Decision-Support Parameters

When you configure virtual shared memory on your system, you must decide what portion to reserve for decision-support queries. Decision-support queries use large amounts of the virtual portion of shared memory to perform joins and sort operations.

On 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.

Use the following parameters to control how decision-support queries are processed and to control the amount of memory that the database server allocates to decision-support queries. For more information about tuning these configuration parameters, refer to your IBM Informix: Performance Guide.

Configuration
Parameter

Description

DATASKIP

Controls whether the database server skips an unavailable table fragment.

DS_ADM_POLICY

Specifies how the Resource Grant Manager (RGM) should schedule queries.

DS_MAX_QUERIES

Specifies the maximum number of queries that can run concurrently.

DS_TOTAL_MEMORY

Specifies the amount of memory available for PDQ queries.

Set the DS_TOTAL_MEMORY configuration parameter to any value not greater than the quantity (SHMVIRTSIZE - 10 megabytes).

MAX_PDQPRIORTY

Limits the amount of resources that a query can use.

OPTCOMPIND

Advises the optimizer on an appropriate join strategy for your applications.

PDQPRIORTY

Requests an amount of memory that a query can use.

If your communication interface between nodes requires configurable buffers, you also need to consider the amount of space that these message buffers take up in the virtual portion. For more details on these configurable buffers, refer to your machine notes file.

For DSS-only applications that do not need to balance resources against OLTP applications, you can allocate all of the virtual portion to your decision-support queries. Set the DS_TOTAL_MEMORY configuration parameter to any value not greater than the quantity (SHMVIRTSIZE - 10 megabytes).

For strategies to improve performance with fragmentation and PDQ, refer to your IBM Informix: Performance Guide.