DATABASE SYSTEM WITH DYNAMIC DATABASE CACHING

Abstract

A fully transactional mid-tier database system services database transactions. A cache manager dynamically loads database entries from a fully transactional backend-tier database system into the mid-tier database system according to the received database transactions. Time based aging or usage based aging can be assigned to selected tables in the mid-tier database system. Database entries contained in the selected tables are then automatically removed according to assigned aging constraints.

Description

TECHNICAL FIELD

The present disclosure relates generally to database systems.

BACKGROUND

A disk-based Relational Database Management System (RDBMS) uses disk storage to store and access large amounts of data. Much of the work performed by a conventional, disk-optimized RDBMS assumes that data primarily resides on disk. Optimization algorithms, buffer pool management, and indexed retrieval techniques are designed based on this fundamental assumption. One problem with disk storage is that access to the data is relatively slow.

Even when an RDBMS is configured to hold data in main memory, performance is still hobbled by assumptions of disk-based data residency. These assumptions cannot be easily reversed due to hard-coded processing logic, indexing schemes, and data access mechanisms.

In-memory resident relational database systems are deployed in the application-tier and operate in physical memory using standard Sequential Query Language (SQL) interfaces. By managing data in memory and optimizing data structures and access algorithms, in-memory database systems can provide improved responsiveness and throughput compared even to fully cached disk-based RDBMS. For example, the in-memory database can be designed with the knowledge that data resides in main memory and can take more direct routes to data, reducing the length of the code path and simplifying algorithms and structure.

When the assumption of disk-residency is removed, complexity is dramatically reduced. The number of machine instructions drop, buffer pool management disappears, extra data copies are not needed, and index pages shrink. The database design becomes simple and more compact, and data requests are executed faster. However, in-memory database systems currently can only operate on a relatively small static portion of the data contained in a disk-based database system.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic block diagram showing a database system that provides dynamic database caching.

FIG. 2 is a schematic block diagram showing in more detail how a cache manager caches database entries in a secondary/application-tier database system.

FIG. 3 is a flow diagram explaining operations performed by the cache manager in FIG. 2.

FIG. 4 is a schematic block diagram showing how the cache manager operates with cache groups.

FIG. 5 is a flow diagram describing in more detail how dynamic database caching operates with cache groups.

FIG. 6 is a schematic block diagram showing how dynamic database caching selectively ages out database entries.

FIG. 7 is a flow diagram describing usage based aging in more detail.

FIG. 8 is a flow diagram describing time based aging in more detail.

INTRODUCTION

A fully transactional mid-tier database system services database transactions. A cache manager dynamically loads database entries from a fully transactional backend-tier database system into the mid-tier database system according to the received database transactions. Time based aging or usage based aging can be assigned to selected tables in the mid-tier database system. Database entries contained in the selected tables are then automatically removed according to assigned aging constraints.

DETAILED DESCRIPTION

FIG. 1 is a schematic representation of a multi-tiered database system. A primary database system 140 can be any conventional fully-relational database system, such as a disk-based Relational Database Management System (RDBMS). The primary database system 140 typically uses disk storage to store and access large amounts of data that in one example includes multiple different tables 144. The primary database system 140 is alternatively referred to as a backend database system or a backend-tier database system.

A secondary database system 122 typically operates on a server 100 that is remote from primary database system 140 and includes a storage manager that stores and manages different tables 127 that contain different database entries. The secondary database 122 in one example is an in-memory fully-relational database that is deployed in an application tier and operates in physical memory of the server 100. The secondary database system 122 is alternatively referred to as an application-tier database system or an in-memory database system.

Applications 112A and 112B are initiated by clients 102A and 102B, respectively, via a local or wide area network 110. The network 110 is alternatively referred to as the Internet. The applications 112 can be any software program that accesses or references database entries in a database. For example, the applications 112 could be software programs used for booking airline reservations, ordering products over the Internet, managing financial transactions for banks or investment institutions, or tracking telephone call usage. Of course these are just a few examples of the essentially limitless number of data management applications that may be used with the database systems shown in FIG. 1.

Connections from the clients 102 can either be direct connections or client/server connections. Direct connections refer to Sequential Query Language (SQL) libraries and routines that implement a direct driver. The application 112A can create a direct driver connection when it runs on the same server 100 that operates the secondary database system 122. In a direct driver connection, the direct driver directly loads the secondary database 122 into the application's heap space or a shared memory segment. The application 112A then uses the direct driver to access a memory image of the secondary database 122. Because no inter-process communication is required, a direct driver connection provides fast performance.

The client/server connection accommodates connections from the remote client 102B to secondary database 100 over network 110. Applications 112B on the client 102B issue calls to local client driver libraries 114B that communicate with a server/child process 113 on the server 100 containing secondary database 122. The server/child process 113, in turn, issues native requests to the direct driver provided by the server libraries for accessing the secondary database 122. If a client 102 and server 100 reside on separate nodes in a network, then communication is provided using sockets and Transmission Control Protocol/Internet Protocol (TCP/IP) communications.

The secondary database 122 maintains durability through a combination of transaction logs 124 and periodic refreshes of a disk-resident version of the secondary database 122. The transaction logs 124 are written to disk asynchronously or synchronous with the completion of transactions 118 and are controlled by the applications 112 at the transaction level. The transaction logs 124 can be used to recover a transaction 118 if the application 112 or database 122 fails, undo transactions 118 that are rolled back, replicate changes to other databases, replicate changes in the secondary database 122 to the primary database 140, or enable applications 112 to detect changes to database entries.

Checkpoint files 126 are used to keep a snap shot of the secondary database 122. In the event of a system failure, the checkpoint files 126 are used to restore the secondary database 122 to a last transactionally consistent state. A checkpoint operation scans the secondary database 122 for blocks that have changed since the last checkpoint and updates the checkpoint files 126 with the changes and removes any transaction log files 124 that are no longer needed.

The applications 112 create and manage the tables 127 that may exist only in secondary database 122. The applications 112, through cache manager 150, can also cache frequently used subsets of database entries from the primary database 140. The tables 127 managed exclusively by the secondary database 122 and the tables 127 cached from primary database 140 may all coexist in the same secondary database 122, and are all persistent and recoverable.

Queries and updates to the tables 127 are performed by the applications 112 through standard SQL. Applications 112 running on other different mid-tier servers may cache different or overlapping subsets of the data in primary database 140.

The cache manager 150 can cache entire tables or table fragments from the primary database 140 to the secondary database 122 operating on server 100. The table fragments are described through an extended SQL syntax and are cached into corresponding tables. For example, tables 128A, 130A, and 132A from primary database 140 are cached into corresponding tables 128B, 130B, and 132B in the secondary database 122. The cached tables 128B, 130B, or 132B may comprise the entire corresponding tables 128A, 130A, or 132B from primary database 140 of may only include selected database entries from the primary database tables 128A, 130A, or 132B. The database entries can be any record, tuple, column, row or other data item that typically exists in a fully transactional database system.

The secondary database 122 dynamically caches performance-critical subsets of the primary database 140, enabling both reads and updates, and automatically manages data consistency between the cached secondary database 122 and the primary database 140. The applications 112 read and update the cached tables 127 using standard SQL, and the cache manager 150 automatically propagates updates from the primary database 140 to the secondary database 122 and vice versa.

Thus, the cached secondary database 122 offers applications 112 the full generality and functionality of a fully-relational database, the transparent maintenance of cache consistency with the primary database 140, and the real-time performance of an application-tier in-memory database system.

FIG. 2 shows the operations performed by the cache manager 150 in more detail. The cache manager 150 dynamically varies what subset of tables 127 are cached from the primary database system 140 into the secondary database system 122 according to the transactions 118 received from the applications 112 in FIG. 1.

The cache manager 150 first determines what transactions 118 can be serviced by the secondary database 122. For example, the cache manager 150 determines if the referenced tables 200A and referenced primary keys 200B in SQL statement 200 reside in secondary database 122. If the referenced database entries reside in the secondary database 122, the transaction 118 is serviced by the secondary database 122.

When the database entries referenced by the transaction 118 do not reside in the secondary database 122, the cache manager 150 may query the primary database 140 for the missing database entries. For example, table identifier 119A and primary key identifier 119B reference a database entry 204 in a table 130B having a primary key value PK=1. Since the database entry 204 is not currently located in the secondary database 122, the cache manager 150 queries the primary database 140. The referenced database entry 204 in primary database 140 is then inserted into table 130B in the secondary database 122. The transaction 118 may then be serviced by the secondary database 122 using the uploaded database entry 204.

FIG. 3 explains some of the operations performed by the cache manager 150 in more detail. Referring both to FIGS. 2 and 3, in operation 250 the cache manager 150 receives or monitors the database transactions 118 directed to secondary database 122. In operation 252, the cache manager 150 may identify the database entries associated with the database transaction. For example, the cache manager 150 obtains the table identifiers and keys referenced by the transaction 118. The cache manager 150 in operation 254 searches the secondary database 122 for the referenced database entries.

If the secondary database 122 contains the referenced database entries in operation 254, the transaction is serviced by the secondary database in operation 256. Otherwise, the cache manager 150 sends one or more queries to the primary database 140 that reference the database entries that are not contained in the secondary database 122.

In some embodiments, the secondary database 122 may contain some, but not all, of the database entries referenced by the transaction 118. In this situation, the cache manager 150 may send queries referencing only the missing database entries. In other embodiments, when only some of the database entries referenced by the transaction 118 are currently located in the secondary database 122, the cache manager 150 may query the primary database 140 for all of the database entries referenced by the transaction 118.

The database entries accessed in the primary database 140 are then uploaded into the secondary database 122 in operation 260. For example, the cache manager 150 may generate additional SQL statements that cause the primary database entries to be inserted into the secondary database 122. Any required commitment is performed on the uploaded database entries 204 in operation 262. The transaction 118 is then serviced by the secondary database in operation 256.

Cache Groups

FIG. 4 shows how dynamic database caching is used in conjunction with cache groups. A cache instance or cache group is a collection of related records that are uniquely identifiable, and is used to model a complex object. For example, a cache group 314 may be a group of rows that correspond to a set of frequently used tables that are related to each other through foreign key constraints.

Cache instances/cache groups can be used when both loading data from primary database 140 into the secondary database 122 and via versa and when database entries are aged out of the secondary database 122. The cache group 314 may be configured to contain entire tables or configured to contain only subsets of table rows and/or table columns.

The following SQL syntax is one example of how the cache group 314 is created that includes different database entries from both CUSTOMER table 302 and ORDER table 304.

create cache group cache_customer from
customer(pk1 int not null primary key),
orders(pk2 int not null primary key,
fk2 int, foreign key (fk2) references customer(pk1));

In this example, each customer in the CUSTOMER table 302 has a primary key on its ID. One customer may have many orders in the ORDER table 304, where each order has a foreign key (fk2) that references a CUSTOMER(ID). Configuring cache group 314 causes the cache manager 150 to treat all of the order information and associated customer information associated with the transaction as a single cache instance. For example, a transaction may only reference one of the database entries associated with cache group 314. If the referenced database entry is not contained in secondary database 122, the cache manager 150 uploads all of the database entries associated with the cache instance from the primary database 140 at the same time.

In this example, CUSTOMER table 302 is considered a root table and ORDERS table 304 is considered a child table. Database entries can be uploaded or flushed based on the root table 302. For example, all child rows for a root table currently located in the secondary database 122 can also be presumed to be currently located in the secondary database 122. This prevents the cache manager 150 from having to determine if all of the foreign keys for a cache group exist in the secondary database 122.

Referring to FIGS. 4 and 5, the cache manager 150 receives a transaction 306 in operation 350. The transaction 306 selects all of the orders for a customer having an ID=100. This transaction 306 may be implemented using the following SQL statement 307.

• • select * from orders with a customer_ID=100

The cache manager 150 in operation 352 determines if the secondary database 122 contains any database entries in the ORDERS table 304 with the customer_ID=100. If so, the SQL statement 307 is serviced by the secondary database 122 in operation 354 by returning the requested database entries.

When the secondary database 122 does not contain orders with customer_ID=100, the cache manager 150 sends the following query 308 to the primary database 140 in operation 356 selecting the database entries from the ORDER table with customer_ID=100.

• • select * from orders with a customer_ID=100

However, the cache manager 150 also determines that the referenced database entries are part of the cache group 314. The cache manager 150 identifies the cache group via the foreign keys assigned to orders in table 304. Accordingly, the cache manager 150 in operation 358 sends the following second query 310 that selects all of the rows from the CUSTOMER table in the primary database 140 that have an ID=100.

• • select * from customer where ID=100

The different database entries accessed in the primary database with queries 308 and 310 are referred to as cache instance 320. It should be understood that the two different queries 308 and 310 select more database entries 320A-320C from the primary database 150 than what was actually referenced by the transaction 306. Thus, in one embodiment, all of the database entries associated with a same cache instance are loaded into the secondary database at the same time.

The cache manager 150 also sends insert commands 312 to the primary database 140 in operation 360 which cause the rows 320A-320C associated with cache instance 320 to be inserted into the secondary database in operation 360. The transaction 306 is then serviced by the secondary database in operation 354.

Aging

FIG. 6 shows how different aging parameters are used to automatically remove database entries from the secondary database 122. Assigning different aging parameters to different database tables allows the secondary database 122 to dynamically cache more relevant user content. For example, an airlines reservation system may use the secondary database 122 for caching a subset of customer flight reservations and caching a subset of airline flight schedules.

The customer flight reservation tables may be more effectively cached based on usage. For instance, it may be advantageous to maintain customer information in the secondary database 122 for customers who frequently or most recently book airline reservations. This allows faster database response to user reservation queries and further may reduce the amount of traffic between the secondary database 122 and primary database 140. This is referred to generally as “usage based aging.”

Other types of data may be more “time based.” For example, airline flight schedules may be highly queried for some period of time. However, after the airline flight arrives at a destination, that flight information is much less likely to be queried again by users. Accordingly, it may be advantageous to remove this type of “time-based” data from the secondary database 122 after a specified time period. The cache manager 150 can be programmed to selectively associate different tables in secondary database 122 with these different usage based and time based aging constraints.

Usage Based Aging

The following SQL, statements may be used for configuring tables A and C in FIG. 6 as usage based aging tables.

CREATE TABLE A (C1 INT , C2 INT );
ALTER table A ADD AGING LRU;
TT AgingLRUConfig (LowUsageThreshold,
HighUsageThreshold,AgingCycle)

CREATE TABLE C (C1 INT , C2 INT );
ALTER table C ADD AGING LRU;
TT AgingLRUConfig (LowUsageThreshold,
HighUsageThreshold,AgingCycle)

A listing 400 identifies in column 400A the tables in secondary database 122 that are configured with usage based aging constraints. In this example, the Least Recently Used (LRU) database entries in tables A and C are periodically removed according the amount of available space in the secondary database 122.

High Usage Threshold (HUT) values 400B identify a percentage of used memory space in the secondary database 122 that trigger the cache manager 150 to remove least recently used database entries. Low Usage Threshold (LUT) values 400C can also be assigned to the tables A and B and identify a second lower percentage of used memory space in the secondary database 122. When the HUT value 400B is reached, the cache manager 150 removes least recently used database entries until the storage space in secondary database 122 reaches the LUT value 400C.

Aging Cycle (AC) values 400D in listing 400 indicates how often the cache manager 150 evaluates the least recently used database entries in tables A and C. For example, a counter or clock 404 is monitored by cache manager 150. The cache manager 150 periodically checks the amount of used memory space in the secondary database 122 after counter/clock 404 indicates the expiration of each aging cycle 400D. If the amount of used memory space reaches HUT 400B, the cache manager 150 removes the least recently used database entries from the associated tables A and/or C.

Last Used (LU) tags 408 and 412 indicate when the database entries in tables A and C were respectively last used. The LU tags 408 and 412 may use the value provided by counter/clock 404 at the time the associated database item was last accessed or referenced. The LU tags 408 and 412 can then be updated with a current time from counter/clock 404 whenever the associated database entries in tables A or C are accessed or referenced again by another transaction or when the database entries are uploaded again from the primary database 140.

Time Based Aging

Selected tables in the secondary database 122 can also be assigned time based aging constraints. For example, the following SQL statement configures time based aging constraints for table D.

• • CREATE TABLE D (Timestamp C1, Timestamp c2, c3 INT) AGING USE c1 LIFETIME {MINUTES|HOURS|DAYS} CYCLE {MINUTES|HOURS|DAYS}

The time based aging SQL statement causes table D to be listed in column 402A of time based aging listing 402. A lifetime value 402B in listing 402 designates how long database entries in table D should reside in the secondary database 122. A cycle time value 402C defines a time period for the cache manager 150 to periodically evaluate the database entries in table D. If different cycle times 402C are defined for different tables, then the cache manager 150 may wake up based on an a single cycle time value for all of the tables, or may wake up according to the cycle times for each individual table.

The time based aging SQL statement above also configures a column in table D with timestamp values 414. The timestamp values 414 could be a date and time value from counter/clock 404 or could alternatively be a counter value from counter/clock 404 that is continuously incremented until reaching a reset value. In one embodiment, the timestamp values 414 are set to the value of counter/clock 404 when the associated database entries are first loaded into the secondary database 122.

Referring both to FIGS. 6 and 7, in operation 450 the tables in the secondary database 122 are configured with different usage based aging constraints and time based aging constraints as described above. In operation 452, the cache manager 150 identifies the different usage based aging tables and time based aging tables as defined in listings 400 and 402. In operation 454, the different aging parameters in listing 400 are identified for the usage based tables. For example, the cache manager 150 in operation 454 identifies the high usage thresholds 400B, low usage thresholds 400C and the aging cycles 400D for tables A and C.

In operation 456 the cache manager 150 waits for one of the aging cycles to be reached for one of the tables A or C. For example, the cache manager 150 determines when the counter/clock 404 reaches the aging cycle 400D for one of the tables A or C. When an aging cycle is reached in operation 456, the cache manager 150 determines the amount of memory space currently being used in the secondary database 122. If the high usage threshold value 400A is reached for either table A or table C in operation 458, the least recently used database entries for that table are removed in operation 462.

For example, the high usage threshold value 400B for table A may be set to 75% and the high usage threshold value 400B for table C may be set to 85%. If storage in the secondary database 122 is 80% full when the counter/clock 404 reaches a next aging cycle time 400D, the least recently used database entry in table A is removed in operation 462. The LU tag value LU=2 indicates that database entry 416 is the least recently used entry in table A and is accordingly removed from the secondary database 122 in operation 462.

In operation 464, the cache manager 150 determines the amount of used storage space after the database entry 416 is removed in operation 462. If the percentage of used memory space does not drop below the low usage threshold value 400C for table A in operation 464, the next least recently used database entry is removed from table A in operation 462. Database entries are removed from tables A until the amount of utilized space in the secondary database drops below the low usage threshold value in operation 464. A next aging cycle is started for table A in operation 460 and the cache manager 150 waits for the expiration of the next aging cycle 400D in operation 456 before conducting the next usage based purge in operation 458.

If the aging cycles 400D for tables A and C are different, then a same aging cycle value 400D may be used. If the HUT values 400B for both table A and table C are reached at the next common aging cycle, then database entries may be removed from both table A and table C in a round robin fashion. Alternatively, different LRU aging sessions may be separately conducted for tables A and C and LRU database entries for each table removed independently according to their associated HUT values 400B, LUT values 400C and aging cycles 400D.

For example, at the next aging cycle for table C, memory utilization in secondary database 122 may exceed the 85% high usage threshold value 400B assigned to table C. Accordingly, least recently used database entries are removed from table C in operation 462 until the database storage reaches the low usage threshold value 400C for table C. In this example, the cache manager 150 removes the least recently used database entries 418 (DB entries #1, #3, and #4) from table C in order to reach the low usage threshold value 400C associated with table C.

Higher priority data in a particular table may be assigned larger high usage threshold values 400B and/or larger low usage threshold values 400C. In addition, the aging cycles 400D for high priority data may be set to longer time periods. These larger threshold values 400B, 400C, and 400D cause the cache manager 150 to remove the least recently used database items for those tables less frequently. Thus, the usage based aging parameters 400 allow automatic customized removal of different types of selectable data from the secondary database 122.

FIG. 8 explains in more detail how the cache manager 150 conducts time based aging. Referring to FIGS. 6 and 8, any tables having time based aging constraints are identified in operation 480. In this example, table D is assigned a lifetime value 402B and an associated cycle time value 402C in listing 402. The cache manager 150 uses the counter/timer 404 in operation 482 to determine when a next cycle time 402C is reached. Any rows in table D that have timestamps 414 exceeding the lifetime value 402B are identified in operation 484 and removed in operation 486.

For example, the lifetime value 402B for table D may be set to a particular counter value of say LIFETIME=20. When the cycle time 402C is reached in operation 482, the value for counter/clock 404 is compared with the timestamp values 414 in table D. The difference between the value of counter/clock 404 and the timestamp values 414 are determined in operation 484. In this example, the counter 404 may have a current value of 28. The difference between the current value of counter 404 (28) and the timestamp value for database entry 420 in table D (TS=7) is greater than the lifetime value 402B (LIFETIME=20). Accordingly, the database entry 420 is removed from table D in operation 486. Any other database entries in table D with expired lifetimes are also removed in operation 486.

Similar to usage based aging, different tables can be assigned different lifetime values 402B and cycle times 402C. Higher priority data may be assigned larger lifetime values 402B and/or may be evaluated less frequently by assigning larger cycle time values 402C.

Tables associated with even higher priority data might not be assigned any aging constraints. For example, table B in FIG. 6 is not assigned any aging constraints. Accordingly, the database entries in table B remain in the secondary database 122 until replaced by updates from the primary database 140.

Cache Group Aging

Referring back to both operation 462 in FIG. 7 and operation 486 in FIG. 8, database entries associated with the same cache group may be removed according to usage based or time based aging constraints. Referring also back to FIG. 4, the CUSTOMER table 302 may both be assigned usage based aging constraints. The high usage threshold may be reached for CUSTOMER table 302 and the database entry PK1=100 in the CUSTOMER table 302 may be the least recently used. Accordingly, the entire cache instance 320 may be removed in operation 462 in FIG. 7. For example, the root database entry PK1=100 is removed from the CUSTOMER table 302 and the child database entries PK2=14 and PK2=20 with referencing foreign keys are removed from the ORDERS table 304.

In another example, usage based aging may be assigned to the child ORDERS table 304 in FIG. 4. When storage in secondary database 122 reaches the high usage threshold value for ORDERS table 304, database entry PK1=14 may be the least recently used.

In one embodiment, all of the root and child database entries for the same cache instance 320 are removed from the secondary database in operation 462 in FIG. 7. In another embodiment, cache groups are only aged based on the database entries in the root table 302. For example, since database entry PK2=14 is located in child table 304 and not located in root table 302, no usage based aging is performed.

For time based aging, database entries associated with the same cache group may also be controlled by the root table. For example, timestamps may only be applied to the database entries in the root CUSTOMER table 302 in FIG. 4. Whenever one of the database entries in CUSTOMER table 302 resides in the secondary database 122 beyond a specified lifetime value, all of the database entries associated with that cache instance are removed at the same time.

For example in FIG. 4, database entry PK1=100 may reside in the secondary database 122 beyond a lifetime value assigned to CUSTOMER table 302. Accordingly, both database entry PK1=100 in CUSTOMER table 302 and the other database entries PK2=14 and PK2=20 in ORDERS table 304 associated with the same cache instance 320 are removed by the cache manager 150.

The system described above can use dedicated processor systems, micro controllers, programmable logic devices, or microprocessors that perform some or all of the operations. Some of the operations described above may be implemented in software and other operations may be implemented in hardware.

For the sake of convenience, the operations are described as various interconnected functional blocks or distinct software modules. This is not necessary, however, and there may be cases where these functional blocks or modules are equivalently aggregated into a single logic device, program or operation with unclear boundaries. In any event, the functional blocks and software modules or features of the flexible interface can be implemented by themselves, or in combination with other operations in either hardware or software.

Having described and illustrated the principles of the invention in a preferred embodiment thereof, it should be apparent that the invention may be modified in arrangement and detail without departing from such principles. Claim is made to all modifications and variation coming within the spirit and scope of the following claims.

Claims

1. A method, comprising: operating a fully transactional mid-tier database;receiving one or more database transactions at the mid-tier database; anddynamically loading database entries from a fully transactional backend-tier database into the mid-tier database according to the received database transactions.

2. The method according to claim 1 further comprising operating the mid-tier database as an in-memory database.

3. The method according to claim 1 further comprising: identifying database entries associated with the database transactions;servicing the database transactions with the mid-tier database when the identified database entries are contained in the mid-tier database; andquerying database entries in the backend-tier database when the identified database entries are not contained in the mid-tier database.

4. The method according to claim 3 further comprising: uploading the queried database entries into the mid-tier database; andservicing the database transactions with the mid-tier database using at least some of the uploaded database entries from the backend-tier database.

5. The method according to claim 1 further comprising: identifying tables and primary keys referenced by the database transactions;searching the mid-tier database for the identified tables and primary keys;servicing the database transactions with the mid-tier database when the identified tables and primary keys are located in the mid-tier database; andaccessing the backend-tier database when the identified tables and primary keys are not located in the mid-tier database.

6. The method according to claim 1 further comprising: associating database entries from different tables with a same cache group;identifying a database transaction that references one or more database entries associated with the cache group;uploading the entire cache group from the backend-tier database into the mid-tier database when the referenced database entries are not contained in the mid-tier database.

7. The method according to claim 6 further comprising: identifying a first database entry in a first table;identifying a primary key associated with the first database entry in the first table;identifying a second database entry in a second table having a foreign key referencing the primary key in the first table; andassociating both the first database entry in the first table and the second database entry in the second table with the same cache group.

8. A method, comprising: operating a database system;assigning aging parameters to selected tables in the database system; andremoving database items from the selected tables in the database system according to the assigned aging parameters.

9. The method according to claim 8 further comprising: associating time based aging with the selected tables in the database system; andremoving the database items from the selected tables according to how long the database items have resided in the database system.

10. The method according to claim 8 further comprising: associating usage based aging with the selected tables in the database system; andremoving the database items from the selected tables according to how recently the database items have been used in the database system.

11. The method according to claim 10 further comprising: assigning a high usage threshold value, a low usage threshold value, and an aging cycle to the selected tables in the database system;identifying an amount of available space in the database system for each aging cycle;removing at least some of the least recently used database items from the selected tables when an amount of storage space in the database system reaches the high usage threshold value; andremoving least recently used database items from the selected tables until the amount of storage space in the database system reaches the low usage threshold value.

12. The method according to claim 8 further comprising: configuring different time based aging tables and usage based aging tables in the database system;removing at least some of the database items from the time based aging tables that have resided in the database system beyond a configured time period;removing at some of the least recently used database items from the usage based aging tables when storage in the database system reaches a configured usage threshold; andskipping removal of database items from non-time based and non-usage based tables in the database system.

13. The method according to claim 8 further comprising: associating some of the database items from different selected tables with a same cache group;assigning an aging parameter to a root table of the cache group; andremoving the database items from both the root table and one or more child tables associated with the cache group according to the aging parameter assigned to the root table.

14. Computer readable media containing instructions that when executed by a computer, comprise: operating a fully-transactional secondary database;receiving database transactions directed to the secondary database;searching the secondary database for data items referenced by the transactions;servicing the database transactions with the secondary database when the secondary database contains the referenced data items;querying a fully-transactional primary database when the referenced data items are not contained in the secondary database; andupdating the secondary database with at least some of the data items queried in the primary database.

15. The computer readable media according to claim 14 further comprising instructions that when executed result in: assigning aging parameters to selectable data items in the secondary database; andautomatically removing data items from the secondary database according to the assigned aging parameters.

16. The computer readable media according to claim 14 further comprising instructions that when executed result in: assigning either a time based aging parameter or a least recently used aging parameter to programmably selectable tables in the secondary database; andremoving the data items according to the time based aging parameter or least recently used aging parameter assigned to the tables containing the data items.

17. The computer readable media according to claim 14 further comprising instructions that when executed result in: configuring time based aging tables and usage based aging tables in the secondary database;removing at least some of the data items from the time based aging tables that have resided in the secondary database beyond a configured time period; andremoving at some of the least recently used data items from the usage based aging tables when storage in the secondary database reaches a configured usage threshold value.

18. The computer readable media according to claim 14 further comprising instructions that when executed result in: associating at least some of the data items from different tables with a same cache group;identifying transactions referencing one or more of the data items from the cache group;uploading the entire cache group from the primary database into the secondary database when any of the referenced data items are not contained in the secondary database.

19. The computer readable media according to claim 18 further comprising instructions that when executed result in: assigning an aging parameter to a root table for the cache group; andremoving all of the data items associated with the cache group according to the aging parameter assigned to the root table.

20. A database management system, comprising: an application-tier database system receiving requests from database applications; anda cache manager configured to maintain at least a sub-set of database entries from a backend-tier database system in the application-tier database system, the cache manager dynamically varying what database entries are loaded from the backend-tier database system into the application-tier database system according to the requests from the database applications.

21. The database management system according to claim 20 wherein the cache manager is further configured to: monitor the different requests received by the application-tier database system;identify what database entries are referenced by the requests;service the requests with the application-tier database system when the referenced database entries are contained in the application-tier database system;temporarily stall the requests when the referenced database entries are not contained in the application-tier database system;query the backend-tier database system for the referenced database entries;load the database entries queried from the backend database system into the application-tier database system; andservice the requests using at least some of the database entries loaded into the application-tier database system from the backend-tier database system.

22. The database management system according to claim 20 wherein the cache manager is further configured to: identify cache groups that include both a root table with one or more database entries having primary keys and one or more child tables having one or more database entries having foreign keys referencing the primary keys in the root table; anddynamically removing the identified cache groups from the application-tier database system or dynamically loading the identified cache groups from the backend-tier database system into the application-tier database system.

23. The database management system according to claim 20 wherein usage based aging is assigned to selected tables in the application-tier database system and the cache manager removes database entries contained in the selected tables according to how recently the database entries have been used in the application-tier database system.

24. The database management system according to claim 20 wherein time-based aging is assigned to selected tables in the application-tier database system and the cache manager removes database entries contained in the selected tables according to how long the database entries have resided in the application-tier database system.