Abstract: A database query is partitioned into an initial partition including a plurality of parallel groups, and is executed, via an execution plan, based on the initial partition. A sampling subset of data is identified from the plurality of parallel groups. Substantially in parallel with the executing of the query, the execution plan is executed on the sampling subset of data as a sampling thread. The execution plan is modified based on feedback from the execution of the execution plan on the sampling subset of data.

Abstract: An apparatus, system, and method are disclosed for improving query execution plan efficiency in a database management system. The present invention binds both static database queries and dynamic database queries to new query execution plans (QEPs) that produce the same result set as a former QEP bound to the database queries. Next, the present invention identifies one of the former QEPs and the new QEP as an active QEP and automatically collects execution statistics for one of the former QEPs and the new QEP. Finally, the present invention automatically selects one of the former QEPs and the new QEP as the active QEP in response to completion of the automatic collection of execution statistics, the active QEP selected such that the automatically selected QEP provides optimal execution performance based on performance criteria in comparison to the automatically unselected QEP, the active QEP selected based on the execution statistics.

Abstract: A system and computer readable medium for executing a query to access data stored in a database, wherein the query includes a plurality of DISTINCT keys, is disclosed. The system and computer readable medium includes a capture module for identifying each of the plurality of DISTINCT keys in the query and a sort module coupled to the capture module for determining if more than one sort is needed to execute the query, performing a first DISTINCT operation on a first DISTINCT key of the plurality of DISTINCT keys, storing data fetched from the first DISTINCT operation in a master workfile only if more than one sort process is needed to execute the query, and utilizing the master workfile to perform subsequent DISTINCT operations on the other of the plurality of DISTINCT keys.

Abstract: Techniques for database table look-up are provided. The techniques include storing one or more column attributes of a database table in a data structure, wherein the data structure also comprises a record identification (RID) column of a table, one or more predicate columns corresponding to the RID column, and a sequence number column that is associated with one or more updated records, generating a key using one or more portions from one or more of the one or more predicate columns, using the key to partition the data structure, wherein partitioning the data structure comprises partitioning the one or more predicate columns for evaluation, and evaluating the one or more predicate columns against the data structure for each matching predicate column-data structure partition.

Abstract: Disclosed is a technique for maintaining statistics. A query is received. Column group generation is performed to form one or more column groups. Column group refinement is performed based on statistics data generated for a previously received query. A statistics check is automatically performed. A statistics collection command is generated to obtain statistics.

Abstract: Techniques are provided for processing a query using an index. It is determined whether a number of consecutive keys that evaluate to a false-positive when applying a screening predicate in the query exceeds a false-positive predefined value. One or more keys in the index are skipped in response to determining that the number of consecutive keys exceeds the false-positive predefined value.

Abstract: A method, computer program product, and system for managing row identifier (RID) list processing on an index are provided. The method, computer program product, and system provide for accessing one or more key values in the index based on one or more keys specified in a query, retrieving a plurality of row identifiers corresponding to the one or more key values from the index, and predicting an actual number of row identifiers to be retrieved from the index based on the one or more key values accessed and the plurality of row identifiers retrieved.

Abstract: A method, storage server, and computer readable medium for off-loading star-join operations from a host information processing system to a storage server. At least a first and second set of keys from a first and second dimension table, respectively are received from a host system. Each of the first and second set of keys is associated with at least one fact table. A set of locations associated with a set of foreign key indexes are received from the host system. A set of fact table indexes are traversed. At least a first set of Row Identifiers (“RIDs”) associated with the first set of keys and at least a second set of RIDs associated with the second set of keys are identified. An operation is performed on the first and second sets of RIDs to identify an intersecting set of RIDs. The intersecting set of RIDs are then stored.

Abstract: Various embodiments of a computer-implemented method, system and computer program product maintain a logical page having a predetermined size. Data is added to an uncompressed area of the logical page. The uncompressed area of the logical page is associated with an uncompressed area of a physical page. The logical page also has a compressed area associated with a compressed area of a physical page. In response to exhausting the uncompressed area, data in the uncompressed area is included in the compressed area. The uncompressed area is adjusted.

Abstract: A method of, a computer program product for and a computer system for managing statistics in a database management system are provided. A value of a volatility of a statistic is determined. A value of a staleness of the statistic is determined using the value of the volatility. A statistics collection task is generated in response to, at least in part, the value of the staleness exceeding a value of a staleness threshold. The statistics collection task is executed to collect an additional value of the statistic. The additional value of the statistic is stored.

Abstract: Various embodiments of a computer-implemented method, system and computer program product are provided. A first plurality of key entries of a first index page are compressed in accordance with an order specified by a first keymap of the first index page. The first keymap also indicates respective positions of the key entries of the first plurality of key entries. A second keymap is generated indicating the order and also indicating respective post-compression positions of the key entries of the first plurality of key entries. The compressed first plurality of key entries is stored on a second index page with the second keymap.

Abstract: A method for performing disjunctive single-index access on a database is disclosed. The method includes a query engine determining whether a first OR predicate and a second OR predicate map to a shared index. Responsive to the query engine determining that the first OR predicate and the second OR predicate map to the shared index, the first OR predicate and the second OR predicate are ordered in an ascending sequence. A first range of the first OR predicate is queued. A first row of the shared index is probed for the first range of the first OR predicate. Whether the first row is disqualified by the first OR predicate is determined. Responsive to determining that the first row is not disqualified by the first OR predicate, the row is reported in a result reporting structure.

Abstract: An apparatus, system, and method are disclosed for improving query execution plan efficiency in a database management system. The present invention binds both static database queries and dynamic database queries to new query execution plans (QEPs) that produce the same result set as a former QEP bound to the database queries. Next, the present invention identifies one of the former QEPs and the new QEP as an active QEP and automatically collects execution statistics for one of the former QEPs and the new QEP. Finally, the present invention automatically selects one of the former QEPs and the new QEP as the active QEP in response to completion of the automatic collection of execution statistics, the active QEP selected such that the automatically selected QEP provides optimal execution performance based on performance criteria in comparison to the automatically unselected QEP, the active QEP selected based on the execution statistics.

Abstract: A database query is partitioned into an initial partition including a plurality of parallel groups, and is executed, via an execution plan, based on the initial partition. A sampling subset of data is identified from the plurality of parallel groups. Substantially in parallel with the executing of the query, the execution plan is executed on the sampling subset of data as a sampling thread. The execution plan is modified based on feedback from the execution of the execution plan on the sampling subset of data.

Abstract: A partial index availability system places, in a restricted state, all pages in the index associated with a structure modification, when an error occurs in processing a log of the said structure modification. This maintains traversability of the rest of the index that is not in restricted state. The system locates and marks a left sentinel and a right sentinel associated with a non-leaf page that is in a restricted state preventing an undo of a transaction. The sentinels prevent a transaction from accessing an uncommitted change associated with the non-leaf page. After a recovery procedure is run the entire index is made available. During the period between the placement of the index pages in LPL or rebuild pending to the time of final removal of these pages from their restrictive states as a result of a recovery procedure being run, the users are given access to the non-restricted portion of the index.

Abstract: An apparatus, system, and method are disclosed for ensuring query execution plan stability in a database management system. The present invention binds a static database query to a new query execution plan (QEP) that produces the same result set as an original QEP bound to the static database query. Next, the present invention identifies one of the original QEP and the new QEP as a primary QEP and automatically collects execution statistics for the original QEP and the new QEP. Finally, the present invention automatically selects one of the original QEP and the new QEP as the primary QEP in response to completion of the automatic collection of execution statistics, the primary QEP selected such that the automatically selected QEP provides optimal execution performance based on performance criteria in comparison to the automatically unselected QEP, the primary QEP selected based on the execution statistics.

Abstract: A method, system, and computer program product are provided for retrieving records into a main memory. A first number of gaps and a first total gap size are received for a list of records from a database subsystem. A determination is made of a first average gap size using the first number of gaps and the first total gap size. A determination is made as to whether the first average gap size is greater than a prestaging threshold value. Responsive to the first average gap size being equal to or less than the prestaging threshold value, a prestaging flag is set for the list of records. Then, the list of records is retrieved into the main memory using prestaging.

Abstract: Provided are a techniques for processing a query including semi-joins. At execution time, a next semi-join is selected from the semi-joins for execution in a current round of semi-join executions. A reporting threshold is determined that indicates a number of record-identifiers to be retrieved for the determined semi-join. The selected semi-join is executed until the determined number of record identifiers are retrieved.

Abstract: A method, storage server, and computer readable medium for off-loading star-join operations from a host information processing system to a storage server. At least a first and second set of keys from a first and second dimension table, respectively are received from a host system. Each of the first and second set of keys is associated with at least one fact table. A set of locations associated with a set of foreign key indexes are received from the host system. A set of fact table indexes are traversed. At least a first set of Row Identifiers (“RIDs”) associated with the first set of keys and at least a second set of RIDs associated with the second set of keys are identified. An operation is performed on the first and second sets of RIDs to identify an intersecting set of RIDs. The intersecting set of RIDs are then stored.

Abstract: Techniques are provided for processing a query using an index. It is determined whether a number of consecutive keys that evaluate to a false-positive when applying a screening predicate in the query exceeds a false-positive predefined value. One or more keys in the index are skipped in response to determining that the number of consecutive keys exceeds the false-positive predefined value.