Database Server Use of PDQ

In Extended Parallel Server, parallel execution automatically occurs when the database operation involves data that is fragmented across multiple dbspaces and multiple CPU VPs are available. Parallel processing can occur on both a single coserver and across multiple coservers.

• Single coserver execution

  Extended Parallel Server executes database operations in parallel when:

  • The involved tables are fragmented across separate dbspaces on separate disks that are local to one coserver.
  • An SQL operator in the query plan processes a large amount of data so that it dynamically allocates multiple threads to execute in parallel across available CPU VPs on the (single) local coserver.
• Multiple coserver execution

  Extended Parallel Server achieves a high degree of parallelism for a query when the involved tables are fragmented across all coservers and multiple CPU VPs on multiple coservers execute the SQL operators within the query plan.

Extended Parallel Server executes the following database operations in parallel when the involved tables are fragmented into separate dbspaces:

• DSS queries, which are usually complex SELECT statements that involve a large number of rows to scan, join, aggregate, sort, or group
• INSERT, DELETE, and UPDATE statements that process nonlocal data
• SPL routines
• Sorts
• Index builds
• Update statistics
• onutil commands, such as CHECK DATA and CHECK INDEX command options
• Uncorrelated subqueries
• Correlated subqueries if they cannot be unnested

For more information on how Extended Parallel Server uses PDQ to execute these database operations, refer to your IBM Informix: Performance Guide.