storage device. The access density variation analysis means monitors access to the storage device, and analyzes variation in access density. The cache holding priority setting means sets the priority of data to be held in the cache on the basis of the analysis result of the access density variation analysis means. The cache data control means controls the arrangement of the data held in the cache on the basis of the priority.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a cache device and a cache device control method.

  In recent years, the capacity of storage systems has been increasing. In a large-capacity storage system, a large-capacity storage device represented by a disk array is used as a storage means. However, the mass storage devices described above often have low response performance. For this reason, the response performance is improved by using a cache having a higher response performance and a smaller capacity than the storage device.

  The cache temporarily holds data accessed by the storage device. If the next access request is made while this data is held, the response can be returned without accessing the storage device, so that the response can be speeded up.

  As a general operation, when certain data is accessed, the data is arranged as new data at the head of the cache retention order. At this time, if the cache is full, the last data in the retention order is discarded from the cache. Therefore, there is a case where there is data on the cache when there is an access request and a case where there is no data on the cache. When the data is on the cache, it is called a cache hit. On the other hand, when the requested data is not in the cache, it is called a cache miss hit. As is apparent from the above operation, a mechanism for increasing the cache hit rate is required to improve storage performance.

  The most common method is LRU (Least Recently Used). In the LRU, the data that has not been referred to for the longest time among the data held in the cache is discarded. This is a concept based on the assumption that data that has not been referenced for a long time has a low probability of being referred to in the future. However, in LRU, there is a problem that if there is data that has been recently referenced but is not used frequently, it remains in the cache for a long time without being used.

  One of the methods for solving the problem is LFU (Least Frequently Used) that discards data that is least frequently used. In the LFU, a table of the number of times each data is referred to in a certain period is created, and data to be discarded is determined based on the table. For this reason, the problem that data with low usage frequency remains is solved. However, in the LFU, there is a difference between a period in which the table is created and a period in which data is controlled using the table. For this reason, there is a problem that it does not function effectively in an environment where the access frequency changes dynamically.

  Various methods for solving the above-mentioned problems of LRU and LFU have been proposed. For example, Patent Document 1 discloses a technique corresponding to an environment where the access frequency dynamically changes, which is a problem of LFU. In this technique, an access frequency collection table and a data control table are prepared at the same time, and are switched alternately at a predetermined interval (for example, 1 second). As a result, the time lag between the collection and control of the access frequency is reduced, and even if the access frequency changes dynamically, it can be followed.

  Further, for example, Patent Document 2 discloses a technology that configures a hierarchical cache system that is hierarchized with high response performance and holds data in an appropriate hierarchy even in a dynamic access environment. In this technique, a threshold value is set to determine whether the lower cache accepts data demoted from the upper cache, and the threshold is dynamically adjusted. Specifically, the hit count of the data MRI inserted last in the lower cache and the cache MRE of the data deleted last is compared. If MRI> MRE, the threshold is decreased, and if MRI <MRE, Increase the threshold. By dynamically adjusting the threshold in this way, appropriate data can be held in the cache even if the access frequency changes dynamically.

Re-specialized WO1099 / 40516 Special table 2014-535106 gazette

  However, the techniques of Patent Documents 1 and 2 are effective when the total amount of access does not change so much and the frequency of access to individual data changes, but when the total amount of access changes. There is a problem that it is not valid.

  For example, it is not effective for a logon storm, which has been a problem in recent years. A logon storm is a problem in which a large number of client terminals log on to a server or storage all at once in a virtual desktop environment or the like, resulting in a significant decrease in response. For example, taking the start time of a company or the like as an example, a logon storm occurs once a day on weekdays. Here, the data on which access is concentrated are rarely referred to at other times, and therefore, in any of the techniques of Patent Documents 1 and 2, they are discarded from the cache before the start of the next day. For this reason, at the start of the next day, access is concentrated again, and cache data is exchanged all at once. As a result, the load becomes larger than when there is no cache.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide a cache device that can hold data in a cache so as to cope with time fluctuations in access density.

  In order to solve the above-described problems, the cache device of the present invention includes a cache, an access density fluctuation analysis unit, a cache retention priority setting unit, and a cache data control unit. The cache temporarily holds data in the storage device. The access density fluctuation analyzing means monitors access to the storage device and analyzes the density fluctuation of the access. The cache holding priority setting means sets the priority of data held in the cache based on the analysis result of the access density fluctuation analyzing means. The cache data control means controls the arrangement of the data held in the cache based on the priority.

  An effect of the present invention is to provide a cache device that can hold data in a cache so as to cope with a change in access density.

It is a block diagram which shows 1st Embodiment. It is a block diagram which shows 2nd Embodiment. It is an example of the graph showing the fluctuation | variation of access density. It is a flowchart which shows operation | movement of 2nd Embodiment. It is a flowchart which shows the defined process of 2nd Embodiment. It is a conceptual diagram which shows the example of data arrangement | positioning of 3rd Embodiment. It is a table | surface which shows the example of weighting of 3rd Embodiment. It is a schematic diagram which shows an example of operation | movement of 3rd Embodiment. It is a schematic diagram which shows another example of operation | movement of 3rd Embodiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. However, the preferred embodiments described below are technically preferable for carrying out the present invention, but the scope of the invention is not limited to the following. In addition, the same number is attached | subjected to the same component of each drawing, and description may be abbreviate | omitted.

(First embodiment)
FIG. 1 is a block diagram showing the first embodiment. The cache device includes a cache 1, an access density fluctuation analysis unit 2, a cache retention priority setting unit 3, and a cache data control unit 4. The cache 1 temporarily holds data in the storage device. The access density fluctuation analyzing means 2 monitors access to the storage device and analyzes the access density fluctuation. The cache holding priority setting means 3 sets the priority of data held in the cache based on the analysis result of the access density fluctuation analyzing means. The cache data control means 4 controls the arrangement of the data held in the cache based on the priority.

  With the above configuration, it is possible to set the priority according to the fluctuation of the access density to the data and hold the data in the cache.

(Second Embodiment)
A storage system can be configured using the cache device of the first embodiment. FIG. 2 is a block diagram showing the storage system of this embodiment. The storage system includes a storage device 100, a cache device 200, and a storage control device 300. In FIG. 2, only one storage device 100 is illustrated, but a plurality of storage devices 100 may be provided.

  The storage device 100 stores data permanently. Specifically, a magnetic disk, an optical disk, a disk array obtained by arraying them, and the like can be used. Note that the present embodiment is not limited to the storage method of the storage device.

  The cache device 200 includes an access density fluctuation analysis unit 210, a cache retention priority setting unit 220, a data arrangement control unit 230, and a cache 240.

  The cache 240 is a storage device that has higher access performance than the storage device 100 and temporarily holds data. For example, an SSD (solid state drive) using a nonvolatile memory can be used. However, the present invention is not limited to this.

  The access density fluctuation analysis unit 210 counts the number of accesses in a predetermined unit time as the access density, and analyzes the fluctuation. By analysis, it is possible to grasp a tendency of access density fluctuation such as a period of time when the access density is regularly increased.

  Based on the analysis result of the access density fluctuation analysis unit 220, the cache retention priority setting unit 230 sets a priority for continuing the retention for the data held in the cache.

  The data arrangement control unit 230 arranges data in the cache 210 according to the priority set by the cache holding priority setting unit 220. The arrangement is such that, for example, the one with the highest priority is arranged in order.

  Before describing the operation of the present embodiment, the problem solved by the present embodiment will be described.

  In recent years, virtual machines (Virtual Machines, VMs) and desktop virtualization (Virtual Desktop Infrastructures, VDIs) have been introduced. In such an environment, there is a problem that access concentrates at a specific time. FIG. 2 is a graph showing an example of time variation of access density. In this example, the state is a logon storm where access is concentrated in the time zone from 8:20 to 8:50. Such a situation occurs periodically, for example, in the same time zone at the start of work on weekdays.

  As described above, in a general technique, a cache device optimizes data held in a cache for an access frequency in a short period. The total amount of access during the predetermined period, that is, the variation in density as described above is not a problem to be solved by the cache device. In the present embodiment, a cache device is actively used to improve performance degradation caused by a rapid increase in access density, such as the above-described logon storm.

  Next, the operation of the cache device in the storage system will be described. FIG. 4 is a flowchart showing the operation of the cache device.

  First, the access density fluctuation analysis unit analyzes the access density fluctuation (S1). Then, for example, a time zone in which the access density rapidly increases and a storage device area accessed at that time are extracted as analysis results. Specifically, an analysis result that the access density rapidly increases between 8:20 and 8:50 every day is obtained from the fluctuation of the access density as shown in FIG.

  Next, a weight a is set as a coefficient to be given to the data for each time zone (S2). Next, when the storage system accepts access to data not held in the cache, the cache device newly holds the data in the cache (S3). At this time, the data with the lowest priority is discarded from the cache, and the weight coefficient a is assigned to the newly held data and held at the head of the cache (position with the highest priority) (S4). Thereafter, a predefined process for setting the cache holding priority is performed (S5). Next, the data arrangement control unit arranges data in the cache in the order of the priority set in S5 (S6).

  Next, an example of the operation for setting the cache holding priority in S5 will be described. Here, a method for quantifying priority with cumulative weight will be described. The cumulative weight is a numerical value of the priority based on the magnitude. The cumulative weight is assigned to each data and updated in the following procedure.

  First, the priority setting unit determines the weight coefficient a based on the access density fluctuation analysis result. The weighting factor a is set such that a ≧ 1 and increases in the time zone when the access density is high and decreases in the time zone where the access density is low.

  First, when data that is not held in the cache is accessed and is newly held in the cache, the accumulated weight when the data is not held is set to 1, and 1 × a = a is assigned as an initial value at the holding stage.

  Next, when there is an access to data held in the cache, the cumulative weight is updated with a value obtained by multiplying the current cumulative weight by a.

  The above operation is shown in a flowchart in FIG. First, access to data held in the cache is accepted by an external access request (S51). When there is an access to the data, the cache retention priority setting unit calculates a value obtained by multiplying the original accumulated weight of the data by the weighting factor a, and updates the accumulated weight using this as a new value (S52). . Thus, the update of the accumulated weight by one access is completed.

  According to the above operation, the cumulative weight increases every time data is accessed. Therefore, the higher the access, the higher the cache retention priority.

  Here, a time zone determined as high density (high load) by the access density fluctuation analysis will be considered. First, a large weighting factor is given to data accessed in this time zone. Further, the data is repeatedly accessed during this time period, and the cumulative weight increases each time. As a result, the cumulative weight of the data becomes very large, and the period for holding it on the secondary cache becomes very long. On the other hand, data that is not accessed during this time zone is preferentially discarded from the cache because the accumulated weight is small. With the above operation, the data hit frequently in a specific time zone has a high cache hit rate in the time zone, and the data can be output to the host without accessing a low-speed storage device. In this way, it is possible to suppress the performance degradation of the storage system.

  In the above description, the calculation method of multiplying the weighting factor by the number of accesses is exemplified, but the priority may be quantified by another calculation method in which the priority increases as the number of accesses increases.

As described above, according to the present embodiment, it is possible to increase the cache hit rate of data that is accessed during a period of time when the access density increases rapidly. As a result, the performance of the storage system in the above time zone can be greatly improved.
(Third embodiment)
In this embodiment, a specific example of the configuration and operation of the cache device will be described.

  FIG. 6 is a conceptual diagram showing data arrangement in the cache. Here, it is assumed that the cache has a hierarchical structure of a primary cache 241 and a secondary cache 242. The primary cache 241 operates by FIFO (First-In First-Out), and the secondary cache 242 operates according to the priority setting by the cumulative weight of the second embodiment.

  When data is accessed, the data is placed at the head of the primary cache 241. In FIG. 6, data is arranged in the order of H, G, and F in the order of new access. This arrangement is performed by FIFO, and the cumulative weight of each data is ignored.

  Data evicted from the primary cache 241 is held in the secondary cache 242. In the secondary cache 242, data is prioritized and held by the cumulative weight. In FIG. 6, the data A, C, B, E, D,. In this case, the eviction target when the new data is held next is ZZ.

  FIG. 7 is a table showing an example of setting weighting factors. In this example, a weighting coefficient in each time zone is set for each logical disk. In this example, in the logical disk 00, the weighting coefficient is switched from 1 to 2 at 08:20, and is 2 from 08:20 to 08:30 (other time zones are not shown). The fact that the weighting factor is switched from 1 to 2 at 08:20 reflects the rapid increase in access density as shown in FIG. 3, for example. The same applies to the logical disk 10. In this example, the same weighting factor is set for the logical disk 00 and the logical disk 10, but different values may be used. The weighting factor may be set separately for each area divided into logical disks or storage devices. A large weighting factor may be set for access from a specific host. For example, a large weighting factor can be set for a host that has installed a license that guarantees a high-speed response. This makes it possible to provide a high-speed response service.

  Next, a specific example of the data arrangement operation will be described. Here, it is assumed that the operation is in a time zone in which the weighting factor is 2. FIG. 8A shows an initial data arrangement in the cache. If there is an access to the data K stored in the storage device 100 that is not in the cache, the data K is added with a weighting factor 2 as shown in FIG. 2B and placed at the head of the primary cache 241 (see FIG. 2). Solid line arrow). On the other hand, the ZZ having the lowest priority in the secondary cache 242 is evicted and stored in the storage device 100 (dotted line arrow in the figure).

  When the data H in the primary cache is accessed from the state of FIG. 8B, the original cumulative weight is multiplied by the weighting factor 2, and 16 × 2 = 32 is set at the head of the primary cache 241 as the cumulative weight. (FIG. 9A). Next, when the data A in the secondary cache 242 is accessed, the original accumulated weight is multiplied by the weighting factor 2, and the accumulated weight is set to 64 × 2 = 128 at the head of the primary cache 241 (FIG. 9B). )). On the other hand, since the data G evicted from the primary cache has a cumulative weight of 2, it is arranged under the data B having a cumulative weight of 16 in the secondary cache 242.

  As described above, according to this embodiment, data accessed in a time zone with a high access density is likely to stay in the secondary cache. For this reason, in a scene where the access density increases regularly, the probability that the data of the secondary cache can be used increases, and the performance can be improved.

  The present invention has been described above using the above-described embodiment as an exemplary example. However, the present invention is not limited to the above embodiment. That is, the present invention can apply various modes that can be understood by those skilled in the art within the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Cache 2 Access density fluctuation | variation analysis means 3 Cache retention priority setting means 4 Cache data control means 100 Storage device 200 Cache apparatus 210 Access density fluctuation analysis part 220 Cache retention priority setting part 230 Data arrangement control part 240 Cache 241 Primary cache 242 Secondary cache 300 storage controller

Claims (10)

A cache that temporarily holds data in the storage device;
Access density fluctuation analysis means for monitoring access to the storage device and analyzing fluctuations in the access density;
Cache holding priority setting means for setting the priority of data held in the cache based on the analysis result of the access density fluctuation analyzing means;
Cache data control means for controlling the arrangement of data held in the cache based on the priority;
A cache device comprising: The access density fluctuation analysis means
Extracting a time period in which the access density to the storage device is regularly larger than a predetermined value;
A cache device. The cache retention priority setting means includes
Giving higher priority to data accessed during the time period than data not accessed during the time period;
The cache device according to claim 2. The cache retention priority setting means includes
Set a priority weighting factor corresponding to the access density based on the analysis result,
The priority is quantified based on the weighting factor and the number of accesses, and set in the data.
The cache device according to claim 1. A cache device according to any one of claims 1 to 4,
A storage device having lower access performance than the cache and permanently storing data;
A data control device that controls data held in the cache and data stored in the storage device;
A storage system comprising: Temporarily cache storage device data,
Monitor accesses to the storage device and analyze variations in the access density;
Set the priority of data held in the cache based on the analysis result of the access density variation,
Controlling the placement of data held in the cache based on the priority;
And a control method for the cache device. The access density fluctuation analysis means
Extracting a time period in which the access density to the storage device is regularly larger than a predetermined value;
The method of controlling a cache device according to claim 6. The cache retention priority setting means includes
Giving higher priority to data accessed during the time period than data not accessed during the time period;
The method of controlling a cache device according to claim 7. The cache retention priority setting means includes
Set a priority weighting factor corresponding to the access density based on the analysis result,
The priority is digitized based on the weighting factor and the number of accesses, and set in data.
The method of controlling a cache device according to claim 7. Temporarily caching data in the storage device;
Monitoring access to the storage device and analyzing variations in the density of the access;
Setting the priority of the data held in the cache based on the analysis result of the variation in the access density;
Controlling the placement of data held in the cache based on the priority;
A control program for a cache device, comprising:

Images