Anshul Bansal
Posted on September 26, 2019
Problem
In general, the hibernate entities (domains) are set to use database sequence as Id generator.
In such a case, for every insert hibernate makes two round trips to the database. It'll first make a round trip to get the next value of the sequence to set the identifier of the record. Then make another round trip to insert the record.
Assume, we want to insert 1000 records in the database. It'll result in a total of 2000 round trips to the database (1000 round trips each, to get the next value of a sequence and insert the records).
Even in a case of very low network latency, performing a few thousand inserts may require a significant amount of time.
The first major problem is to perform each insert separately.
The second major problem is to make a round trip to get the next value of the sequence every time.
Theoretical Solution
We should try to reduce the network round trips to the database for bulk inserts by batch processing of inserts.
Along with that, reduce the network round trips to get the next value of the sequence for every insert.
Hibernate Solution
First problem:
We know the obvious.
Use the JDBC batching provided by Hibernate.
Set the following properties in the hibernate configuration file.
<property name="hibernate.jdbc.batch_size" value="100"/>
<property name="hibernate.order_inserts" value="true"/>
Now, hibernate will make a batch of 100 inserts and orders them. Then, it will make a single network round trip to the database to insert 100 records.
Therefore, the initial 1000 round trips to insert the records will reduce to 10.
Second problem:
We'll use enhance sequence identifier with an optimizer strategy like pooled or pooled-lo, which provides in-memory identifiers. Therefore, they reserve ids in memory to be used later.
Let's see what a 'pooled' optimizer strategy can do!
To enable it, we'll require to set the 'INCREMENT BY' of the database sequence to 100.
Then, set the pooled optimizer strategy with increment_size = 100 in the entity:
@Id
@GenericGenerator(
name = "sequenceGenerator",
strategy = "org.hibernate.id.enhanced.SequenceStyleGenerator",
parameters = {
@Parameter(name = "sequence_name", value = "hibernate_sequence"),
@Parameter(name = "optimizer", value = "pooled"),
@Parameter(name = "initial_value", value = "1"),
@Parameter(name = "increment_size", value = "100")
}
)
@GeneratedValue(
strategy = GenerationType.SEQUENCE,
generator = "sequenceGenerator"
)
private Long id;
Now, hibernate will make a round trip to the database and set the sequence nextval to next 100 value. And, reserve the 100 values in memory to set the ids for the inserts.
With this approach, we'll make 1 round trip to fetch the next value of the sequence for every 100 records to insert.
Hence, the initial 1000 round trips to get the next value of the sequence will reduce to 10.
Therefore applying both of the solutions, we can reduce the round trips to just 20 for 1000 inserts.
We all may know about batch insert optimization. But, the sequence optimizer is quite a winner.
Not many of us know this kind of optimization strategy available already.
Isn't it sound cool and smart.
Guys, Please do share your thoughts with me.
Posted on September 26, 2019
Join Our Newsletter. No Spam, Only the good stuff.
Sign up to receive the latest update from our blog.