Allocation of resources for concurrent query execution via adaptive segmentation

1. An automated method to allocate resources of a highly parallel computing system for processing a database query, the method comprising:

receiving the database query from an application at a client system;

compiling the query and computing the number of executive server processes (ESPs) in each ESP layer of the query;

generating an affinity value, wherein the affinity value specifies a subset of processors for a range of processor subset sizes;

executing the query wherein placement of ESP layers of the query onto processors of the computing system is determined using the affinity value; and

returning results of the execution to the application.

2. The method of claim 1 , further comprising computing a maximum degree of parallelism for the query.

3. The method of claim 2 , wherein said computing of the maximum degree of parallelism includes computing an estimated memory resource required by the query and an estimated CPU resource required by the query.

4. The method of claim 1 , wherein an interleaved ordering of processors is used in forming CPU subsets.

5. The method of claim 1 , wherein the subset of processors specified by the affinity value is given by a variable multiplied by a skip value plus an offset value, where the variable goes from zero to the number of executive server processes in an ESP layer minus one.

6. The method of claim 1 , wherein cyclical placement is used when placing ESP layers onto CPU subsets.

7. The method of claim 1 , wherein the affinity value is based on a current S run-time state of the system.

8. The method of claim 1 , wherein the affinity value is randomly generated.

9. The method of claim 1 , wherein the affinity value is based on a source of the query.

10. A highly parallel computing apparatus for an enterprise data warehouse system or a business intelligence system, the apparatus comprising:

a plurality of processors in the highly parallel computing apparatus;

memory resources in the highly parallel computing apparatus;

a database server configured to receive a database query from an application at a database client system;

a query compiler configured to prepare an execution plan for the query and compute a number of executive server processes (ESPs) in each ESP layer of the query;

a workload management system configured to generate an affinity value based on the current run-time state of the apparatus, wherein the affinity value specifies a subset of processors for a range of processor subset sizes; and

a query executor configured to execute the query wherein placement of ESP layers of the query onto processors of the computing system is determined using the affinity value.

11. The apparatus of claim 10 , further comprising a query optimizer module configured to compute a maximum degree of parallelism for the query.

12. The apparatus of claim 11 , wherein said computing of the maximum degree of parallelism includes computing an estimated memory resource required by the query and an estimated processor resource required by the query.

13. The apparatus of claim 10 , wherein an interleaved ordering of processors is used in forming processor subsets.

14. The apparatus of claim 10 , wherein the subset of processors specified by the affinity value is given by a variable multiplied by a skip value plus an offset value, where the variable goes from zero to the number of executive server processes in an ESP layer minus one.

15. The apparatus of claim 10 , wherein cyclical placement is used when placing ESP layers onto CPU subsets.

16. The apparatus of claim 10 , wherein the affinity value is based on a current run-time state of the apparatus.

17. The apparatus of claim 10 , wherein the affinity value is randomly generated.

18. The apparatus of claim 10 , wherein the affinity value is based on a source of the query.

19. An enterprise warehouse or business intelligence system comprising:

a plurality of processors configured to execute computer-readable instructions;

memory configured to store computer-readable instructions and data;

processor-executable code for a database server which is configured to receive a database query from an application at a database client system;

processor-executable code for a workload management system configured to generate an affinity value, wherein the affinity value specifies a subset of processors for a range of processor subset sizes;

processor-executable code for a query compiler configured to prepare an execution plan for the query and compute a number of executive server processes (ESPs) in each ESP layer of the query; and

processor-executable code for a query executor configured to execute the query wherein placement of ESP layers of the query onto processors of the computing system is determined using the affinity value.

20. The method of claim 1 , comprising:

computing an estimated memory resource (EMR) required by the query, and an estimated CPU resource (ECR) required by the query;

computing a first ratio of the EMR to an amount of memory available per CPU, and a second ratio of the ECR to an amount of work acceptable to be assigned per CPU; and

computing a maximum degree of parallelism for the query as a function of greater of the first ratio and the second ratio.

21. The apparatus of claim 11 , wherein computing the maximum degree of parallelism comprises:

computing an estimated memory resource (EMR) required by the query, and computing an estimated CPU resource (ECR) required by the query;

computing a first ratio of the EMR to a first constant representing an amount of memory available per CPU, and computing a second ratio of the ECR to a second constant representing an amount of work acceptable to be assigned per CPU; and

selecting a maximum value of the first ratio versus the second ratio.