JP2010198258A - Cache maintenance device, method and program thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a cache maintenance technology for suppressing increase in processing loads without deteriorating a reference speed even in an environment where update processing is highly frequently performed on a database. <P>SOLUTION: A cache maintenance device 1 includes: an update information storage part 11 for storing cache entry update information; a maintenance condition storage part 21 for storing maintenance conditions; a maintenance condition generation part 31 for generating maintenance conditions and writing them into a maintenance condition storage part 21; a maintenance object selection par 41 for selecting a cache entry 1CE becoming the object of maintenance; a maintenance determination part 51 for determining whether or not maintenance conditions are satisfied; and a cache maintenance part 61 for performing the maintenance of the cache entry 1CE. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a cache maintenance technique for maintaining cache entries in a distributed memory cache that stores data from a database in a cache constituted by one or more computer memories.

  Today, with the explosion of the Internet, many services are provided through the Internet. Examples of services provided through the Internet include online shopping, online banking, and social networking service (SNS). In most cases, a system that provides these services includes a database (DB) that stores data such as user account information, product information, and posted articles by the user, and an application. This application issues a query to the database in accordance with the service, and creates a result to be presented to the user by using data returned from the database.

  In these systems, when the number of accesses to the database greatly increases due to an increase in the number of users, the database bandwidth and CPU resources may be exhausted, resulting in performance degradation and server down. In these systems, even if the number of accesses does not increase, if the query to the database is complicated, it takes a long time to return the query result.

  Therefore, as a technique to solve the above-mentioned database processing load and reference performance problems, data created by using this database query result in a storage area (cache) that can be accessed at high speed separately from the database. Is being stored. A system using this cache can be composed of, for example, a relational database system (RDBMS) excellent in persistence and query processing for a database, and a distributed memory cache excellent in responsiveness and scalability.

  Here, when acquiring data from the cache, the application checks whether the necessary data exists on the cache. If the data does not exist in the cache, the data associated with the request from the user is stored in the cache at the same time as obtaining the data from the database. Thereafter, when the same data is required, the application reads this data from the cache and uses it. This eliminates the need to access the database, so that a system using a cache can make it possible to refer to data at high speed and reduce the number of accesses to the database to reduce the load on the database.

  On the other hand, when the cache is used, if update processing (registration, update, deletion) occurs in the database, there is a possibility that inconsistency occurs between the data stored in the cache and the data stored in the database. For example, when referring to data updated in the database in the cache, the data stored in the cache is not updated, and thus the application refers to old data before update.

  Conventionally, in order to solve this problem, a cache maintenance technique for maintaining cache consistency when a database is updated has been proposed.

  As this technique, one that invalidates a cache has been proposed (see, for example, Non-Patent Document 1). In the invalidation method described in Non-Patent Document 1, when an update that causes inconsistency in the cache is performed on the database, the database transmits an invalidation message to the cache, and the cache is not valid. Data in which matching occurs is invalidated, and data before update stored in the cache is prevented from being referred to.

  In addition, a re-query method has also been proposed in which all data stored in the cache is replaced and kept up-to-date by making a query to the database again after the invalidation (see, for example, Non-Patent Document 2 and Non-Patent Document 5). ).

  Also, in an environment where the data stored in the cache is reusable, instead of replacing all the data stored in the cache by the re-query method, only the updated difference is calculated and this difference is stored in the cache. There has also been proposed a differential update method that reflects and maintains data that is already stored in the cache (see, for example, Non-Patent Document 3 and Non-Patent Document 4).

<Conventional cache maintenance technology: first example>
Here, a first example of the conventional cache maintenance technique will be described with reference to FIG. As shown in FIG. 24, the cache maintenance system 900 includes a cache maintenance device 9A, a database 9B, and a cache 9C. Here, the cache maintenance system 900 connects the cache maintenance device 9A, the database 9B, and the cache 9C via a network N such as the Internet.

  The cache maintenance device 9A performs maintenance of the cache 9C by the invalidation method or the re-inquiry method when inconsistency occurs in the data stored in the cache 9C. The cache maintenance device 9A includes a maintenance target selection unit 91 and a cache maintenance unit 92.

  The database 9B is a general database such as a relational database that stores data to be cached 9C. The database 9B stores various data according to the type of application.

  The cache 9C is a distributed memory cache composed of one or more computer memories. Note that illustration of each computer constituting the cache 9C is omitted.

  Hereinafter, the maintenance of the cache 9C will be described in detail. When the update process is performed on the database 9B, the maintenance target selection unit 91 is affected by the update process of the database and requires maintenance based on the contents of the update. ) Is selected. Here, the maintenance target selection unit 91 uses, for example, a method of preparing a mapping table that makes it possible to search for a maintenance target based on a part or all of an update inquiry to the database 9B. In addition, the maintenance target selection unit 91 uses, for example, a method of preparing an inquiry sentence in advance for searching for a maintenance target by inputting a part or all of an update inquiry to the database 9B. Furthermore, the maintenance target selection unit 91 may use a method in which all cache entries created based on a table that has undergone update processing are subject to maintenance, or may use the technique described in Non-Patent Document 3 described above. good.

  The cache maintenance unit 92 transmits an invalidation message for invalidating the maintenance target cache entry selected by the maintenance target selection unit 91. Further, when the re-inquiry method is adopted, the cache maintenance unit 92 issues an inquiry for acquiring the data of the cache entry to the database 9B after transmitting the invalidation message, and the inquiry result is transmitted to the cache 2C. To store. In this way, the cache maintenance system 900 performs cache maintenance. The details of the invalidation method and the re-inquiry method are described in Non-Patent Documents 1, 2, 3, and 5 described above.

<Conventional cache maintenance technology: second example>
Further, a second example of the conventional cache maintenance technique will be described with reference to FIG. 25 (see FIG. 24 as appropriate). In FIG. 25, the cache maintenance device 9A performs maintenance of the cache 9C by the differential update method when inconsistency occurs in the data stored in the cache 9C.

  As illustrated in FIG. 25, the database 9B stores a table 9T1 called “message” and a table 9T2 called “follower”. Further, the cache 9C stores a cache entry 9CE “timeline” that is a result of joining (JOIN) the table 9T1 and the table 9T2.

  Here, when a new record (id = 6 in FIG. 25) is registered (INSERT) in the table 9T1, an inquiry 9Q for obtaining difference data is performed based on this record. As a result, the cache entry 9CE is updated based on the obtained difference data 9D based on the key. Thus, in the conventional differential update method, it is necessary to perform this cache update process every time an update process occurs for the database 9B.

Greg Barish, Katia obraczka, “World Wide Web Caching: Trends and techniques”, IEEE Communications Magazine 2000.5 James Gwertzman, Margo Seltzer, “World Wide Web Cache Consistency”, USENIX technical Conference, 1996 Ashish Gupta, lnderpal Singh Mumick, “Maintenance of Materialized Views: Problems, technologies, and Applications”, IEEE Data Engineering Bulletin, 1995 Jingren Zhou, Per Ake Larson, Hicham G. Elmongui, “Lazy maintenance of materialized views”, VLDB Conference, 2007 Cal Henderson, “Scalable Web Site”, O'Reilly Japan, 2006

  However, in the invalidation method described above, when update processing occurs frequently in the database, cache invalidation occurs frequently, so that the cache hit rate is greatly reduced, and the cache effect (improvement of reference speed and processing) There is a problem that sufficient load reduction cannot be obtained.

  Further, in the above-described re-query method and differential update method, if update processing occurs frequently in the database, the database is inquired every time during cache maintenance. Increase, leading to performance degradation and server down.

  Accordingly, an object of the present invention is to provide a cache maintenance technique that minimizes an increase in processing load without reducing the reference speed even in an environment where update processing is performed on a database with high frequency.

  In order to solve the above-described problem, a cache maintenance apparatus according to the first invention of the present application stores the data in the cache in a distributed memory cache that stores data from a database in a cache constituted by one or more computer memories. In a cache maintenance device that maintains a cache entry as a unit to perform maintenance, a maintenance condition storage unit that stores a maintenance condition that is a condition for maintaining the cache entry, a maintenance condition stored in the maintenance condition storage unit, and the maintenance A maintenance determination unit that determines that maintenance is possible when a condition is satisfied, and a cache member that maintains the cache entry when the maintenance determination unit determines that maintenance is possible. Characterized in that it comprises a nonce portion.

  According to such a configuration, when a mismatch occurs between the cache entry and the data in the database due to the database update process, the cache maintenance device does not always maintain the cache entry but satisfies a predetermined maintenance condition. Only in this case, it was decided to maintain the cache entry. As a result, the cache maintenance device can minimize the increase in processing load without reducing the reference speed even in an environment where update processing is frequently performed on the database.

  Note that the distributed memory cache is a technique that handles the memory of one or more computers in cooperation with each other as a huge logical memory (cache). Instead of providing only simple access using a pair of key and value, hash distribution is used. It is characterized by having high scalability.

  The cache entry is a unit for storing data from the database in the cache. For example, a cache entry is stored in the cache in association with the key.

  Note that maintenance means that when an inconsistency occurs between the data and the cache entry corresponding to this data due to database update processing, this cache entry is invalidated, re-inquired, or differentially updated. Or invalidate this data.

  The invalidation method is to invalidate the cache entry when inconsistency occurs between this data and the cache entry corresponding to this data by updating the database, and the inconsistent data is transferred from the cache entry. This method prevents reading.

  The re-inquiry method is to perform an inquiry corresponding to the cache entry to the database when inconsistency occurs between this data and the cache entry corresponding to the data due to the database update process. In this method, the cache entry is overwritten with the cache result so that the cache entry and the data are in a consistent state.

  Further, the cache maintenance device according to the second aspect of the present invention stores update information including at least one of a previous maintenance time and an update count indicating the number of times the data has been updated since the previous maintenance time. A storage unit, and a maintenance condition generation unit that receives at least one of a preset allowable delay number and an allowable delay time, generates the maintenance condition based on the input, and writes the maintenance condition in the maintenance condition storage unit; The maintenance condition generation unit generates the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count when only the allowable delay count is input, and includes only the allowable delay time. Is entered, it shows the time from the previous maintenance time to the current time. When the maintenance condition that enables the maintenance is generated when time exceeds the allowable delay time, and the allowable delay count and the allowable delay time are input, the update count exceeds the allowable delay count The maintenance condition that enables the maintenance is generated when the time or the elapsed time exceeds the allowable delay time.

  According to such a configuration, the cache maintenance device performs the maintenance of the cache entry only when the maintenance condition is satisfied, and controls the maintenance frequency. As a result, when the cache maintenance apparatus adopts the invalidation method, the frequency at which the cache entry is invalidated by maintenance is reduced, so that the cache hit rate is greatly improved. Further, when the re-inquiry method is adopted, the cache maintenance device can significantly reduce the maintenance cost because the processing load for determining the maintenance condition is sufficiently smaller than the maintenance cost that can be reduced by reducing the frequency of maintenance.

  The number of updates (maintenance count) indicates the number of times data has been updated since the previous maintenance time for each cache entry.

  Note that the allowable number of delays means that maintenance is performed once when the number of inconsistencies specified by the allowable number of delays occurs between this data and the cache entry corresponding to this data due to the database update process. A numerical value indicating what to do.

  The allowable delay time refers to the maintenance of cache entries that have exceeded the allowable delay time from the previous maintenance time when a mismatch occurs between this data and the cache entry corresponding to this data due to the database update process. It is time to show that.

  Note that the maintenance cost is an arithmetic processing amount such as CPU time spent for maintenance in the database server.

  In the cache maintenance device according to the third aspect of the present invention, at least one of a preset allowable delay count and an allowable delay time is input, and based on the input, the maintenance condition is generated and stored in the maintenance condition storage unit. A maintenance condition generation unit for writing, and when only the allowable delay number is input, the maintenance condition generation unit generates the maintenance condition that enables the maintenance when the number of updates exceeds the allowable delay number When only the allowable delay time is input, the maintenance condition that enables the maintenance when the elapsed time exceeds the allowable delay time is generated, and the allowable delay count and the allowable delay time are input. When the update count exceeds the allowable delay count or the elapsed time The maintenance condition that enables maintenance when a delay time is exceeded is generated, and the maintenance determination unit refers to a difference table that stores difference data in the database table and updates the number of data in the difference table And calculating from the time when the oldest data in the difference table is stored to the current time as the elapsed time, and determining whether the maintenance condition is satisfied using the calculated update count and the elapsed time. The cache maintenance unit collectively updates the cache entry with the difference data stored in the difference table, and deletes the updated difference data from the difference table.

  According to such a configuration, the cache maintenance device updates the cache entry with the difference data only when the maintenance condition is satisfied, and controls the maintenance frequency. As a result, when the differential maintenance method is adopted, the cache maintenance device greatly reduces the maintenance cost because the overhead of inquiry processing and communication processing is significantly less than when incremental data is sequentially updated to the cache entry by maintenance. Can be reduced.

  The difference update method is the difference required to update the cache entry to a consistent state when inconsistency occurs between this data and the cache entry corresponding to this data due to the database update process. In this method, the data is calculated and the difference data is reflected in the cache entry so that the cache entry and the data are in a consistent state.

Further, the cache maintenance device according to the fourth invention of the present application stores update information that stores update information including at least one of a previous maintenance time and an update count indicating the number of times the data has been updated since the previous maintenance time. A storage unit, at least one of a preset allowable delay number and allowable delay time, and a maintenance load target load value are input, and based on the input, the maintenance condition is generated for each cache entry, and the maintenance condition is generated. A maintenance condition generation unit that writes to the maintenance condition storage unit, and a statistical information calculation unit that calculates, as statistical information, a cumulative value of the number of maintenance times indicating the number of times the cache entry is maintained in a preset statistical period, and The maintenance condition generation unit When the reduction target value is calculated by multiplying the cumulative load value by the target load value and only the allowable delay count is input, the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count And the reduction value calculated by the expression (1) is subtracted from the reduction target value and only the allowable delay time is input, the elapsed time indicating the previous maintenance time to the current time is the allowable delay. The maintenance condition that enables the maintenance when the time is exceeded is generated, the reduction value calculated by Equation (2) is subtracted from the reduction target value, and the allowable delay count and the allowable delay time are input. When the update count exceeds the allowable delay count or when the elapsed time exceeds the allowable delay time, the maintenance condition is enabled. When the value obtained by subtracting the reduction value calculated by Equation (3) from the reduction target value becomes 0 or less, the generation of the maintenance condition is terminated. It is characterized by doing.
A = Mi × (N−1) / N (1)
A = Mi × (1−T / S) (2)
A = MIN [{Mi × (N−1) / N}, {Mi × (1-T / S)}] (3)
In Equations (1) to (3), A is a reduction value, Mi is a cumulative value of the number of maintenance times, N is an allowable delay number, T is an allowable delay time, and S is a statistical period.

  According to such a configuration, the cache maintenance device performs the maintenance of the cache entry only when the maintenance condition is satisfied, and controls the maintenance frequency. As a result, when the cache maintenance apparatus adopts the invalidation method, the frequency at which the cache entry is invalidated by maintenance is reduced, so that the cache hit rate is greatly improved. Further, when the re-inquiry method is adopted, the cache maintenance device can significantly reduce the maintenance cost because the processing load for determining the maintenance condition is sufficiently smaller than the maintenance cost that can be reduced by reducing the frequency of maintenance. Furthermore, the cache maintenance device can further reduce the processing load without performing unnecessary maintenance after achieving the target load.

  The target load value is a numerical value that is a target for reducing the maintenance cost.

  Further, in the cache maintenance device according to the fifth invention of the present application, the maintenance condition generation unit does not generate the maintenance condition, and the maintenance condition for the cache entry in which the cumulative value of the number of maintenance times is maximized. Is generated.

  According to such a configuration, the cache maintenance device selects cache entries in descending order of the cumulative number of maintenance times, and performs cache entry maintenance in this order. Here, the effect of reducing the maintenance frequency for cache entries with higher maintenance frequency (for example, improving the cache hit rate when the invalidation method is used, and reducing the processing load when the requery method is used). growing. Therefore, the cache maintenance device sets the maintenance conditions and minimizes the number of cache entries to be maintained, thereby minimizing the deterioration in the freshness of the cache entries.

  Note that the freshness is a ratio at which data read from a cache entry is the latest, that is, a ratio of data in which inconsistency does not occur. The data read from the cache entry is the latest when the data obtained as a result of the inquiry to the database matches the data read from the cache entry.

  Further, in the cache maintenance device according to the sixth aspect of the present invention, the statistical information calculation unit calculates, as the statistical information, an accumulated value of the number of times the cache entry is referred to in the statistical period, and the maintenance condition generation unit Is characterized in that the maintenance condition is generated for the cache entry in which the maintenance condition is not generated and the cumulative value of the reference count is minimized.

  According to such a configuration, the cache maintenance device selects cache entries in ascending order of the cumulative number of reference times, and performs cache entry maintenance in this order. As a result, the cache maintenance device performs maintenance in order from the cache entries that are less adversely affected by the maintenance, thereby minimizing the deterioration in the freshness of the cache entries.

  Further, the cache maintenance device according to the seventh invention of the present application stores update information including at least one of a previous maintenance time and an update count indicating the number of times the data has been updated since the previous maintenance time. A storage unit, a maintenance condition generation unit that receives at least one of a preset allowable delay count and an allowable delay time, generates the maintenance condition based on the input, and writes the maintenance condition in the maintenance condition storage unit; and the database A database load measurement unit that measures a load value indicating the load of the maintenance condition generation unit, when only the allowable delay count is input, the update count exceeds the allowable delay count, and The maintenance is possible when the load value satisfies a preset load condition. When the maintenance condition is generated and only the allowable delay time is input, the elapsed time from the previous maintenance time to the current time exceeds the allowable delay time, and the load value satisfies the load condition When the maintenance condition that enables the maintenance is generated and the allowable delay count and the allowable delay time are input, the update count exceeds the allowable delay count or the elapsed time exceeds the allowable delay time. When the load value exceeds the load condition, the maintenance condition that enables the maintenance is generated.

  According to such a configuration, the cache maintenance device performs the maintenance of the cache entry only when the maintenance condition is satisfied, and controls the maintenance frequency. As a result, when the cache maintenance apparatus adopts the invalidation method, the frequency at which the cache entry is invalidated by maintenance is reduced, so that the cache hit rate is greatly improved. Further, when the re-inquiry method is adopted, the cache maintenance device can significantly reduce the maintenance cost because the processing load for determining the maintenance condition is sufficiently smaller than the maintenance cost that can be reduced by reducing the frequency of maintenance. Furthermore, the cache maintenance device can control the frequency of maintenance according to the load of the database by setting the load condition. Thus, for example, the cache maintenance device can prevent the situation where the CPU resource is saved by reducing the frequency of maintenance when the CPU resource of the database server is sufficient, and the database can sufficiently exhibit the performance. .

  The cache maintenance method according to the eighth aspect of the present invention is a cache entry as a unit for storing the data in the cache in a distributed memory cache that stores data from a database in a cache including one or more computer memories. In the cache maintenance method for maintaining the cache entry, a maintenance determination step that determines that maintenance is possible when the maintenance condition is satisfied is referred to with reference to a maintenance condition that is a condition for maintaining the cache entry, and maintenance is determined in the maintenance determination step A cache maintenance step for maintaining the cache entry.

  According to such a configuration, the cache maintenance method does not always maintain the cache entry when inconsistency occurs between the data stored in the cache entry by the database update process and the data stored in the database. Only when a predetermined maintenance condition is satisfied, the cache entry is maintained. As a result, the cache maintenance method can minimize the increase in processing load without reducing the reference speed even in an environment where update processing is frequently performed on the database.

  A cache maintenance program according to the ninth invention of the present application causes a computer to function as each means constituting the cache maintenance device according to any of the first to seventh inventions of the present application.

  According to such a configuration, the cache maintenance program does not always maintain the cache entry when inconsistency occurs between the data stored in the cache entry by the database update process and the data stored in the database. Only when a predetermined maintenance condition is satisfied, the cache entry is maintained. As a result, the cache maintenance program can minimize the increase in processing load without reducing the reference speed even in an environment where update processing is frequently performed on the database.

  According to the present invention, the cache entry is maintained only when a predetermined maintenance condition is satisfied. Therefore, even in an environment where update processing is frequently performed on the database, the reference speed does not decrease and the processing load increases. Can be minimized.

1 is a block diagram illustrating a configuration of a cache maintenance device according to a first embodiment of the present invention. It is a figure which shows the data structure of the cache entry update information which the update information storage part of FIG. 1 memorize | stores. It is a figure which shows the data structure of the maintenance conditions which the maintenance condition memory | storage part of FIG. 1 memorize | stores. It is a flowchart which shows operation | movement of the cache maintenance apparatus of FIG. It is a flowchart which shows the detail of step S2 of FIG. It is a flowchart which shows the detail of step S23 of FIG. It is a figure showing the example which realized the cash maintenance device concerning a 1st embodiment of the present invention with the computer. It is a block diagram which shows the structure of the cash maintenance apparatus which concerns on 2nd Embodiment of this invention. It is explanatory drawing explaining the maintenance by the cache difference maintenance part of FIG. 9 is a flowchart showing details of step S2 of FIG. 4 in the cache maintenance device of FIG. It is a flowchart which shows the detail of step S32 of FIG. It is a block diagram which shows the structure of the cash maintenance apparatus which concerns on 3rd Embodiment of this invention. It is a figure which shows the data structure of the cache entry update information which the update information storage part of FIG. 12 memorize | stores. It is a figure which shows the data structure of the maintenance conditions which the maintenance condition memory | storage part of FIG. 12 memorize | stores. 14 is a flowchart showing details of step S1 in FIG. 4 in the cache maintenance device in FIG. It is a flowchart which shows the detail of step S40 of FIG. It is a flowchart which shows the 1st example of step S43 of FIG. It is a flowchart which shows the 2nd example of step S43 of FIG. It is a block diagram which shows the structure of the cache maintenance apparatus which concerns on 4th Embodiment of this invention. It is a figure which shows the data structure of the maintenance conditions which the maintenance condition memory | storage part of FIG. 18 memorize | stores. 19 is a flowchart showing details of step S1 in FIG. 4 in the cache maintenance device in FIG. 19 is a flowchart showing details of step S23 in FIG. 5 in the cache maintenance device in FIG. It is a figure which shows the relationship between the frequency | count of permissible delay and the limit value of update throughput in an Example. It is explanatory drawing explaining the 1st example of the conventional cache maintenance technique. It is explanatory drawing explaining the 2nd example of the conventional cache maintenance technique.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings as appropriate. In each embodiment, means having the same function and the same configuration are denoted by the same reference numerals and description thereof is omitted.

(First embodiment: invalidation method or re-inquiry method)
[Outline of cache maintenance system]
An outline of the cache maintenance system 100 will be described with reference to FIG. As shown in FIG. 1, the cache maintenance system 100 includes a cache maintenance apparatus 1, a database 1B, and a cache 1C. Here, the cache maintenance system 100 connects the cache maintenance device 1, the database 1B, and the cache 1C via a network N such as the Internet.

  The cache maintenance device 1 maintains a cache entry 1CE in a distributed memory cache that stores data stored in the database 1B in the cache 1C. Here, an invalidation method or a requery method is performed. The configuration of the cache maintenance device 1 will be described later.

  The database 1B is a general database such as a relational database that stores data to be cached 1C.

The cache 1C is a distributed memory cache composed of one or more computer memories. Here, the cache 1C has one or more cache entries 1CE ₁ , 1CE ₂ ... 1CE _n (where n is an integer of 1 or more).

  The cache entry 1CE stores data in the database 1B and provides data to the application. Here, the data stored in the cache entry 1CE and the database 1B may be stored as they are, or the data stored in the database 1B may be processed (for example, processed into HTML format for a Web application) and stored. good.

  When a mismatch occurs between the data stored in the database 1B and the data stored in the cache entry 1CE, the cache maintenance device 1 performs maintenance of the cache entry 1CE when a maintenance condition described later is satisfied.

[Configuration of cache maintenance device]
Hereinafter, the configuration of the cache maintenance device 1 will be described with reference to FIG. As shown in FIG. 1, the cache maintenance device 1 includes an update information storage unit 11, a maintenance condition storage unit 21, a maintenance condition generation unit 31, a maintenance target selection unit 41, a maintenance determination unit 51, and a cache maintenance unit. 61.

  The update information storage unit 11 is a storage unit such as a hard disk that stores cache entry update information (update information). This cache entry update information includes at least one of a previous maintenance time and an update count (maintenance count) indicating the number of times data has been updated since the previous maintenance time. Here, as shown in FIG. 2, the cache entry update information includes a cache entry identifier that uniquely identifies the cache entry 1CE, a maintenance count that indicates the number of times the data has been updated since the previous maintenance time, , And the previous maintenance time which is the time when the maintenance was performed last time.

  The maintenance condition storage unit 21 in FIG. 1 is a storage unit such as a hard disk that stores maintenance conditions. This maintenance condition indicates a condition for maintaining the cache entry 1CE.

  The maintenance condition generation unit 31 receives at least one of a preset allowable delay number and an allowable delay time, generates a maintenance condition based on the input, and writes the maintenance condition in the maintenance condition storage unit 21. Here, the maintenance condition generation unit 31 reads at least one of the allowable delay number and the allowable delay time from the setting file, or uses an input unit such as a keyboard or a mouse (not shown) to at least the allowable delay number and the allowable delay time. One is entered.

  And the maintenance condition production | generation part 31 produces | generates a maintenance condition. Specifically, when only the allowable delay count is input, the maintenance condition generation unit 31 generates a maintenance condition that enables maintenance when the maintenance count exceeds the allowable delay count. For example, when the allowable delay count “10 times” is input, the maintenance condition generation unit 31 generates a maintenance condition “maintenance count> 10 times” as illustrated in FIG.

  In addition, when only the allowable delay time is input, the maintenance condition generating unit 31 generates a maintenance condition that enables maintenance when the elapsed time exceeds the allowable delay time. For example, when the allowable delay time “10 minutes” is input, the maintenance condition generation unit 31 generates a maintenance condition “elapsed time> 10 minutes” as illustrated in FIG. The elapsed time is the time from the previous maintenance time to the current time. Further, as the current time, for example, the time of the internal clock of the computer or the time acquired by NTP (Network Time Protocol) can be used.

  In addition, when both the allowable delay count and the allowable delay time are input, the maintenance condition generation unit 31 can perform maintenance when the maintenance count exceeds the allowable delay count or when the elapsed time exceeds the allowable delay time. A maintenance condition is generated. For example, when the allowable delay number “20 times” and the allowable delay time “10 minutes” are input, the maintenance condition generation unit 31 performs “(maintenance count> 20 times) OR” as shown in FIG. (Elapsed time> 10 minutes) ”is generated. Note that “OR” in FIG. 3C indicates an OR condition. The maintenance condition generation unit 31 generates one maintenance condition that is common to all cache entries 1CE.

  The maintenance target selection unit 41 in FIG. 1 selects a cache entry 1CE that is a maintenance target from among all the cache entries 1CE of the cache 1C. Here, the maintenance target selection unit 41 may select the cache entry 1CE and update the cache entry update information. For example, the maintenance target selection unit 41 increments the maintenance count and updates the previous maintenance time to the current time. At this time, the maintenance target selection unit 41 uses a method of preparing a mapping table that makes it possible to search for a maintenance target based on part or all of an update inquiry (registration, update, deletion) to the database 1B. Further, for example, the maintenance target selection unit 41 has a method of preparing an inquiry sentence in advance for searching for a maintenance target by inputting a part or all of an update inquiry to the database 1B. Furthermore, the maintenance target selection unit 41 may use a method in which all cache entries 1CE created based on the table that has undergone the update process are subject to maintenance, using the technique described in Non-Patent Document 3 described above. Also good.

  The maintenance determination unit 51 refers to the maintenance conditions stored in the maintenance condition storage unit 21 and determines that maintenance is possible when the maintenance conditions are satisfied. Here, when the cache entry update information stored in the update information storage unit 11 is updated, the maintenance determination unit 51 determines the maintenance count corresponding to the cache entry 1CE selected by the maintenance target selection unit 41 and the previous maintenance time. Read from the update information storage unit 11. Then, the maintenance determination unit 51 assigns at least one of the read maintenance count and the previous maintenance time to the maintenance condition, and determines whether the maintenance condition is satisfied.

  The cache maintenance unit 61 maintains the cache entry 1CE when the maintenance determination unit 51 determines that maintenance is possible. When the invalidation method is employed, the cache maintenance unit 61 issues an invalidation message indicating that the cache entry 1CE selected by the maintenance target selection unit 41 is invalidated when the maintenance determination unit 51 determines that maintenance is possible. Send to cache 1C.

  When the re-inquiry method is adopted, the cache maintenance unit 61 transmits an invalidation message and then makes an inquiry to the database 1B for acquiring data to be stored in the cache entry 1CE. Then, the cache maintenance unit 61 stores the inquiry result in the cache 1C as the cache entry 1CE. The details of the invalidation method and the re-inquiry method are described in Non-Patent Documents 1, 2, 3, and 5 described above.

[Operation of the cache maintenance device]
<Overall operation>
Hereinafter, the overall operation of the cache maintenance apparatus 1 will be described with reference to FIG. 4 (see FIG. 1 as appropriate). First, as shown in FIG. 4, in the cache maintenance device 1, at least one of the allowable delay number and the allowable delay time is input by the maintenance condition generation unit 31, and the maintenance condition is generated based on the input. Write to the storage unit 21 (step S1).

  Following the processing in step S1, the cache maintenance device 1 performs maintenance on the cache entry 1CE (step S2). Details of step S2 will be described later.

  Following the processing of step S2, the cache maintenance device 1 determines whether or not the maintenance is to be ended by the maintenance determination unit 51 (step S3). For example, when an end command is input from the user to the input unit, the cache maintenance device 1 determines that the maintenance is ended by the maintenance determination unit 51.

  If it is determined that the maintenance is to be ended (Yes in step S3), the cache maintenance device 1 ends the process. On the other hand, when it is determined that the maintenance is not finished (No in step S3), the cache maintenance device 1 returns to the process of step S2.

<Details of Step S2: Maintenance>
The details of step S2 (maintenance) in FIG. 4 will be described below with reference to FIG. First, in the cache maintenance device 1, the maintenance target selection unit 41 determines a cache entry 1CE that is a maintenance target among all the cache entries 1CE of the cache 1C (step S20). Specifically, the cache maintenance device 1 outputs a list (unprocessed list) of unprocessed cache entry identification values that are not maintained by the maintenance target selection unit 41.

  Following the processing in step S20, the cache maintenance device 1 arbitrarily selects the cache entry 1CE to be maintained based on the unprocessed list output in step S20 by the maintenance target selection unit 41 (step S21). For example, the cache maintenance device 1 sorts the cache entry identifiers included in the unprocessed list by the maintenance target selection unit 41, and selects the cache entries 1CE in the sort order. In addition, for example, the cache maintenance device 1 uses the maintenance target selection unit 41 to randomly select the cache entry 1CE included in this unprocessed list. Then, the cache maintenance device 1 outputs a cache entry identifier indicating the selected cache entry 1CE by the maintenance target selection unit 41.

  Following the processing in step S21, the cache maintenance device 1 updates (increments) the maintenance count of the cache entry update information for the cache entry 1CE corresponding to the cache entry identifier output in step S20 by the maintenance target selection unit 41. (Step S22).

  Following the processing of step S22, the cache maintenance device 1 evaluates the maintenance conditions by the maintenance determination unit 51 (step S23). Then, the cache maintenance device 1 outputs a truth value corresponding to the evaluation result by the maintenance determination unit 51. Details of step S23 will be described later.

  Following the processing of step S23, the cache maintenance device 1 determines whether the maintenance condition is satisfied by the maintenance determination unit 51 (step S24). For example, when the true / false value output in step S23 is “true” by the maintenance determination unit 51, the cache maintenance device 1 determines that the maintenance condition is satisfied and maintenance is possible. Further, for example, the cache maintenance device 1 determines that maintenance is not possible because the maintenance determination unit 51 does not satisfy the maintenance condition when the true / false value output in step S23 is “false”.

  If it is determined that maintenance is possible (Yes in step 24), the cache maintenance device 1 proceeds to the process of step S25. On the other hand, when it is determined that maintenance is not possible (No in step S24), the cache maintenance device 1 proceeds to the process of step S27.

  Following the processing in step 24, the cache maintenance device 1 performs maintenance of the cache entry 1CE corresponding to the cache entry identifier output in step S20 by the cache maintenance unit 61 (step S25).

  Following the processing in step 25, the cache maintenance device 1 updates the cache entry update information for the cache entry 1CE corresponding to the cache entry identifier output in step S20 by the cache maintenance unit 61 (step S26). For example, the cache maintenance device 1 causes the cache maintenance unit 61 to update (reset) the maintenance count of the cache entry update information to “0” and update the previous maintenance time with the current time. In step S26, when the cache entry reference count is stored and used as the cache entry update information, these values may be updated (reset).

  Following the processing in step 26 or No in step S24, the cache maintenance device 1 processes all the cache entries 1CE to be maintained included in the unprocessed list output in step S20 by the cache maintenance unit 61 ( It is determined whether or not maintenance has been performed (step S27).

  If it is determined that all the maintenance targets have been processed (Yes in step S27), the cache maintenance device 1 ends the maintenance. On the other hand, when it is determined that not all the maintenance targets have been processed (No in step S27), the cache maintenance device 1 returns to the process of step S21.

<Details of Step S23: Evaluation of Maintenance Conditions>
The details of step S23 (evaluation of maintenance conditions) in FIG. 5 will be described below with reference to FIG. First, the cache maintenance device 1 acquires the maintenance condition for the cache entry 1CE corresponding to the cache entry identifier output in step S20 by the maintenance determination unit 51 with reference to the maintenance condition storage unit 21, and updates the information storage unit 11 To obtain cache entry update information (step S231).

  Subsequent to the processing of step S231, the cache maintenance device 1 uses the maintenance determination unit 51 to evaluate the maintenance condition by substituting the cache entry update information acquired in step S231 for the maintenance condition acquired in step S231. A true / false value corresponding to is output (step S232).

  As described above, since the cache maintenance device 1 according to the first embodiment performs maintenance of the cache entry 1CE only when the maintenance condition is satisfied, the cache maintenance device 1 is referred even in an environment where update processing is frequently performed on the database 1B. The speed does not decrease, and the increase in processing load can be minimized. Further, when the cache maintenance apparatus 1 adopts the invalidation method, the frequency of invalidating the cache entry 1CE due to maintenance is reduced, so that the cache hit rate is greatly improved. Furthermore, when the re-inquiry method is adopted, the cache maintenance apparatus 1 can significantly reduce the maintenance cost because the processing load for determining the maintenance condition is sufficiently smaller than the maintenance cost that can be reduced by reducing the frequency of maintenance. .

  Although the cache maintenance device 1 has been described as an independent device, the present invention is not limited to this. For example, as shown in FIG. 7, the present invention provides a system bus B that includes a display unit O such as a display, an input unit I such as a keyboard and a mouse, an arithmetic device C, and a storage device H such as a memory and a hard disk. It can be realized with a connected general computer. Specifically, the present invention causes the storage device H of this computer to operate cooperatively as the update information storage unit 11 and the maintenance condition storage unit 21, and the arithmetic device C to operate the maintenance condition generation unit 31, the maintenance target selection unit 41, the maintenance determination. This can be realized by a cache maintenance program that operates in cooperation as the unit 51 and the cache maintenance unit 61. This program may be distributed via the network N, or may be distributed by writing in a recording medium such as a CD-ROM or a flash memory. This also applies to the following embodiments.

(Second embodiment: differential update method)
[Outline of cache maintenance system]
Hereinafter, with reference to FIG. 8, the outline of the cache maintenance system 200 according to the second embodiment will be mainly described in terms of differences from the first embodiment. Here, when performing maintenance, the differential update method in which only a part in the cache entry 2CE is updated is more general than the requery method in which all the data in the inconsistent cache entry 2CE is discarded and then recalculated. Therefore, the calculation amount is reduced. This differential update method is not applicable to any cache 2C and needs to satisfy a certain database environment. The cache maintenance system 200 is an example when such a differential update method is adopted.

  The database 2B is a relational database that stores a difference table ΔT for temporarily storing difference data for each table T in addition to the table T stored from the beginning. This difference table ΔT stores only records registered (INSERT) in the table T. In FIG. 8, one difference table ΔT and one table T are shown, but there may be a plurality of difference tables ΔT and tables T.

  The cache 2C is the same as the cache 1C in FIG. The cache entry 2CE stores the inquiry result of the database 2B as a storage target.

[Configuration of cache maintenance device]
In the following, the configuration of the cache maintenance device 2 according to the second embodiment will be described mainly with respect to differences from the first embodiment. As shown in FIG. 8, the cache maintenance device 2 includes a maintenance condition storage unit 22, a maintenance condition generation unit 32, a maintenance determination unit 52, and a cache differential maintenance unit (cache maintenance unit) 62. The cache difference maintenance unit 62 corresponds to the cache maintenance unit described in the claims.

  The maintenance condition storage unit 22 is a storage unit such as a hard disk that stores maintenance conditions. Whether or not the maintenance determination unit 52 described later updates the cache entry 2CE corresponding to the table T by comparing the information with information such as the number of data stored in the difference table ΔT and the time of the oldest data. This is a conditional expression for determining whether or not. For example, the maintenance condition is a condition such as “An> 10” when the number of difference data in the difference table ΔT is An.

  The maintenance condition generation unit 32 generates a maintenance condition for each table T stored in the database 2B and writes it in the maintenance condition storage unit 24. The maintenance condition generation method of the maintenance condition generation unit 32 is the same as that of the maintenance condition generation unit 31 of FIG.

  The maintenance determination unit 52 refers to the difference table ΔT storing the difference data in the table T of the database 2B, and calculates the number of data items in the difference table ΔT as a maintenance count. Further, the maintenance determination unit 52 calculates the elapsed time from the time when the oldest data in the difference table ΔT is stored to the current time. Then, the maintenance determination unit 52 refers to the maintenance condition stored in the maintenance condition storage unit 22, substitutes the calculated update count and elapsed time into the maintenance condition, and determines that maintenance is possible when this maintenance condition is satisfied. .

  The cache difference maintenance unit 62 updates (maintains) the difference data stored in the difference table ΔT in the cache entry 2CE for the cache entry 2CE that the maintenance determination unit 52 has determined that maintenance is possible. Then, the cache difference maintenance unit 62 deletes the difference data updated together from the difference table ΔT.

  Hereinafter, the maintenance by the cache difference maintenance unit 62 will be specifically described with reference to FIG. 9 (see FIG. 8 as appropriate). As shown in FIG. 9, the database 2B stores a table “message” (not shown), a difference table ΔT “Δmessage”, and a table T “follower”. This difference table ΔT stores four records with ids of 6-9.

  Here, when updating (maintenance) the cache entry 2CE, the cache difference maintenance unit 62 obtains batch difference data for updating a plurality of difference data in one cache entry 2CE in a batch. For example, the cache difference maintenance unit 62 makes an inquiry Q2 to the database 2B in order to obtain the batch difference data. This query Q2 differs from the query 9Q in FIG. 25 in that all records in the difference table ΔT are used and that grouping is performed using the key of the cache entry 2CE. Then, the cache difference maintenance unit 62 updates the batch difference data to the cache entry 2CE for each group corresponding to the key of the cache entry 2CE. That is, the cache difference maintenance unit 62 collectively updates (maintains) one or more pieces of difference data to the cache entry 2CE.

[Operation of the cache maintenance device]
<Details of Step S2: Maintenance>
The cache maintenance device 2 performs the same overall operation as in FIG. Hereinafter, since the operation of step S2 in FIG. 4 in the cache maintenance device 2 is different, this will be described with reference to FIG.

  In advance, the database 2B accumulates updates to the database 2B, that is, generates a difference table (step S30). Here, the database 2B receives a registration (INSERT) query for the database 2B and registers the record described in the query in the table T. At this time, the database 2B also registers this record in the difference table ΔT. In addition, in order to clarify that the process of step S30 is not operation | movement of the cache maintenance apparatus 2, it illustrated in the broken line.

  The processing in step S31 and step S32 is the same as that in steps 23 and S24 in FIG.

  Following the processing of step 32, the cache maintenance device 2 performs batch update (maintenance) of the cache entry 2CE by the cache difference maintenance unit 62 (step S33).

<Details of Step S33: Batch Update of Cache Entries>
Hereinafter, with reference to FIG. 11, the details of step S33 (batch update of cache entries) of FIG. 10 will be described. First, the cache maintenance device 2 obtains batch difference data from the difference table ΔT and the table T by the cache difference maintenance unit 62 (step S331). Here, the cache maintenance device 2 uses the cache difference maintenance unit 62 to make an inquiry Q2 in FIG. 9 to the database 2B to obtain batch difference data for all the cache entries 2CE.

  Following the processing in step S331, the cache maintenance device 2 uses the cache difference maintenance unit 62 to batch update one cache entry 2CE with the batch difference data obtained in step S331 (step S332). Then, the cache maintenance device 2 uses the cache difference maintenance unit 62 to identify the identifier of the table T (for example, the name of the table T) corresponding to the cache entry 2CE that has updated the batch difference data, and the cache entry that has not been batch updated. Unprocessed batch differential data that is 2CE batch differential data is output.

  Following the processing in step S332, the cache maintenance device 2 determines whether or not all the batch difference data has been processed by the cache difference maintenance unit 62 (step S333). Here, the cache maintenance device 2 determines that all the batch difference data has been processed when the cache difference maintenance unit 62 does not output the unprocessed batch difference data in step S332.

  If it is determined that all the batch difference data has been processed (Yes in step S333), the cache maintenance device 2 proceeds to the process of step S334. On the other hand, when it is determined that all the batch difference data has not been processed (No in step S333), the cache maintenance device 2 returns to the process of step S332.

  Following the processing in step S333, the cache maintenance device 2 deletes the batch difference data updated in step S332 from the difference table ΔT by the cache difference maintenance unit 62 (step S334). Here, the cache maintenance device 2 deletes (DELETE) all the records in the table T for all the tables T indicated by the identifier output in step S332 by the cache difference maintenance unit 62.

  As described above, since the cache maintenance device 2 according to the second embodiment performs maintenance of the cache entry 2CE only when the maintenance condition is satisfied, the cache maintenance device 2 is referred even in an environment where update processing is frequently performed on the database 2B. The speed does not decrease, and the increase in processing load can be minimized. In addition, the cache maintenance device 2 updates the cache entry 2CE in batch with the difference data only when the maintenance condition is satisfied, and controls the maintenance frequency. As a result, when the differential maintenance method is adopted, the cache maintenance device 2 has much lower overhead of inquiry processing and communication processing than the case where sequential differential data is updated to the cache entry 2CCE by maintenance. Can be greatly reduced.

(Third embodiment: Invalidation method or re-inquiry method)
[Configuration of cache maintenance device]
Hereinafter, with reference to FIG. 12, the configuration of the cache maintenance device 3 according to the third embodiment will be described mainly with respect to differences from the first embodiment. The cache maintenance device 3 is greatly different from the cache maintenance device 1 in that maintenance is performed until a preset target load is achieved using statistical information.

  Therefore, as shown in FIG. 12, the cache maintenance device 3 includes an update information storage unit 13, a maintenance condition storage unit 23, a maintenance condition generation unit 33, a maintenance target selection unit 43, a maintenance determination unit 53, A cache maintenance unit 63 and a statistical information calculation unit 73 are provided.

  The update information storage unit 13 is a storage unit such as a hard disk that stores cache entry update information. Here, as shown in FIG. 13, the cache entry update information includes a cache entry identifier, a maintenance count, a previous maintenance time, a cumulative reference count, and a cumulative maintenance count. This cumulative reference count is a cumulative value of the reference count for which the cache entry 3CE is referenced in a preset statistical period. The cumulative maintenance count is a cumulative value of the maintenance count indicating the number of times that the cache entry 3CE is maintained in the statistical period. That is, among the data items of the cache entry update information, the cumulative reference count and the cumulative maintenance count become the cache entry statistical information (statistical information). The accumulated reference count and the accumulated maintenance count are written in the cache entry update information by the statistical information calculation unit 73 described later.

  The maintenance condition storage unit 23 in FIG. 12 is a storage unit such as a hard disk that stores maintenance conditions. Here, as shown in FIG. 14, the maintenance condition is generated for each cache entry 3CE.

  The maintenance condition generation unit 33 in FIG. 12 receives at least one of a preset allowable delay count and allowable delay time and a maintenance load target load value, and performs maintenance for each cache entry 3CE based on this input. A condition is generated and written in the maintenance condition storage unit 23. Here, the maintenance condition generation unit 33 receives at least one of the allowable delay count and the allowable delay time for each cache entry 3CE. Further, the maintenance condition generation unit 33 reads the target load value from the setting file, or the target load value is input by the input unit. Here, the target load value is input as a value exceeding 0 and less than 100, for example.

  Then, the maintenance condition generation unit 33 generates a maintenance condition for each cache entry 3CE until the target load is achieved or the maintenance conditions for all the cache entries 3CE are generated. At this time, it is preferable that the maintenance condition generation unit 33 selects the cache entry 3CE in which the maintenance condition generation unit 33 has not generated the maintenance condition and the cumulative value of the number of times of maintenance is maximized. Alternatively, it is preferable that the maintenance target selection unit 43 selects the cache entry 3CE in which the maintenance condition generation unit 33 has not generated the maintenance condition and the cumulative value of the reference count is minimized. Details of generation of maintenance conditions will be described later.

  The maintenance target selection unit 43 selects the cache entry 3CE, similarly to the maintenance target selection unit 41 of FIG.

  As with the maintenance determination unit 51 in FIG. 1, the maintenance determination unit 53 refers to the maintenance condition stored in the maintenance condition storage unit 21 for the cache entry 3CE selected by the maintenance target selection unit 43 and satisfies the maintenance condition. It is determined that maintenance is possible.

  Similarly to the cache maintenance unit 61 in FIG. 1, the cache maintenance unit 63 maintains the cache entry 3CE when the maintenance determination unit 53 determines that maintenance is possible for the cache entry selected by the maintenance target selection unit 43. .

  The statistical information calculation unit 73 calculates a cumulative reference count and a cumulative maintenance count in a preset statistical period. Here, the statistical information calculation unit 73 operates the cache maintenance device 3 for a statistical period, and calculates the cumulative reference count and the cumulative maintenance count. Then, the statistical information calculation unit 73 writes the cumulative reference count and the cumulative maintenance count as cache entry update information in the update information storage unit 13.

[Operation of the cache maintenance device]
The cache maintenance device 3 performs the same overall operation as in FIG. Hereinafter, since the operation of step S1 in FIG. 4 in the cache maintenance device 3 is different, this will be described with reference to FIG.

<Details of Step S1: Generation of Maintenance Conditions>
As shown in FIG. 15, the cache maintenance device 3 acquires (calculates) cache entry statistical information by the statistical information calculation unit 73 (step S40). Then, the cache maintenance device 3 writes the cache entry statistical information in the update information storage unit 13 by the statistical information calculation unit 73. Details of step S40 will be described later.

  Following the processing of step S40, the cache maintenance device 3 acquires (inputs) at least one of the allowable delay count and the allowable delay time and the target load value of the maintenance cost by the maintenance condition generation unit 33 (step S41). .

Following the processing of step S41, the cache maintenance device 3 calculates a reduction target value based on the target load value acquired in step S41 by the maintenance condition generation unit 33 (step S42). Here, the cache maintenance device 3 uses the maintenance condition generation unit 33 to refer to the cache entry statistical information and calculate the reduction target value using Equation (4). In equation (4), Dg represents the reduction target value, ΣMi represents the cumulative maintenance count of all cache entries, and G represents the target load value.
Dg = ΣMi × G / 100 (4)

  Following the processing of step S42, the cache maintenance device 3 selects one cache entry 3CE from the cache entries 3CE that have not generated the maintenance condition by the maintenance condition generation unit 33 (step S43). Then, the cache maintenance device 3 outputs a cache entry identifier indicating one selected cache entry 3CE by the maintenance condition generation unit 33. Details of the processing in step S43 will be described later.

Following the processing in step S43, the cache maintenance device 3 generates maintenance conditions and calculates a reduction value for the cache entry 3CE corresponding to the cache entry identifier output in step S43 by the maintenance condition generation unit 33 ( Step S44). Here, when only the allowable number of delays is input in step S41, the cache maintenance device 3 calculates the reduction value using the formula (1) by the maintenance condition generation unit 33, and is expressed by the formula (5). Generate maintenance conditions.
A = Mi × (N−1) / N (1)
Mc> N (5)

When only the allowable delay time is input in step S41, the cache maintenance device 3 calculates a reduction value using the equation (2) by the maintenance condition generation unit 33, and the maintenance represented by the equation (6). Generate a condition.
A = Mi × (1−T / S) (2)
Pt> T (6)

  When both the allowable delay count and the allowable delay time are input in step S41, the cache maintenance device 3 calculates a reduction value using the equation (3) by the maintenance condition generation unit 33, and the equation (7 ) Is generated.

A = MIN [{Mi × (N−1) / N}, {Mi × (1-T / S)}] (3)
(Mc> N) OR (Pt> T) (7)

  In Equations (1) to (3), A is a reduction value, Mi is a cumulative value of the number of maintenance times, N is an allowable delay number, T is an allowable delay time, and S is a statistical period. Further, in the above formulas (5) to (7), Mc represents a maintenance count, Pt represents an elapsed time, OR represents an OR condition, and MIN is a function that outputs a minimum value.

  Further, the cache maintenance device 3 subtracts the reduction value (A) from the reduction target value (Dg) every time the maintenance condition is generated by the maintenance condition generation unit 33. Then, the cache maintenance device 3 outputs the reduction target value after the subtraction by the maintenance condition generation unit 33.

  Following the processing in step S44, the cache maintenance device 3 determines whether the maintenance condition generation unit 33 has achieved the target load or whether the maintenance conditions for all the cache entries 3CE have been generated (step S45). . Here, the cache maintenance device 3 determines that the target load has been achieved when the maintenance condition generation unit 33 causes the reduction target value after subtraction output in step S44 to be 0 or less.

  If it is determined that the target load has been achieved or the maintenance conditions for all the cache entries 3CE have been generated (Yes in step 45), the cache maintenance device 3 proceeds to the process of step S46. When it is determined that the target load has not been achieved and the maintenance conditions for all the cache entries 3CE have not been generated (No in step 45), the cache maintenance device 3 returns to the process of step S43.

Following the processing of step S45, when there is a cache entry 3CE for which no maintenance condition has been generated by the maintenance condition generating unit 33, the cache maintenance device 3 is expressed by equation (8) for the cache entry 3CE. Maintenance conditions are generated (step S46). Equation (8) indicates that when a mismatch occurs between the data in the database 3B and the corresponding cache entry 3CE, maintenance is always performed regardless of the target load.
Mc> 1 Formula (8)

<Details of Step S40: Acquisition of Cache Entry Statistics Information>
Hereinafter, the details of step S40 (acquisition of cache entry statistical information) in FIG. 15 will be described with reference to FIG. First, the cache maintenance device 3 reads the statistical period from the setting file by the statistical information calculation unit 73 or inputs the statistical period by the input unit (step S401).

  Subsequent to the processing in step S401, the cache maintenance device 3 calculates the reference count and the maintenance count in the statistical period input in step S401 by the statistical information calculation unit 73, and writes it in the cache entry statistical information (step S402). . The number of maintenance times is the number of write requests to the cache entry 3CE in the invalidation method, and in the re-inquiry method and the differential update method.

  Following the processing of step S402, the cache maintenance device 3 determines whether or not the statistical period input in step S401 has ended by the statistical information calculation unit 73 (step S403). If it is determined that the statistical period has ended (Yes in step 403), the cache maintenance device 3 ends the process. On the other hand, when it is determined that the statistical period has not ended (No in step 403), the cache maintenance device 3 returns to the process of step S402.

<Details of Step S43: Selection of Cache Entry: First Example>
Hereinafter, a first example of step S43 (selection of cache entry) in FIG. 15 will be described with reference to FIG. In the cache maintenance device 3, the maintenance condition generator 33 selects a cache entry 3CE that has not generated a maintenance condition and has the maximum accumulated maintenance count (step S431).

  As a result, the cache maintenance device 3 can improve the cache hit rate with the higher maintenance frequency (for example, the cache hit rate is improved when the invalidation method is used, and the processing is performed when the requery method is used). Load reduction). Therefore, the cache maintenance device 3 sets the maintenance conditions and minimizes the number of caches to be maintained, thereby minimizing the deterioration of cache freshness.

<Details of Step S43: Selection of Cache Entry: Second Example>
Hereinafter, a second example of step S43 (selection of cache entry) in FIG. 15 will be described with reference to FIG. In the cache maintenance device 3, the maintenance condition generation unit 33 selects the cache entry 3CE in which the maintenance condition generation unit 33 has not generated the maintenance condition and the cumulative value of the reference count is the minimum (step S432).

  As a result, the cache maintenance device 3 performs maintenance in order from the cache having the least adverse effect of the maintenance, thereby minimizing the deterioration of the cache freshness.

  As described above, the cache maintenance device 3 according to the third embodiment performs maintenance of the cache entry 3CE only when the maintenance condition is satisfied. Therefore, even in an environment where update processing is frequently performed on the database 3B. The speed does not decrease, and the increase in processing load can be minimized. Further, when the cache maintenance device 3 adopts the invalidation method, the frequency of cache invalidation due to maintenance is reduced, so that the cache hit rate is greatly improved. Furthermore, when the re-inquiry method is adopted, the cache maintenance device 3 can significantly reduce the maintenance cost because the processing load for determining the maintenance condition is sufficiently smaller than the maintenance cost that can be reduced by reducing the frequency of maintenance. . Furthermore, after achieving the target load, the cache maintenance device 3 can further reduce the processing load without performing unnecessary maintenance.

(Fourth embodiment: invalidation method or re-inquiry method)
[Configuration of cache maintenance device]
Hereinafter, the configuration of the cache maintenance device 4 according to the fourth embodiment will be described mainly with respect to differences from the first embodiment with reference to FIG. The cache maintenance device 4 is greatly different from the cache maintenance device 1 in that the load information of the database is reflected in the maintenance conditions.

  For this reason, as shown in FIG. 19, the cache maintenance device 4 includes an update information storage unit 14, a maintenance condition storage unit 24, a maintenance condition generation unit 34, a maintenance target selection unit 44, a maintenance determination unit 54, A cache maintenance unit 64 and a database load measurement unit 84 are provided.

  The maintenance condition storage unit 24 is a storage unit such as a hard disk that stores maintenance conditions. Here, as shown in FIG. 20, the maintenance condition is added such that a load condition described later exceeds a predetermined threshold value. The load condition is read from the setting file, or the load condition is input by the input unit.

  The maintenance condition generation unit 34 in FIG. 19 generates maintenance conditions and writes them in the maintenance condition storage unit 24 in the same manner as the maintenance condition generation unit 33 in FIG. At this time, the maintenance condition generating unit 34 adds a condition that the load value measured by the database load measuring unit 80 described later satisfies the load condition to the maintenance condition. For example, when the maintenance condition is “maintenance count> 10 times”, the maintenance condition generation unit 34 adds a load condition such as “(maintenance count> 10 times) AND (CPU usage rate> 90%)” under the AND condition. To do.

  The database load measuring unit 84 measures a load value indicating the load of the database 4B. Here, the database load measurement unit 84 measures load values such as a CPU usage rate, an interface usage rate, a memory usage rate, and a disk I / O value of the server having the database 4B. The load condition indicates a condition that the load value is equal to or greater than a preset threshold value or the load value is equal to or less than this threshold value.

  The update information storage unit 14, the maintenance target selection unit 44, the maintenance determination unit 54, and the cache maintenance unit 64 are the same as the respective units in FIG.

[Operation of the cache maintenance device]
<Details of Step S1: Generation of Maintenance Conditions>
The cache maintenance device 4 performs the same overall operation as in FIG. Hereinafter, since the operation of step S1 in FIG. 4 in the cache maintenance device 4 is different, this will be described with reference to FIG.

  First, the cache maintenance device 4 acquires (inputs) at least one of the allowable delay number and the allowable delay time by the maintenance condition generation unit 34 (step S50). Further, the cache maintenance device 4 generates a maintenance condition by the maintenance condition generation unit 34 (step S51). Then, the cache maintenance device 4 adds the load condition to the maintenance condition as an AND condition by the maintenance condition generation unit 34 (step S52).

<Details of Step S23: Evaluation of Maintenance Conditions>
Hereinafter, since the operation of step S23 in FIG. 5 in the cache maintenance device 4 is different, this will be described with reference to FIG. First, the cache maintenance device 4 uses the maintenance determination unit 54 to acquire maintenance conditions for the cache entry 4CE to be maintained with reference to the maintenance condition storage unit 24, and updates the cache entry with reference to the update information storage unit 14 Information is acquired (step S60).

  Following the processing of step S60, the cache maintenance device 4 measures the load value by the database load measuring unit 84 (step S61).

  Following the processing of step S61, the cache maintenance device 4 uses the maintenance determination unit 54 to substitute the cache entry update information acquired in step S61 and the load value measured in step S62 for the maintenance condition acquired in step S60. The maintenance condition is evaluated, and a true / false value corresponding to the evaluation result is output (step S62).

  As described above, the cache maintenance device 4 according to the fourth embodiment performs maintenance of the cache entry 4CE only when the maintenance condition is satisfied, and controls the maintenance frequency. As a result, when the cache maintenance device 4 adopts the invalidation method, the frequency of cache invalidation due to maintenance is reduced, so that the cache hit rate is greatly improved. In addition, when the re-inquiry method is adopted, the cache maintenance device 4 can significantly reduce the maintenance cost because the processing load for determining the maintenance condition is sufficiently smaller than the maintenance cost that can be reduced by reducing the frequency of maintenance. . Furthermore, the cache maintenance device 4 can control the frequency of maintenance according to the load of the database by setting the load condition. As a result, the cache maintenance device 4 prevents, for example, a situation in which the CPU frequency of the database server is low and the CPU frequency is saved by reducing the frequency of maintenance, and allows the database 4B to perform sufficiently. Can do.

  Although the cache maintenance device 4 has been described as setting maintenance conditions by the same method as in the first embodiment, the present invention is not limited to this. Similarly to the third embodiment, the cache maintenance device 4 may generate maintenance conditions loaded with load conditions until the target load is achieved. In this case, the cache maintenance device 4 may generate the maintenance conditions in the descending order of the maintenance frequency or in ascending order of the reference frequency.

  Examples of the present invention will be described below (see FIGS. 8 and 9 as appropriate). Here, the cache maintenance system 200 shown in FIG. 8 is mounted in the following environment, and the verification result of the cache maintenance device 2 at this time is shown and described.

In this embodiment, the application server is configured by one server described below.
CPU: Xeon (registered trademark) X5470 3.3 GHz × 1
Memory: 32GB Hard disk: 1.5TB (RAID10)
Middleware: ruby 1.8.5, MySQL5.0

In this embodiment, the database server 2B is composed of one server as described below.
CPU: Xeon (registered trademark) X5355 2.66 GHz × 2
Memory: 16 Gbytes Hard disk: 1.5 Tbytes (RAID10)
Middleware: MySQL5.0, libmcached0.23

In this embodiment, the cache 2C is composed of the following 10 servers.
CPU: Celeron (registered trademark) X430 1.8 GHz × 1
Memory: 2 GB Hard disk: 500 GB Middleware: memcached 1.2.6

Also, in this embodiment, crawling is performed for the “blog” that is a mini blog service, information on friendship between users (“follower” table), and information on messages posted by each user (“message” table) ) And stored as sample data in the database 2B. Details of this sample data are shown below.
“Message” table: 1128300 “follower” table: 70803 Effective number of users: 16596 Target period: May to November 2008

  Then, an inquiry for registering (INSERT) a part of the above-described data in the database 2B was performed at a high load for 2 minutes, and the limit value of the update throughput at this time was measured. A case where the allowable number of delays is set to 10 is referred to as Example 1. In addition, under the same conditions, the limit value of the update throughput according to the conventional technique (always maintenance method) was also measured. This is referred to as Comparative Example 1.

The limit values of the update throughput of Examples 1 and 2 and Comparative Example 1 are shown below. The limit value of the update throughput indicates that the higher the numerical value, the lower the maintenance cost.
Example 1: 213.4 cases / second Comparative Example 1: 89.6 cases / second

  From the above measurement results, the limit value of the update throughput in Comparative Example 1 is 89.6 cases / second, whereas the limit value of the update throughput of Example 1 is 213.4 cases / second, It can be seen that the limit value of the update throughput is improved and the maintenance cost is reduced.

  In the cache maintenance system 200, an example in which the number of allowable delays is gradually increased will be described as a second embodiment. Then, the relationship between the allowable number of delays and the update throughput limit in Example 2 was measured, and the result is shown in FIG. In FIG. 23, the horizontal axis represents the allowable number of delays, and the vertical axis represents the update throughput limit value.

  From the results of FIG. 23, it can be seen that in Example 2, the limit value of the update throughput is improved as the allowable number of delays is increased. Here, in comparison with Comparative Example 1, in Example 2, the limit value of the update throughput is 27 times at the maximum. That is, it can be seen that the cache maintenance device 2 according to the present invention can reduce the maintenance cost as the allowable number of delays is increased.

  In FIG. 23, when the allowable number of delays is 1 (the left end of FIG. 23), a supplementary explanation will be made on the point that Example 2 gives a measurement result substantially equivalent to Comparative Example 1. In the second embodiment, when the allowable number of delays is one, maintenance is performed at the same frequency as in the first comparative example, and in addition, a maintenance determination cost is additionally generated. For this reason, as shown in the left end of FIG. 23, it is thought that the measurement result of Example 2 and Comparative Example 1 becomes substantially equivalent.

Cache maintenance device 1B, 2B, 3B, 4B Database 1C, 2C, 3C, 4C Cache 1CE, 2CE, 3CE, 4CE Cache entries 11, 13, 14 Update information storage unit 21, 22, 23, 24 Maintenance condition storage unit 31 , 32, 33, 34 Maintenance condition generation units 41, 43, 44 Maintenance target selection units 51, 52, 53, 54 Maintenance determination units 61, 63, 64 Cache maintenance unit 62 Cache difference maintenance unit (cache maintenance unit)
73 Statistical information calculation unit 84 Database load measurement unit 100, 200, 300, 400 Cache maintenance system

Claims (9)

In a cache maintenance device for maintaining a cache entry as a unit for storing the data in the cache in a distributed memory cache that stores data from a database in a cache constituted by one or more computer memories,
A maintenance condition storage unit that stores a maintenance condition that is a condition for maintaining the cache entry;
A maintenance determination unit that refers to the maintenance condition stored in the maintenance condition storage unit and determines that maintenance is possible when the maintenance condition is satisfied;
A cache maintenance unit that maintains the cache entry when the maintenance determination unit determines that maintenance is possible;
A cache maintenance device comprising: An update information storage unit that stores update information including at least one of a previous maintenance time and an update count indicating the number of times the data has been updated since the previous maintenance time;
A maintenance condition generation unit that receives at least one of a preset allowable delay number and an allowable delay time, generates the maintenance condition based on the input, and writes the maintenance condition in the maintenance condition storage unit; and
The maintenance condition generation unit
When only the allowable delay count is input, the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count is generated,
When only the allowable delay time is input, the maintenance condition that enables the maintenance when the elapsed time indicating the previous maintenance time to the current time exceeds the allowable delay time is generated,
When the allowable delay count and the allowable delay time are input, the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count or when the elapsed time exceeds the allowable delay time The cache maintenance device according to claim 1, wherein: A maintenance condition generation unit that receives at least one of a preset allowable delay number and an allowable delay time, generates the maintenance condition based on the input, and writes the maintenance condition in the maintenance condition storage unit;
The maintenance condition generation unit
When only the allowable delay count is input, the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count is generated,
When only the allowable delay time is input, the maintenance condition that enables the maintenance when an elapsed time exceeds the allowable delay time is generated,
When the allowable delay count and the allowable delay time are input, the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count or when the elapsed time exceeds the allowable delay time Produces
The maintenance determination unit
The difference table storing the difference data in the database table is referred to, the number of data items in the difference table is calculated as the update count, and the elapsed time from the time when the oldest data is stored in the difference table to the current time is calculated. To determine whether the maintenance condition is satisfied using the calculated number of updates and the elapsed time,
The cache maintenance unit
The cache maintenance device according to claim 1, wherein the cache entry is updated collectively with the difference data stored in the difference table, and the updated difference data is deleted from the difference table. An update information storage unit that stores update information including at least one of a previous maintenance time and an update count indicating the number of times the data has been updated since the previous maintenance time;
At least one of an allowable delay number and an allowable delay time set in advance and a target load value of maintenance cost are input, and based on the input, the maintenance condition is generated for each cache entry, and the maintenance condition storage unit A maintenance condition generation unit to write to,
A statistical information calculation unit that calculates, as statistical information, a cumulative value of maintenance counts indicating the number of times that the cache entry has been maintained in a preset statistical period;
The maintenance condition generation unit
Multiplying the cumulative value of the maintenance times by the target load value to calculate a reduction target value,
When only the allowable delay count is input, the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count is generated, and the reduction value calculated by Expression (1) is set as the reduction target value. Subtract from
When only the allowable delay time is input, the maintenance condition that enables the maintenance when the elapsed time from the previous maintenance time to the current time exceeds the allowable delay time is generated, and Expression (2) Subtract the reduction value calculated in step 1 from the reduction target value,
When the allowable delay count and the allowable delay time are input, the maintenance condition that enables the maintenance when the update count exceeds the allowable delay count or when the elapsed time exceeds the allowable delay time And subtracting the reduction value calculated by Equation (3) from the reduction target value,
2. The cache maintenance device according to claim 1, wherein when the value obtained by subtracting the reduction value from the reduction target value becomes 0 or less, the generation of the maintenance condition is terminated.
A = Mi × (N−1) / N (1)
A = Mi × (1−T / S) (2)
A = MIN [{Mi × (N−1) / N}, {Mi × (1-T / S)}] (3)
In Equations (1) to (3), A is a reduction value, Mi is a cumulative value of the number of maintenance times, N is an allowable delay number, T is an allowable delay time, and S is a statistical period.   5. The maintenance condition generation unit according to claim 4, wherein the maintenance condition generation unit generates the maintenance condition for the cache entry in which the maintenance condition is not generated and the cumulative value of the number of maintenance times is maximized. Cash maintenance equipment. The statistical information calculation unit calculates, as the statistical information, a cumulative value of the number of times the cache entry is referenced in the statistical period,
5. The maintenance condition generation unit according to claim 4, wherein the maintenance condition generation unit generates the maintenance condition for the cache entry in which the maintenance condition is not generated and the cumulative value of the reference count is minimized. Cash maintenance equipment. An update information storage unit that stores update information including at least one of a previous maintenance time and an update count indicating the number of times the data has been updated since the previous maintenance time;
A maintenance condition generation unit that receives at least one of a preset allowable delay number and an allowable delay time, generates the maintenance condition based on the input, and writes the maintenance condition in the maintenance condition storage unit;
A database load measuring unit that measures a load value indicating the load of the database, and
The maintenance condition generation unit
When only the allowable delay number is input, the maintenance condition that enables the maintenance when the update number exceeds the allowable delay number and the load value satisfies a preset load condition is generated,
When only the allowable delay time is input, the maintenance is possible when the elapsed time from the previous maintenance time to the current time exceeds the allowable delay time and the load value satisfies the load condition. Generating the maintenance condition;
When the allowable delay count and the allowable delay time are input, the update count exceeds the allowable delay count or the elapsed time exceeds the allowable delay time, and the load value satisfies the load condition The cache maintenance apparatus according to claim 1, wherein the maintenance condition that enables the maintenance is generated. In a cache maintenance method for maintaining a cache entry as a unit for storing the data in the cache in a distributed memory cache that stores data from a database in a cache configured by one or more computer memories,
A maintenance determination step that refers to a maintenance condition that is a condition for maintaining the cache entry and determines that maintenance is possible when the maintenance condition is satisfied;
If it is determined that maintenance is possible in the maintenance determination step, a cache maintenance step for maintaining the cache entry;
A cache maintenance method comprising:   A cache maintenance program for causing a computer to function as each means constituting the cache maintenance device according to any one of claims 1 to 7.

Images