Comparison of DataNucleus with H2 embedded vs OpenJPA with Derby embedded

Each of the following tables focuses on a specific database operation, where the last table presents average results comparison.

Speed comparison of JPA database persistence operations (normalized score, higher is better)

Transaction Size =>	Few Entities		Many Entities		Average Score
	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded
Basic Person Test	19.2	4.8	8.0	2.9	13.6	3.9
Element Collection Test	6.6	2.8	2.8	1.9	4.7	2.4
Inheritance Test	15.6	4.1	4.4	2.4	10.0	3.3
Indexing Test	24.4	6.1	11.8	4.1	18.1	5.1
Graph (Binary Tree) Test	9.2	1.5	6.2	1.4	7.7	1.4
Multithreading Test	34.9	8.7	7.1	3.3	21.0	6.0
All Tests	18.3	4.7	6.7	2.7	12.5	3.7

The results above show that in general DataNucleus with H2 embedded is much more efficient than OpenJPA with Derby embedded in persisting JPA entity objects to the database. Comparing the normalized speed of OpenJPA with Derby embedded database (3.7) to the normalized speed of DataNucleus with H2 embedded database (12.5) reveals that in these tests, DataNucleus with H2 embedded is 3.4 times faster than OpenJPA with Derby embedded.

A large performance gap has been detected when using graphs of objects with small transaction size. Comparing the normalized speed of OpenJPA with Derby embedded database (1.5) to the normalized speed of DataNucleus with H2 embedded database (9.2) reveals that in that case, DataNucleus with H2 embedded is 6.1 times faster than OpenJPA with Derby embedded.

Speed comparison of JPA database retrieval operations (normalized score, higher is better)

Retrieval Size =>	Few Entities		Many Entities		Average Score
	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded
Basic Person Test	13.2	14.8	13.5	22.6	13.3	18.7
Element Collection Test	6.4	0.0031	8.9	2.3	7.7	1.2
Inheritance Test	12.2	0.015	12.7	7.0	12.4	3.5
Indexing Test	8.1	11.7	16.5	26.5	12.3	19.1
Graph (Binary Tree) Test	10.6	0.93	20.9	1.2	15.8	1.1
Multithreading Test	13.3	25.2	9.9	28.6	11.6	26.9
All Tests	10.6	8.8	13.7	14.7	12.2	11.7

The results above show that in general DataNucleus with H2 embedded is slightly more efficient than OpenJPA with Derby embedded in retrieving JPA entity objects from the database.

A huge performance gap has been detected when using JPA element collections with small retrieval size. Comparing the normalized speed of OpenJPA with Derby embedded database (0.0031) to the normalized speed of DataNucleus with H2 embedded database (6.4) reveals that in that case, DataNucleus with H2 embedded is 2,065 times faster than OpenJPA with Derby embedded.

On the other hand, DataNucleus with H2 embedded is slower, for instance, when using multithreading with large retrieval size. Comparing the normalized speed of DataNucleus with H2 embedded database (9.9) to the normalized speed of OpenJPA with Derby embedded database (28.6) reveals that in that case, DataNucleus with H2 embedded is 2.9 times slower than OpenJPA with Derby embedded.

Speed comparison of JPA database query operations (normalized score, higher is better)

Retrieval Size =>	Few Entities		Many Entities		Average Score
	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded
Basic Person Test	4.7	49.7	1.4	9.9	3.1	29.8
Element Collection Test	5.0	2.7	2.4	9.6	3.7	6.1
Inheritance Test	3.8	2.3	1.6	1.3	2.7	1.8
Indexing Test	0.0041	7.0	1.8	17.1	0.91	12.1
Multithreading Test	1.2	40.9	0.79	11.7	1.0	26.3
All Tests	3.0	20.5	1.6	9.9	2.3	15.2

The results above show that in general OpenJPA with Derby embedded is much more efficient than DataNucleus with H2 embedded in executing the tested JPA queries. Comparing the normalized speed of DataNucleus with H2 embedded database (2.3) to the normalized speed of OpenJPA with Derby embedded database (15.2) reveals that in these tests, OpenJPA with Derby embedded is 6.6 times faster than DataNucleus with H2 embedded.

A huge performance gap has been detected when using database indexes with small retrieval size. Comparing the normalized speed of DataNucleus with H2 embedded database (0.0041) to the normalized speed of OpenJPA with Derby embedded database (7.0) reveals that in that case, OpenJPA with Derby embedded is 1,707 times faster than DataNucleus with H2 embedded.

Speed comparison of JPA database update operations (normalized score, higher is better)

Transaction Size =>	Few Entities		Many Entities		Average Score
	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded
Basic Person Test	10.0	5.6	3.9	5.2	6.9	5.4
Element Collection Test	7.6	0.0090	3.9	2.2	5.8	1.1
Inheritance Test	9.4	0.045	4.4	4.6	6.9	2.3
Indexing Test	10.6	6.1	5.9	8.7	8.2	7.4
Graph (Binary Tree) Test	5.9	1.6	3.7	0.91	4.8	1.3
Multithreading Test	failed	18.3	failed	4.5	failed	11.4
All Tests	8.7	5.3	4.4	4.4	6.5	4.8

DataNucleus with H2 embedded has failed in 2 tests (see exceptions).

The results above show that in general DataNucleus with H2 embedded is more efficient than OpenJPA with Derby embedded in updating JPA entity objects in the database.

A huge performance gap has been detected when using JPA element collections with small transaction size. Comparing the normalized speed of OpenJPA with Derby embedded database (0.0090) to the normalized speed of DataNucleus with H2 embedded database (7.6) reveals that in that case, DataNucleus with H2 embedded is 844 times faster than OpenJPA with Derby embedded.

Speed comparison of JPA database removal operations (normalized score, higher is better)

Transaction Size =>	Few Entities		Many Entities		Average Score
	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded
Basic Person Test	4.2	4.2	1.3	4.5	2.8	4.3
Element Collection Test	4.9	0.0071	1.6	1.6	3.2	0.80
Inheritance Test	4.0	0.035	1.3	3.0	2.7	1.5
Indexing Test	7.5	5.3	1.9	3.6	4.7	4.4
Graph (Binary Tree) Test	5.6	0.80	4.1	0.93	4.8	0.86
Multithreading Test	failed	5.7	failed	6.2	failed	6.0
All Tests	5.2	2.7	2.0	3.3	3.6	3.0

DataNucleus with H2 embedded has failed in 2 tests (see exceptions).

The results above show that in general DataNucleus with H2 embedded is slightly more efficient than OpenJPA with Derby embedded in deleting JPA entity objects from the database.

A huge performance gap has been detected when using JPA element collections with small transaction size. Comparing the normalized speed of OpenJPA with Derby embedded database (0.0071) to the normalized speed of DataNucleus with H2 embedded database (4.9) reveals that in that case, DataNucleus with H2 embedded is 690 times faster than OpenJPA with Derby embedded.

On the other hand, DataNucleus with H2 embedded is slower, for instance, when using simple basic entities with large transaction size. Comparing the normalized speed of DataNucleus with H2 embedded database (1.3) to the normalized speed of OpenJPA with Derby embedded database (4.5) reveals that in that case, DataNucleus with H2 embedded is 3.5 times slower than OpenJPA with Derby embedded.

Comparison of database storage efficiency (normalized score, higher is better)

Transaction Size =>	Few Entities		Many Entities		Average Score
	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded
Basic Person Test	27.0	52.8	27.0	52.8	27.0	52.8
Element Collection Test	22.4	42.4	22.4	42.4	22.4	42.4
Inheritance Test	27.0	44.2	27.0	44.2	27.0	44.2
Indexing Test	31.6	53.1	32.1	54.0	31.9	53.5
Graph (Binary Tree) Test	20.0	12.8	15.0	9.6	17.5	11.2
Multithreading Test	33.9	50.7	9.2	9.7	21.5	30.2
All Tests	27.0	42.7	22.1	35.5	24.6	39.1

The results above show that in general OpenJPA with Derby embedded is more efficient than DataNucleus with H2 embedded in using disk space.

Comparison of JPA/Database speed - the averages (normalized score, higher is better)

Transaction/Retrieval Size	Few Entities		Many Entities		Average Score
	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded	DataNucleus H2 embedded	OpenJPA Derby embedded
Basic Person Test	10.3	15.8	5.6	9.0	7.9	12.4
Element Collection Test	6.1	1.1	3.9	3.5	5.0	2.3
Inheritance Test	9.0	1.3	4.9	3.7	6.9	2.5
Indexing Test	10.1	7.2	7.6	12.0	8.9	9.6
Graph (Binary Tree) Test	7.8	1.2	8.7	1.1	8.3	1.2
Multithreading Test	16.5	19.8	5.9	10.9	11.2	15.3
All Tests	9.6	8.0	6.0	6.9	7.8	7.4

The results above show that in general DataNucleus with H2 embedded is slightly more efficient than OpenJPA with Derby embedded in performing JPA database operations.

A large performance gap has been detected when using graphs of objects with large transaction/retrieval size. Comparing the normalized speed of OpenJPA with Derby embedded database (1.1) to the normalized speed of DataNucleus with H2 embedded database (8.7) reveals that in that case, DataNucleus with H2 embedded is 7.9 times faster than OpenJPA with Derby embedded.

JPA Performance Benchmark (JPAB)