Sap Se

Sap Ag

Yongsik Yoon

Yong Sik Kwon

Sang Kyun Cha

Chanho Jeong

Juchang Lee

Chang Bin Song

Joo Yeon Lee

Sung Heun Wi

Oliver M Steinau

Sang Il Song

Markus Hartmann

Ki Hong Kim

Boyung Lee

Jeong Ae Han

Christoph Weyerhaeuser

Tobias Mindnich

Johannes Merx

Taehyung Lee

Junhee Woo

Norman May

Di Wu

Alexander Boehm

Joo Young Yoon

Won Seok Kim

Sanghyun Choi

An in-memory database server hosting a tenant of a multi-tenant software architecture can receive a definition of a custom data field that is unique to an organization having isolated access to the tenant. The custom data field can extend a standard table defined by central metadata stored at a system tenant of the multi-tenant software architecture. Tenant private metadata that includes the definition can be stored in memory accessible only to the tenant. A tenant-dependent table that includes the custom data field can be formed, for example by retrieving central metadata defining the standard table from the system tenant and adding the custom data field using the definition. The tenant-dependent table can be presented for access via a database client at the organization. Related systems, articles of manufacture, and computer-implemented methods are disclosed.

In-memory database for multi-tenancy

Methods and apparatus, including computer program products, are provided for a case join. In one aspect, there is provided method, which may include receiving a query for a predefined view that is configured to inhibit modification; detecting whether the predefined view includes a database table extended to include an additional column; and generating, in response to the query, a view based on a case join, when the predefined view includes the database table extended to include the additional column. Related apparatus, systems, methods, and articles are also described.

Case join decompositions

A hybrid database table is stored as both a row and a column store. One or more techniques may be employed alone or in combination to enhance performance of the hybrid table by regulating access to, and/or the size of, the processing-intensive column store data. For example during an insert operation, the column store data may be searched for a uniqueness violation only after certain filtering and/or boundary conditions have been considered. In another technique, a hybrid table manager may control movement of data to the column store based upon considerations such as frequency of access, or underlying business logic. In still another technique, querying of the hybrid table may result in a search of the column store data only after an initial search of row store data fails to return a result.

Hybrid database table stored as both row and column store

A window operator can be processed according to a variety of techniques that introduce parallelism. Window function sub-results can be calculated separately on different nodes. Overall superior performance can result. Skewness in input data can be accounted for by controlling a degree of parallelism at nodes.

Parallelized execution of window operator

Disclosed herein are system, method, and computer program product embodiments for utilizing join cardinality to improve the performance of query processing. An embodiment operates by a database engine receiving a query including a join operation on a first and a second table using a join predicate, and a tag for the join operation indicating a cardinality relationship between the first and the second table when satisfying the join predicate. Then, the database engine generates a query processing plan for performing the query. For the join operation, the database engine selects one of the first or the second tables to remove from the join operation based on a type of the join operation and the tag. Based on the selecting, the database engine may remove may the join operation and the selected table from the query processing plan.

Query optimization using join cardinality

Embodiments relate to view caching techniques that cache for a limited time, some of the (intermediate) results of a previous query execution, in order to avoid expensive re-computation of query results. Particular embodiments may utilize a cache manager to determine whether information relevant to a subsequent user request can be satisfied by an existing cache instance or view, or whether creation of an additional cache instance is appropriate. At design time, cache defining columns of a view are defined, with user input parameters automatically being cache defining. Cache instances are created for each tuple of literals for the cache defining columns, and for each explicit or implicit group by clause. Certain embodiments may feature enhanced reuse between cache instances, in order to limit memory footprint. Over time a cache instances may be evicted from memory based upon implementation of a policy such as a Least Recently Used (LRU) strategy.

Cached views

A mixed join between database column and row tables employs an algorithm that recognizes both row and column store, and is executable upon the data in its native form (row or column) without requiring conversion between orientations. The native mixed join algorithm accesses the column dictionary of the column table for efficient join processing. The native mixed join algorithm may also exploit an inverted index (if present) to search a record (e.g. docid) with a given value. In particular, the native mixed join algorithm looks up a column dictionary for a join condition, while iterating the row table and returning matched records in a pipelined manner.

Mixed join of row and column database tables in native orientation

Partition-aware calculation of a window operator can be supported. Different nodes can calculate window function sub-results on database partitions locally, in parallel and independently. Recognition of scenarios in which such parallelism is permissible can be performed. Overall superior performance can result.

Partition-aware distributed execution of window operator

According to an aspect, a database system for integrating calculation models into execution plans includes a first engine configured to parse a query to be applied on a database. The first engine is configured to invoke a second engine during query compilation. The second engine is configured to instantiate a calculation model based on the query, and the second engine is configured to derive a converted calculation model by converting the calculation model into a format compatible with the first engine. The first engine is configured to incorporate the converted calculation model into an execution plan during the query compilation and execute the query on the database according to execution plan.

Integration of calculation models into SQL layer

In accordance with aspects of the disclosure, systems and methods are provided for query optimization including generating a knowledge based container for one or more new operators, building a generic operator structure for the knowledge based container generated for each new operator, and applying optimization rules to each new operator based on the generic operator structure built for each new operator to thereby increase extensibility of the query optimizer using a structured query language (SQL).

Generic operator framework

Embodiments relate to re-writing database query plans, and visualizing such re-written query plans. A query re-write framework includes a query normalization engine in communication with a rule catalog comprising query re-write rules in the form of rule classes. The framework receives as input, a query plan graph to be re-written. Based upon the engine's application of re-write rules from the catalog, the framework produces a re-written query plan graph as output. An interface component of the framework may provide a visualization of the re-written query plan graph as part of a dashboard. A user may access the framework to enable/disable existing rules in the catalog, add new rules to the catalog, and/or control a sequence and a precedence in which rules are applied to re-write the query plan. A user may interact with the visualization of the re-written query plan for purposes of de-bugging, re-write optimization, and/or query development.

Growing community of inventors

Seocho-ku, South Korea

Yongsik Yoon