757318b5cd
Unified benchmarks into one for JDBC and JPA. They're now both testing the data access with the plain infrastructure (JdbcTemplate and EntityManager) plus interactions via Spring Data repositories on top of that. Added build profiles to run the same benchmarks against an in-memory H2, a locally running H2 and a locally running Postgres. See the readme for setup instructions. The JdbcFixture contains code to optionally disable event publication for the JDBC operation. Uncomment the line invoking disableEntityCallbacks(…). Changed the benchmarks to use 10 warmup iterations instead of 5 to make sure we're at stable numbers once the measurement happens.
35 lines
1.7 KiB
Plaintext
35 lines
1.7 KiB
Plaintext
= Benchmarks for relational data access with Spring Data
|
|
|
|
This benchmark evaluates various options of relation data access in the Spring Data project family:
|
|
|
|
- JDBC and Spring Data JDBC
|
|
- JPA and Spring Data JPA
|
|
|
|
The primary purpose of the benchmark is to help the team detect degradations in performance quickly or verify optimizations made in various areas of the libraries.
|
|
It also helps justifying differences between numbers in rather clean room contexts (an embedded database) and scenarios that use a more realistic setup like a locally running database.
|
|
That difference alone will help reasoning about the real-world impact of an optimization or degradation.
|
|
|
|
== Benchmark model
|
|
|
|
The benchmarks use a very simple model of a book with a title and pages attribute.
|
|
We deliberately chose a simple model as the benchmarks are supposed to measure the overhead the Spring Data mapping and repository infrastructure adds on top of the raw JDBC and JPA alternatives.
|
|
There are two major benchmark operations:
|
|
|
|
1. Finding all books (8 items)
|
|
2. Finding a single book by title.
|
|
|
|
There are different flavors of those operations to measure the impact of different setups to execute them:
|
|
|
|
- the effect of read-only transactions in the find all case
|
|
- the difference between derived and declared queries in JPA
|
|
|
|
== Infrastructure
|
|
|
|
The benchmarks are run against the following databases:
|
|
|
|
- In-memory H2
|
|
- A locally running H2 (port 9092, database name `benchmark`, user `sa`, empty password)
|
|
- A locally running Postgres (port 5432, database name `benchmark`, no user, no password)
|
|
|
|
The settings can be adapted by tweaking corresponding `application-$database.properties` file in `src/main/resources`.
|
|
|