In this topic we see how the number of dimensions affects session memory. In each of the following tests one measures is defined and no rules are used. The number of dimensions varies from one test to the next.
Tests
Non-MODEL Query: select 'x' as x from t_100k ; Test Complete
1 Dimension: select x from t_100k model dimension by( key_1 ) measures( 'x' as x ) () ; Test Complete
2 Dimensions: select x from t_100k model dimension by( key_1, key_2 ) measures( 'x' as x ) () ; Test Complete
3 Dimensions: select x from t_100k model dimension by( key_1, key_2, key_3 ) measures( 'x' as x ) () ; Test Complete
Results
Non-MODEL
Metric Query 1 Dimension 2 Dimensions 3 Dimensions
-------------------------- --------- ----------- ------------ ------------
session pga memory max (K) 256 12,160 23,936 33,024
session uga memory max (K) 248 11,877 23,266 32,028
See Statistics Descriptions and Latches and Internal Locks for metric descriptions.
Observations
Session memory increased as the number of dimensions increased, even though no rules were defined.
Warning: Results on your own systems with your own data will differ from these results. Results will even differ from one set of test runs to the next on the same machine. Run your own tests and ave rage the results from multiple runs before making performance decisions.
