JP3287300B2 - Temporary storage area management method and device, and machine-readable recording medium recording program - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a technique for effectively using a temporary storage area of an information processing device.

[0002]

2. Description of the Related Art In a conventional cache management method / device or buffer management method / device, as a method used when a temporary storage area (cache buffer) having a limited size is effectively used, LRU (Least Recently) is used. Used)
The method and the LFU (Least Frequently Used) method have been used.

That is, in the LRU method, if the data to be referred to does not exist in the temporary storage area, and if there is no space for storing the data in the temporary storage area, the data is stored in the temporary storage area. By selecting the least recently used data from the existing data and returning that data to the original storage area or invalidating it,
Free space is secured in the temporary storage area. Also,
In the LFU method, data with the least frequency of reference is selected from the data stored in the temporary storage area, and the data is returned to the original storage area or invalidated to secure a free space. ing.

[0004] The LRU method has a disadvantage that even if data is frequently referred to, if the period during which the data is not referred to happens for a long time, the data is expelled from the temporary storage area. Further, in the LFU method, data frequently referred to during a certain period is
Even if it is not referred to thereafter, it remains in the temporary storage area for a long time. As described above, it is difficult for the conventional LRU method or LFU method to effectively use the temporary storage area.

As a method for overcoming such a drawback, a method based on the following paper described in 1993, ACM SIGMOD Conf. Are known. El
izabeth J.O'Neil, Patric E.O'Neil, Gerhard Weikum
Author, The LRU-K Page Replacement Algorithm For Dat
abase Disk Buffering ".

[0006] This is to store the k most recent times referring to the data in the buffer (temporary storage area) for each data,
When making an empty space in the buffer, a method is used in which the time point referred to k times earlier is the oldest, that is, the current time minus the reference time point k times earlier is selected. In this method, the case where k = 1 is equivalent to the ordinary LRU method. In the following description, the above method is referred to as LRU
-Called the K method.

On the other hand, Japanese Patent Application Laid-Open No. 7-152649 describes a caching method in which an access time (reference cost) to an original data storage area (main storage device or the like) is considered. By performing caching in consideration of the reference cost, in a system such as a NUMA (Non Uniform Memory Access) machine having storage areas with different reference costs, the temporary storage area is effective compared to a case where the reference cost is not considered. It can be used. However, in the caching method described in the above publication, when securing a free space in the cache memory, data having a high reference cost among data stored in the cache memory is left, and data having a low reference cost is discarded. It is simple and does not consider the data reference history at all.

[0008]

According to the LRU-K method, the disadvantages of the LRU method and the LFU method can be solved, but there are the following disadvantages. As described above, the case where k = 1 is equivalent to the normal LRU method, and when the value of k is not so large, the past history is hardly considered, so that in terms of performance, There is not much difference from the LRU method. Therefore, in order to solve the disadvantage of the LRU method, it is necessary to increase the value of k. However, if the value of k is increased, the LRU-K method, which needs to record the latest k reference times for each data,
The hardware scale for managing the reference time becomes large, and the processing for managing the data becomes complicated. Further, the LRU-K method has a problem that data whose reference count does not reach k times cannot be handled.

An object of the present invention is to make it possible to effectively use a temporary storage area without increasing the hardware scale and without complicating the processing.

Another object of the present invention is to enable a temporary storage area to be effectively used in a system having storage areas having different reference costs.

[0011]

According to the method of managing a temporary storage area of the present invention , a system having storage areas having different reference costs is provided.
In order to make it possible to effectively use the temporary storage area without increasing the hardware scale or complicating the processing, the method for managing the temporary storage area in which data is temporarily stored includes the steps of: For each data stored in the area, the storage time when first stored in the temporary storage area, the reference count, and the original
And managing the reference cost indicating the access time to the storage area, and when it becomes necessary to make an empty space in the temporary storage area, from among the data stored in the temporary storage area, The estimated value of the reference interval determined by considering the time and the number of references is normalized by the reference cost.
The data having a large value at the time of conversion is selected, and a portion where the selected data is stored is made empty.

[0012]

Further, the temporary storage area management device of the present invention is applicable to a system having storage areas having different reference costs.
And a temporary storage area management device that manages a temporary storage area in which data is temporarily stored so that the temporary storage area can be used effectively without increasing the hardware scale or complicating the processing. In each of the data stored in the temporary storage area, a reference time indicating a storage time when the data is first stored in the temporary storage area, a reference count, and an access time to the original storage area.
A management table for managing costs, and when it is necessary to make room in the temporary storage area, from the data stored in the temporary storage area, the storage time managed in the management table and Selects data having a large value when the estimated value of the reference interval determined by considering the number of times of reference is normalized by the reference cost , and makes the portion where the selected data is saved empty. Management means.

In this configuration, the management table stores, for each data stored in the temporary storage area, a storage time when the data is first stored in the temporary storage area, a reference count ,
Reference cost indicating the access time to the original storage area
When it becomes necessary for the management means to make room in the temporary storage area, the storage time and the number of times of reference managed in the management table are selected from the data stored in the temporary storage area. estimate of the reference interval determined by considering the the
A value having a large value when normalized by the reference cost is selected, and a portion where the selected data is stored is made empty.

[0015]

[0016]

[0017]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 illustrates a technique on which the present invention is based .
, An information processing device 5 including an arithmetic device 1 and a temporary storage device 2, a storage device 3, an information processing device 4, and a recording medium 6.

The temporary storage device 2 in the information processing device 5 has a temporary storage area 23 for temporarily storing reference data.
And management table 22 having management information for managing data
And a management unit 21 that manages the temporary storage area 23 using the management table 22. The arithmetic device 1 stores data stored in the storage device 3 or another information processing device 4 based on a program or an instruction of a user of the information processing device 5.
The process proceeds with reference to the data generated by.

The recording medium 6 connected to the information processing device 5 is a disk, a semiconductor memory, or another recording medium, and is used for operating the information processing device 5 constituted by a computer as a temporary storage area management device. The program has been recorded. This program is read by the information processing device 5 and controls the operation of the information processing device 5 to realize the management unit 21 on the information processing device 5.

The management table 22 includes at least the temporary storage area 2
It is necessary to have the management information of the data stored in No.3. If the management information of the data stored in the temporary storage area 23 is provided,
May have management information on data that has been evicted from (data previously stored in the temporary storage area 23).

As shown in FIG. 2, the management information managed by the management table 22 includes, as management items, an identifier ID for distinguishing data, the latest reference time TR of the data, and a temporary storage area. 23 includes the time (storage time) TF at the time of first storage and the number n of data references from the storage time TF to the current time. Where the identifier I
As D, for example, information indicating the original storage location of the data can be used. If the original storage location of the data is, for example, address 100 on the storage device 3, information indicating that can be used as the identifier ID. Regarding the storage time TF, if the management table 22 is large enough to store management information for all data referenced by the arithmetic device 1,
As described above, this indicates the time when the data is first stored in the temporary storage area 23. However, if the management table 22 is not large enough, the data is evicted from the temporary storage area 23. After that, it may indicate the time when the data is stored in the temporary storage area 23 again.

In the example of the contents of the management table 22 shown in FIG. 2, the data whose identifier ID is "A100"
R, the storage time TF, and the reference count n are “t1”,
The data of the identifier ID “A50” at “t2” and “10” indicates that the latest reference time TR, storage time TF, and reference count n are “t3”, “t4”, and “5”, respectively. I have.

[0024] Next, a description will be given of operation. In the following description, it is assumed that the size of each reference data stored in the temporary storage area 23 is all equal.

When referring to the data, the arithmetic unit 1 outputs a data reference request including the original storage position of the data to be referenced.

When a data reference request is output from the arithmetic unit 1, the management means 21 in the temporary storage device 2 stores the requested data (reference data) in the temporary storage area as shown in the flowchart of FIG. 23 is determined (S31).

If it is determined that the reference information exists, the latest reference time TR is updated to the current time with respect to the management information corresponding to the reference data in the management information stored in the management table 22. An update process of increasing the number n by 1 is performed (S36). On the other hand, when it is determined that the reference data does not exist, it is determined whether or not there is enough free space in the temporary storage area 23 to store the reference data (S3).
2).

If it is determined in S32 that there is a free space, the reference data is stored there (S34). On the other hand, if it is determined that there is no free space, the temporary storage area 2
3 (S33), and stores the reference data therein (S34).

FIG. 4 is a flowchart showing a detailed processing example of S33, and a method of creating an empty space will be described with reference to FIG.

First, an estimated value of the reference interval is obtained for each data stored in the temporary storage area 23 (S41). The estimated value of the reference interval is obtained, for example, as follows. The management information of each data stored in the temporary storage area 23 is acquired from the management table 22. Thereafter, for each data, the latest reference time T in the management information corresponding to each data is set.
The estimated value of the reference interval of each data is obtained by performing the calculation represented by the following equation (1) based on R, the storage time TF, and the number of references n.

Estimated value of reference interval = (TR−TF) / (n−1) (1)

When the estimated value of the reference interval of each data is obtained in S41, one data having the largest estimated value of the reference interval is selected from the data stored in the temporary storage area 23 (S42).

Thereafter, the data selected in S42 is returned to the original storage position or discarded, thereby making a space in that portion (S43). Through the above processing, a space for storing the reference data is created in the temporary storage area 23.

When storing the reference data in the temporary storage area 23 (S34 in FIG. 3), the management means 21 determines whether or not the management information of the reference data exists in the management table 22 (S35).

If it is determined that the reference data exists, the reference time TR is updated to the current time with respect to the management information of the reference data, and an update process of increasing the number of references n by 1 is performed (S36). On the other hand, if it is determined that the management information does not exist, it is determined whether or not there is enough space in the management table 22 to store the management information (S37).

If it is determined in S37 that there is a free space in the management table 22, the management information of the reference data is newly stored there (S39). In this case, the content of each management item of the management information, that is, the identifier ID, the latest reference time T
The contents of R, storage time TF, and reference count n are “identifier of reference data”, “current time”, “current time”,
It becomes "1".

On the other hand, in S37, the management table 2
If it is determined that there is no empty space in the management table 22, a space is created in the management table 22 (S38), and the management information of the reference data is stored there (S39). As a method of creating an empty space in the management table 22, for example, one of the management information in which data corresponding to the temporary storage area 23 is not stored is selected randomly or with the largest estimated value of the reference interval. Alternatively, a method of discarding the selected management information can be adopted.

In the above description , the value obtained by the above equation (1) is used as the estimated value of the reference interval. However, for example, the value obtained by the following equation (2) may be used as the estimated value of the reference interval. . When using the equation (2), the management table 2
In 2, it is not necessary to manage the latest reference time TR, and it is sufficient to manage the identifier ID, the storage time TF, and the number of times of reference n.

Estimated value of reference interval = (current time−TF) / n (2)

Further, a value obtained by the following equation (3) can be used as an estimated value of the reference interval. In addition, in Formula (3), a, b, c, and d are positive values.

Estimated value of reference interval = {a × (current time−TR) + b × (1 / n) + c × (current time−TF) / n + d × (TR−TF) / (n−1)} (A + b + c + d) (3)

Further, the configuration of the above-described base technology can be modified as follows. Apart from the time generally used as the time, a counter value that can increase the value each time data is referenced may be used. The reference data is not only data stored in the storage device 3 or data generated by another information processing device 4, but also information processing device 5.
May be temporarily generated data. Also,
The information processing device 5 may be configured to include the storage device 3. Further, a plurality of storage devices 3 and other information processing devices 4 may be provided.

Further, in the above description , the sizes of the reference data stored in the temporary storage area 23 are all the same, but when the sizes of the reference data are different,
The processing performed in S33 of FIG. 3 may be performed as follows. That is, the process of selecting the data stored in the temporary storage area 23 in order from the one with the largest estimated value of the reference interval will be performed by storing the accumulated value of the size of the selected data in the temporary storage area 23. This is performed until the size of the reference data becomes equal to or larger than the size of the reference data, and then the portion where the selected data is stored may be made empty.

[0044] will now be described real施例of the present invention. Figure 5 is a block diagram of a real施例of the present invention, an information processing apparatus 5a that includes an arithmetic unit 1 and the temporary storage device 2a, a storage device 3, the information processing apparatus 4, is composed of a recording medium 6a I have.

The temporary storage device 2a includes management means 21a,
It is composed of a management table 22a, a temporary storage area 23, and a reference cost storage unit 24.

The reference cost storage unit 24 records the reference costs (access times) of all storage areas accessible by the arithmetic unit 1.

As shown in FIG. 6, the management table 22a stores, as data management information, an identifier ID for distinguishing data, the latest reference time TR of the data, and the data in the temporary storage area 23. The time (storage time) TF at the time of first storage, the number of data references n from the storage time TF to the current time, and the reference cost G are stored. In the example of FIG. 6, the latest reference time TR, storage time TF, reference count n, and cost G of the data having the identifier ID “A100” are “t1,” “t2,” “10,” and “g1,” respectively.
Latest reference time T of data with identifier ID “A50”
R, the storage time TF, the reference count n, and the reference cost G are “t3”, “t4”, “5”, and “g2”, respectively.

The management unit 21a has a function of selecting data to be evicted from the temporary storage area 23 by using an estimated value of the reference interval normalized by the reference cost when creating a free space in the temporary storage area 23, and temporarily storing the selected data. It has a function of creating an empty space in the temporary storage area 23 by eviction from the storage area 23.

The recording medium 6a connected to the information processing apparatus 5a is a disk, a semiconductor memory, or another recording medium, and is used for causing the information processing apparatus 5a formed by a computer to function as a temporary storage area management device. The program has been recorded. This program is read by the information processing device 5a, and controls the operation of the information processing device 5a, thereby realizing the management unit 21a on the information processing device 5a.

Next, the operation of this embodiment will be described. The operation of the present embodiment is performed in S33 of the flowchart shown in FIG.
And that performs the processing shown in the flowchart of FIG. 7 in place of the processing shown in the flowchart of, except for processing S39a of FIG. 8 in place of S39 in FIG. 3, the base technology shown in FIG. 1 Since the processing is the same as that described above, only the processing performed in S33 and the processing performed in S39a will be described.

In S33, when creating a free space in the temporary storage area 23, the management means 21a creates a free space in the temporary storage area 23 according to the procedure shown in the flowchart of FIG.

First, the management unit 21a stores the temporary storage area 2
For each of the data stored in No.3, an estimated value of the reference interval is calculated using the above-described formula (1), formula (2) or formula (3) (S71). Note that the value of the current time −TR or the value of 1 / n may be used as the estimated value of the reference interval.

Thereafter, the estimated value of the reference interval of each data obtained in S71 is calculated by the following equation (4) to obtain the estimated value of the reference interval normalized by the reference cost (S72). ).

Normalized reference interval estimated value = (reference interval estimated value) ÷ (reference cost) (4)

For example, the estimated value of the normalized reference interval of the data of the identifier A100 is (estimated value of the reference interval) ÷ g
The estimated value of the normalized reference interval of the data of the identifier A50 is (estimated value of the reference interval) ÷ g2 (see FIG. 6).

In S72, when the estimated value of the reference interval normalized for each data is obtained, the temporary storage area 23
Is selected from among the data stored in the storage area (S7).
3).

Thereafter, the data selected in S73 is returned to the original storage position or discarded, thereby creating a space in the temporary storage area 23 for storing the reference data (S74). The above is the processing performed in S33. In this way, by making a space in a portion where data with a large estimated value of the normalized reference interval is stored,
If the reference histories are the same, data with a higher reference cost can be left in the temporary storage area 23 for a long period of time, so that in a system having storage areas with different reference costs, the temporary storage area 23 can be effectively used. become.

Next, the processing of S39a shown in FIG. 8 performed in place of S39 will be described. In S39a, FIG.
In S34, as the management information of the reference data stored in the temporary storage area 23, the management information having the following contents set in the respective management items is stored in an empty space of the management table 22a. That is, “identifier of reference data”, “current time”, “current time”, “1”, and “reference data” are assigned to identifier ID, latest reference time TR, storage time TF, reference count n, and reference cost G, respectively. Is set and stored in the management table 22a. The reference cost of the reference data is obtained from the reference cost storage unit 24.

It is needless to say that the modification described in the description of the base technology in FIG. 1 is also possible in this embodiment. Furthermore, the reference cost is not a known value (a statically determined value) as described above, but a value (a dynamically determined value) that is known only when data is actually referenced, and an estimated value. It does not matter. When the reference cost is dynamically determined, the reference cost storage unit 24 is unnecessary.

[0060]

As described above, when it becomes necessary to make a space in the temporary storage area, the cost is reduced by the reference cost.
Since the data to be evicted from the temporary storage area is determined using the normalized reference interval estimation value, the reference cost
LRU in a system with different storage areas
Method and the LFU method can be solved, and the temporary storage area can be effectively used. As a result, it is possible to reduce the processing time of various operations performed using the information processing device.

Information required to be managed in order to obtain an estimated value of the reference interval normalized by the reference cost includes a data storage time, a data reference count , and a reference count .
Since only the illumination cost is required, the temporary storage area can be effectively used with an economical hardware configuration and simple processing compared to the LRU-K method. Further, the LRU-K method cannot handle data whose reference count does not reach k, but the present invention eliminates such a limitation.

[0062]

[Brief description of the drawings]

FIG. 1 is a block diagram of a base technology of the present invention.

FIG. 2 is a diagram showing an example of the contents of a management table 22.

FIG. 3 is a flowchart illustrating a processing example of a management unit 21;

FIG. 4 is a diagram showing a management unit 2 for creating an empty space in the temporary storage area 23.
6 is a flowchart illustrating a processing example of No. 1;

5 is a block diagram of a real施例of the present invention.

FIG. 6 is a diagram showing an example of the contents of a management table 22a.

FIG. 7 shows management means 2 for creating a free space in temporary storage area 23.
It is a flowchart which shows the example of a process of 1a.

FIG. 8 is a flowchart showing a processing example of a management unit 21a when newly saving management information in a management table 22a.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Operation device 2, 2a ... Temporary storage device 21, 21a ... Management means 22, 22a ... Management table 23 ... Temporary storage area 24 ... Reference cost storage part 3 ... Storage device 4 ... Information processing device 5, 5a ... Information processing device 6, 6a ... recording medium

Claims (3)

(57) [Claims] 1. A method of managing a temporary storage area in which data is temporarily stored, wherein, for each data stored in the temporary storage area, a storage time when the data is first stored in the temporary storage area; Reference count and when accessing the original storage area
Manages the reference cost indicating the interval, and when it is necessary to make room in the temporary storage area, from among the data stored in the temporary storage area, the managed storage time and reference count The estimated value of the reference interval determined by considering the
A method for managing a temporary storage area, characterized by selecting data having a large value at the time of conversion , and vacating a portion where the selected data is stored. 2. A temporary storage area management device for managing a temporary storage area in which data is temporarily stored, wherein each data stored in the temporary storage area is first stored in the temporary storage area. Time of storage, reference count, and access to the original storage area
A management table for managing the reference cost indicating the interval, and when it is necessary to make room in the temporary storage area, from the data stored in the temporary storage area, The estimated value of the reference interval, which is determined by considering the
And a management unit for selecting data having a large value when normalized by the data, and vacating a portion where the selected data is stored. 3. A machine-readable recording medium on which a program for managing a temporary storage area in which data is temporarily stored by a computer is recorded, wherein the computer stores each of the programs stored in the temporary storage area. For each of the data, the storage time when the data was first stored in the temporary storage area, the number of references, and the time when the original storage area was accessed.
A process of managing a reference cost indicating a time interval, when it is necessary to make a space in the temporary storage area, from the data stored in the temporary storage area, the managed storage time and reference The estimated value of the reference interval determined by considering the number of times is normalized by the reference cost
A process for selecting data having a large value when the data is converted , and making a portion where the selected data is stored empty, and a program for recording the program for executing the process.