11 Performance Guidelines

This section discusses performance guidelines for applications that use LOB data types.

Note:

From release 23ai onwards, you can experience improved read and write performance for LOBs due to the following enhancements:

• Multiple LOBs in a single transaction are buffered simultaneously. This improves performance when you use mixed workload in a transaction. Mixed workload refers to switching between LOBs while writing within a single transaction. Let's consider that you write to LOB1, then you write to LOB2, and then you want to write again to LOB1 in a single transaction. LOB1 and LOB2 are buffered simultaneously, which provides better throughput and minimizes space fragmentation.
• Various enhancements, such as acceleration of compressed LOB append and compression unit caching, improve the performance of reads and writes to compressed LOBs.
• The input-output buffer is adaptively resized based on size of the input data for large writes to LOBs with the NOCACHE option. This improves the performance for large direct writes, such as writes to file systems on DBFS and OFS.

• LOB Performance Guidelines
  This section provides performance guidelines while using LOBs through Data Interface or LOB APIs.
• Moving Data to LOBs in a Threaded Environment
  Learn about the recommended procedure to follow while moving data to LOBs in this section.
• LOB Access Statistics
  Three session-level statistics specific to LOBs are available to users: LOB reads, LOB writes, and LOB writes unaligned.