Workload-based sampling

1. A system comprising:

a memory storing processor-executable program code; and

a processing unit to execute the processor-executable program code in order to cause the system to:

determine a plurality of queries of a workload;

determine a data source comprising a plurality of data rows; and

determine a sample data source based on a cardinality of each of the plurality of queries with respect to the data source and an estimated cardinality of each of the plurality of queries with respect to the data source by:

for each of the plurality of queries, determining a decrease in a cardinality estimation error associated with addition of each of candidate rows of the data source to the sample data source; and

selecting a candidate row to add to the sample data source based on the determined decreases,

wherein the estimated cardinality of a query with respect to the data source is determined based on the sample data source.

2. A system according to claim 1 , the processing unit to execute the processor-executable program code in order to cause the system to:

receive a runtime query on the data source;

determine an estimated cardinality of the runtime query with respect to the data source based on the sample data source; and

determine a query execution plan for the runtime query based on the estimated cardinality of the runtime query with respect to the data source.

3. A system according to claim 1 , wherein each of the plurality of queries is associated with one or more predicates, and

wherein the candidate rows associated each one of the plurality of queries are rows of the data source selected by the one or more predicates of the query.

4. A system according to claim 3 , wherein determination, for one of the plurality of queries, of a decrease in a cardinality estimation error associated with addition of a candidate row comprises:

determination of a true cardinality of the query with respect to the data source by execution of the query on the data source;

determination of a current estimated cardinality of the query with respect to the data source by execution of the query on the sample data source not including the candidate row;

determination of a current cardinality estimation error based on the true cardinality and the current estimated cardinality;

determination of a new estimated cardinality of the query with respect to the data source by execution of the query on the sample data source including the candidate row;

determination of a new cardinality estimation error based on the true cardinality and the new estimated cardinality; and

determination of the decrease in the cardinality estimation error based on the current cardinality estimation error and the new cardinality estimation error.

5. A system according to claim 1 , wherein determination, for one of the plurality of queries, of a decrease in a cardinality estimation error associated with addition of a candidate row comprises:

determination of a true cardinality of the query with respect to the data source by execution of the query on the data source;

determination of a current estimated cardinality of the query with respect to the data source by execution of the query on the sample data source not including the candidate row;

determination of a current cardinality estimation error based on the true cardinality and the current estimated cardinality;

determination of a new estimated cardinality of the query with respect to the data source by execution of the query on the sample data source including the candidate row;

determination of a new cardinality estimation error based on the true cardinality and the new estimated cardinality; and

determination of the decrease in the cardinality estimation error based on the current cardinality estimation error and the new cardinality estimation error.

6. A computer-implemented method, comprising:

determining a plurality of queries;

determining a data source comprising a plurality of data rows; and

determining a sample data source comprising a plurality of the plurality of data rows based on a cardinality of each of the plurality of queries with respect to the data source and an estimated cardinality of each of the plurality of queries with respect to the data source by:

for each of the plurality of queries, determining a decrease in a cardinality estimation error associated with addition of each of candidate rows of the data source to the sample data source; and

selecting a candidate row to add to the sample data source based on the determined decreases,

wherein the estimated cardinality of a query with respect to the data source is determined based on data rows of the sample data source.

7. A method according to claim 6 , further comprising:

receiving a runtime query on the data source;

determining an estimated cardinality of the runtime query with respect to the data source based on the sample data source; and

determining a query execution plan for the runtime query based on the estimated cardinality of the runtime query with respect to the data source.

8. A method according to claim 6 , wherein each of the plurality of queries is associated with one or more predicates, and

wherein the candidate rows associated each one of the plurality of queries are rows of the data source selected by the one or more predicates of the query.

9. A method according to claim 8 , wherein determining, for one of the plurality of queries, of a decrease in a cardinality estimation error associated with addition of a candidate row comprises:

determining a true cardinality of the query with respect to the data source by execution of the query on the data source;

determining a current estimated cardinality of the query with respect to the data source by execution of the query on the sample data source not including the candidate row;

determining a current cardinality estimation error based on the true cardinality and the current estimated cardinality;

determining a new estimated cardinality of the query with respect to the data source by execution of the query on the sample data source including the candidate row;

determining a new cardinality estimation error based on the true cardinality and the new estimated cardinality; and

determining the decrease in the cardinality estimation error based on the current cardinality estimation error and the new cardinality estimation error.

10. A method according to claim 6 , wherein determining, for one of the plurality of queries, of a decrease in a cardinality estimation error associated with addition of a candidate row comprises:

determining a true cardinality of the query with respect to the data source by execution of the query on the data source;

determining a current estimated cardinality of the query with respect to the data source by execution of the query on the sample data source not including the candidate row;

determining a current cardinality estimation error based on the true cardinality and the current estimated cardinality;

determining a new estimated cardinality of the query with respect to the data source by execution of the query on the sample data source including the candidate row;

determining a new cardinality estimation error based on the true cardinality and the new estimated cardinality; and

determining the decrease in the cardinality estimation error based on the current cardinality estimation error and the new cardinality estimation error.

11. A non-transitory computer-readable medium storing program code executable by a processing unit to:

determine a plurality of queries of a workload; and

determine a sample data source based on a cardinality of each of the plurality of queries with respect to a data source comprising a plurality of data rows and an estimated cardinality of each of the plurality of queries with respect to the data source by:

for each of the plurality of queries, determination of a decrease in a cardinality estimation error associated with addition of each of candidate rows of the data source to the sample data source; and

selection of a candidate row to add to the sample data source based on the determined decreases,

wherein the estimated cardinality of a query with respect to the data source is determined based on the sample data source.

12. A medium according to claim 11 , the program code executable by a processing unit to:

receive a runtime query on the data source;

determine an estimated cardinality of the runtime query with respect to the data source based on the sample data source; and

determine a query execution plan for the runtime query based on the estimated cardinality of the runtime query with respect to the data source.

13. A medium according to claim 11 , wherein each of the plurality of queries is associated with one or more predicates, and

wherein the candidate rows associated each one of the plurality of queries are rows of the data source selected by the one or more predicates of the query.

14. A medium according to claim 13 , wherein determination, for one of the plurality of queries, of a decrease in a cardinality estimation error associated with addition of a candidate row comprises:

determination of a true cardinality of the query with respect to the data source by execution of the query on the data source;

determination of a current estimated cardinality of the query with respect to the data source by execution of the query on the sample data source not including the candidate row;

determination of a current cardinality estimation error based on the true cardinality and the current estimated cardinality;

determination of a new estimated cardinality of the query with respect to the data source by execution of the query on the sample data source including the candidate row;

determination of a new cardinality estimation error based on the true cardinality and the new estimated cardinality; and

determination of the decrease in the cardinality estimation error based on the current cardinality estimation error and the new cardinality estimation error.

15. A medium according to claim 11 , wherein determination, for one of the plurality of queries, of a decrease in a cardinality estimation error associated with addition of a candidate row comprises:

determination of a true cardinality of the query with respect to the data source by execution of the query on the data source;

determination of a current estimated cardinality of the query with respect to the data source by execution of the query on the sample data source not including the candidate row;

determination of a current cardinality estimation error based on the true cardinality and the current estimated cardinality;

determination of a new estimated cardinality of the query with respect to the data source by execution of the query on the sample data source including the candidate row;

determination of a new cardinality estimation error based on the true cardinality and the new estimated cardinality; and

determination of the decrease in the cardinality estimation error based on the current cardinality estimation error and the new cardinality estimation error.