Batch Performance Issues Due To Hibernate Cache On Algorithms (Doc ID 2563370.1)

Last updated on AUGUST 02, 2019

Applies to:

Symptoms

On RMB v2.6.0.1.0 against FW v4.3.0.4.0:
We are facing performance issue whenever we have a hibernate batched data processed in Algorithms. This hibernate batched process took 6 hrs to create approximately. 2000 transactions.
Auto flush triggered by framework takes maximum of time due to which processing in JVM slows down, Primary analysis indicates the Hibernate cache is slowing down as it holds uncommitted entities, eventually causing batch performance issues.

Example :
  ORMB Batch is multithreaded.
  In GetJobWork - Master data id's are passed to threadpoolworkers as workunits.
  In executeWorkUnits - We do BO status transition on this master data, which invokes BO Life Cycle on these entities.
  Life cycle further invoke respective status algorithms and processes hibernate batched inserts/updates (creating ft's etc)

This is single commit process. Initially processing is faster but gradually when uncommitted data increases in VM, Auto flush starts running slow. We are facing this issue in multiple ORMB batches (including consulting and product) where hibernate batched operation is slowed down due to auto flush.

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.