The following major features are new in this release.

• Automatic In-Memory

  This feature uses Heat Map data, column usage statistics, and other relevant information to manage the contents of the IM column store. If a populate job fails because the IM column store is full, then Automatic In-Memory evicts inactive segments to make room for the active segments.

  See "Configuring Automatic In-Memory".

• In-Memory Dynamic Scans

  IM dynamic scans automatically and transparently parallelize table scans by using lightweight process threads. Oracle Resource Manager allocates these threads when it perceives that CPU resources are idle and can be leveraged to speed up the query.

  See "In-Memory Dynamic Scans".

• IM expressions window capture

  You can define an expression capture window of an arbitrary length, which ensures that only the expressions occurring within this window are considered for In-Memory materialization. This mechanism is especially useful when you know of a small interval that is representative of the entire workload. For example, during the trading window, a brokerage firm can gather the set of expressions, and materialize them in the IM column store to speed-up future query processing for the entire workload.

  See "Expression Capture Interval".

• In-Memory support for external tables

  You can populate external tables into the IM column store. This feature is useful for analytical queries that combine internal and external data.

  See "In-Memory External Tables" and "Creating and Populating External Tables: Examples".

• In-Memory Optimized Arithmetic

  For tables compressed with QUERY LOW, NUMBER columns are encoded using an optimized format that enables native calculations in hardware. SIMD vector processing of aggregations and arithmetic operations that use this format can achieve significant performance gains. The feature is enabled when INMEMORY_OPTIMIZED_ARITHMETIC is set to ENABLE.

  See "Optimizing In-Memory Arithmetic".

• Enhanced performance for Large Objects (LOBs)

  In previous releases, although LOBs and LOB pointers were populated in the IM column store, the database satisfied queries by using the buffer cache. In this release, In-Memory queries that apply range predicates to scalar columns or SQL operators to LOB columns benefit from SIMD vector processing.

  The IM column store provides contiguous storage for inline LOBs, which are LOBs less than 4 KB, within the IMCUs. For out-of-line LOBs, the IM column store only stores the LOB locator, which is 40 byes. There is one exception to the preceding rule. An IMEU can allocate up to 32 KB of contiguous storage for JSON columns defined as a LOB data type. The IMEU stores these columns in the OSON (binary JSON) format.

  See "CPU Architecture: SIMD Vector Processing".

• In-Memory join group on one column

  You can use the following syntax to create a join group for a self-join on a single column: CREATE INMEMORY JOIN GROUP jg_name(table_name(column_name)).

  See "Optimizing Joins with Join Groups".