Accessing Data Through Hibernate; What DBAs Should Tell Developers and Vice Versa

Accessing Data Through
Hibernate; What DBAs Should
Tell Developers and Vice Versa
Marco Tusa and Francisco Bordenave
Percona

3
Agenda
1. Basic description of Hibernate architecture
2. Basic description of MySQl architecture
3. What is an Object entity in Hibernate?
4. What it means for a DB?
5. JDBC
6. Simple example with One table (employees) as object
7. What it is employees as DB object?
8. CRUD example with SQL generation
9. What is the overhead in using Hibernate?
10.What is a composite object and what is the impact?
11.Can we do better?

4
Introduction / Disclaimer
the important … is to have clear ideas and coordinated direction What is this fish?
I didn’t ask for sushi
Java?
Why is he talking
about coffee, I
ordered sushi!

5
Hibernate Basic Architecture
Benefit from developer’s point of view:
● Object-relational mapping (ORM)
● Makes it easy to access data
● Rows become objects
● Don’t have to deal with how data is accessed
● Data persist in the application after query
● I can access multiple sources with
same logic/code
In short my life is easier without dealing with SQL

6
DBA
Let’s see a very high level and basic description of internals

9
DBA
Now being serious
MySQL is the RDBMS
InnoDB is the engine
Data is organized in 16kb pages
Clustered Primary Key and then secondary indexes

1
0
DBA
Now being serious
MySQL is the RDBMS
InnoDB is the engine
Data is organized in 16kb pages
Clustered Primary Key and then secondary indexes
InnoDB supports transactions (i.e. BEGIN/COMMIT) and has ACID capabilities:
● Atomicity
● Consistency
● Isolation
● Durability
Isolation means transactions should not affect other transactions when running concurrently.

1
1
Data Access Using Hibernate
● So I have a DB
● My code
● JDBC
● Hibernate
If I have Hibernate why I need to know all
these additional things?
I just need:
● Hibernate SessionFactory (heavy to create
and one for Application or by data store)
● Hibernate Session, light, ephemeral,
Open-close it inside nuclear operations

1
2
What is an Entity? And States?
I need to know that I have this:

1
3
Simple example with One table employees
● Mapping definition
● Code
What is an Entity?
<hibernate-mapping package="net.tc.employees">
<class
name="Employees"
table="employees"
catalog="employees"
optimistic-lock="version">
<id
name="empNo"
type="int">
<column name="emp_no" />
<generator class="assigned" />
</id>
<property
name="birthDate"
type="date"
>
<column name="birth_date"
length="10" not-null="true" />
<property
name="firstName"
type="string"
>
<column name="first_name"
length="14" not-null="true" />
public class Employees implements java.io.Serializable {
private int empNo;
private Date birthDate;
private String firstName;
private String lastName;
private char gender;
private Date hireDate;
private Set titleses = new HashSet(0);
private Set salarieses = new HashSet(0);
private Set deptEmps = new HashSet(0);
private Set deptManagers = new HashSet(0);
public Employees() {
}
public Employees(int empNo, Date birthDate, String
firstName, String lastName, char gender, Date hireDate) {
this.empNo = empNo;
this.birthDate = birthDate;
this.firstName = firstName;
this.lastName = lastName;
this.gender = gender;
this.hireDate = hireDate;
}

DBA
So what’s an entity in relational database?

DBA
So what is an entity in a relational database?
Simple: it is a row in the table
show create table employeesG
*************************** 1. row ***************************
Table: employees
Create Table: CREATE TABLE èmployees` (
èmp_no` int(11) NOT NULL,
`birth_date` date NOT NULL,
`first_name` varchar(14) NOT NULL,
`last_name` varchar(16) NOT NULL,
`gender` enum('M','F') NOT NULL,
`hire_date` date NOT NULL,
PRIMARY KEY (èmp_no`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
1 row in set (0.01 sec)
show create table salariesG
*************************** 1. row ***************************
Table: salaries
Create Table: CREATE TABLE `salaries` (
èmp_no` int(11) NOT NULL,
`salary` int(11) NOT NULL,
`from_date` date NOT NULL,
`to_date` date NOT NULL,
PRIMARY KEY (èmp_no`,`from_date`),
KEY èmp_no` (èmp_no`),
CONSTRAINT `salaries_ibfk_1` FOREIGN KEY (èmp_no`) REFERENCES èmployees` (èmp_no`) ON DELETE CASCADE
) ENGINE=InnoDB DEFAULT CHARSET=latin1
1 row in set (0.00 sec)

1
6
Connection Pool + JDBC Pollution
Connections:
Important to know that a connection=thread in MySQL.
Thread allocates buffers (tmp_table_size, sort_buffer_size, etc) --> Thread = CPU
and Memory work.
The more the threads, the more the CPU and Memory pressure.

1
7
Connections:
Inside MySQL 2 ways of handling threads (thread_handling variable):
- 1 thread per each connection - one-thread-per-connection (thread_cache_size)
- pool of threads - loaded_dinamically (thread_pool_size) - Enterprise feature
- Percona Server has a thread pool library pool-of-threads
Rule of thumb:
- In MySQL most of use cases are handled properly with
one_thread_per_connection
- Thread pool fits only in few use cases: normally high number of threads running
very short queries.

1
8
Connections
Out of MySQL
JDBC
- API for connecting to MySQL from Java App
- Several connection pool handlers
- Hikari -> standard nowadays
- c3p0 (still common but no longer recommended)

1
9
Connection Pool:
Connection pool as external concept VS internal Connection pool
In MySQL: group of connections (threads) constantly opened and re used.
Out of MySQL: group of connections handled by framework which may or not be kept opened.
JDBC pollution (a.k.a. damn defaults):
Simple connection without optimization
/* mysql-connector-java-8.0.12 (Revision: 24766725dc6e017025532146d94c6e6c488fb8f1) */SELECT
@@session.auto_increment_increment AS auto_increment_increment, @@character_set_client AS
character_set_client, @@character_set_connection AS character_set_connection,
@@character_set_results AS character_set_results, @@character_set_server AS character_set_server,
@@collation_server AS collation_server, @@init_connect AS init_connect, @@interactive_timeout AS
interactive_timeout, @@license AS license, @@lower_case_table_names AS lower_case_table_names,
@@max_allowed_packet AS max_allowed_packet, @@net_write_timeout AS net_write_timeout,
@@query_cache_size AS query_cache_size, @@query_cache_type AS query_cache_type, @@sql_mode AS
sql_mode, @@system_time_zone AS system_time_zone, @@time_zone AS time_zone, @@transaction_isolation
AS transaction_isolation, @@wait_timeout AS wait_timeout;

2
0
Noise exists, but can be reduced:
hibernate.hikari.dataSource.cachePrepStmts">true
hibernate.hikari.dataSource.defaultFetchSize">100
hibernate.hikari.dataSource.prepStmtCacheSize">250
hibernate.hikari.dataSource.prepStmtCacheSqlLimit">2048
hibernate.hikari.dataSource.useServerPrepStmts">true
hibernate.hikari.dataSource.useLocalSessionState">true
hibernate.hikari.dataSource.rewriteBatchedStatements">true
hibernate.hikari.dataSource.cacheResultSetMetadata">true
hibernate.hikari.dataSource.cacheServerConfiguration">true
hibernate.hikari.dataSource.elideSetAutoCommits">true
hibernate.hikari.dataSource.maintainTimeStats">false
For Hibernate you must add in URL
jdbc:mysql://127.0.0.1:3306/employees?prepStmtCacheSize=250&am
p;prepStmtCacheSqlLimit=2048&useServerPrepStmts=YES&us
eLocalSessionState=YES&useSSL=false&defaultFetchSize=1
00&rewriteBatchedStatements=YES&cacheResultSetMetadata
=YES&cacheServerConfiguration=YES&elideSetAutoCommits=
YES&maintainTimeStats=false

2
2
Developer CRUD Code Hibernate
Insert
Session se = sessionFactoryEmp2.openSession();
se.setJdbcBatchSize(1);
se.beginTransaction();
int i=1;
while(++i < 500) {
Employees employee = new Employees();
employee.setBirthDate(new Date());
employee.setHireDate(new Date());
employee.setGender('M');
employee.setEmpNo(i);
employee.setFirstName("Franco" + 1);
employee.setLastName("Castagna"+i);
se.save(employee);
}
se.getTransaction().commit();
se.close();
Read
List<Employees> employees = se.createQuery("from Employees where emp_no <999 "
).list();
se.disconnect();
se.close();
Update
List<Employees> employees = se.createQuery("from Employees where emp_no <999 "
).list();
int i = 0;
Iterator it = employees.iterator();
while(it.hasNext()) {
Employees myEmp = (Employees) it.next();
try{myEmp.setHireDate(this.getSimpleDateFormat().parse("2015-"+
getRandomNumberInRange(1,12) +"-10"));}catch(Exception ae ) {ae.printStackTrace();}
se.update(myEmp);
}
se.disconnect();
se.close();

2
3
Developer CRUD Code Simple Code No Batch
Insert
Statement stmt = conn.createStatement();
int i=0;
StringBuffer sb =new StringBuffer();
String sqlHead="INSERT INTO employees
(emp_no,birth_date,first_name,last_name,gender,hire_date)
VALUES (";
stmt.execute("START TRANSACTION");
while(++i < 500) {
sb.append(i);
sb.append(",’”+
this.getSimpleDateFormat().format(new Date())
+"'");
sb.append(",'Franco"+ i +"'");
sb.append(",'Castagna"+ i +"'");
sb.append(",'M'");
sb.append(",’”+
this.getSimpleDateFormat().format(new Date())
+"'");
sb.append(")");
stmt.execute(sqlHead+sb.toString());
sb.delete(0,sb.length());
}
conn.commit();
Read
StringBuffer sb =new StringBuffer();
String sqlHead="SELECT emp_no,birth_date,first_name,last_name,gender,hire_date
FROM employees where emp_no=";
int rowReturned=0;
while(++i < 500) {
sb.append(i);
ResultSet rs = stmt.executeQuery(sqlHead+sb.toString());
HashSet employees = new HashSet();
while(rs.next()) {
employee.setBirthDate(rs.getDate("birth_date"));
employee.setHireDate(rs.getDate("hire_date"));
employee.setGender( rs.getString("gender").charAt(0));
employee.setEmpNo(rs.getInt("emp_no"));
employee.setFirstName(rs.getString("first_name"));
employee.setLastName(rs.getString("last_name"));
employees.add(employee);
}
}
Delete
Statement stmt = conn.createStatement();
int i=0;
String sqlHead="Delete from employees where emp_no=";
stmt.execute("START TRANSACTION");
while(++i < 500) {
stmt.execute(sqlHead+i);
}
conn.commit();

DBA
Where is the overhead in:
- Additional not useful SQL
- too many temporary tables on disk
- causes IO overhead
- Duplicated SQL
- too many show/set commands
- causes CPU/Memory overhead
- Not optimized SQL (like batching and so on)
- Queries coming from framework are generally not properly optimized
- Poorly optimized queries hit everything: CPU/Memory/IO

Evidence
Hibernate
# Time: 2019-05-14T11:27:20.481014Z
# User@Host: hibernatee[hibernatee] @ [127.0.0.1] Id: 1919
# Query_time: 0.000008 Lock_time: 0.000000 Rows_sent: 0 Rows_examined: 0
SET timestamp=1557833240;
# administrator command: Close stmt;
# Time: 2019-05-14T11:27:20.481484Z
# administrator command: Prepare;
# Time: 2019-05-14T11:27:20.482054Z
insert into employees.employees (birth_date, first_name, last_name, gender, hire_date,
emp_no) values ('2019-05-14', 'Franco1', 'Castagna2', 'M', '2019-05-14', 2);

Evidence
Standard call/code
INSERT INTO employees (emp_no,birth_date,first_name,last_name,gender,hire_date) VALUES
(1,'2019-05-14','Franco1','Castagna1','M','2019-05-14');
# Time: 2019-05-14T11:29:36.974274Z
INSERT INTO employees (emp_no,birth_date,first_name,last_name,gender,hire_date) VALUES
(2,'2019-05-14','Franco2','Castagna2','M','2019-05-14');

2
7
Developer - CRUD Results No Batch

What Can Be Done? - Batch
Hibernate configuration
<hibernate-configuration>
<session-factory >
<property name="sessionFactoryName">Hikari</property>
<property name="hibernate.dialect">org.hibernate.dialect.MySQLDialect</property>
<property name="hibernate.jdbc.batch_size">20</property>
Java code Hibernate related
while(i < 500) {
employee.setBirthDate(new Date());
...
se.save(employee);
if ( ++i % 20 == 0 ) { //20, same as the JDBC batch sizeflush a batch of inserts and release
memory:
se.flush();
se.clear();
}
}

What Can Be Done? - Batch
Hibernate
insert into employees.employees (birth_date, first_name, last_name, gender, hire_date,
emp_no) values ('2019-05-14', 'Franco1', 'Castagna1', 'M', '2019-05-14', 1),...,('2019-05-14',
'Franco1', 'Castagna19', 'M', '2019-05-14', 19);
# Time: 2019-05-14T11:34:15.998326Z
# administrator command: Close stmt;
# Time: 2019-05-14T11:34:15.998609Z

3
0
Developer - CRUD Results With Batch

3
1
CRUD Summary With Optimization

DBA - What is That Traffic?
● A lot of data is read from the DB over and over
● Tons of simple queries
● Queries involves also referenced tables and not only the main one

3
3
Developer - Entity and References

3
4
Developer Entity and References
Let us start with the cool things:
Employees myEmp = se.find(Employees.class,10001, LockModeType.PESSIMISTIC_WRITE);
Set salaries = null;
salaries = myEmp.getSalarieses();
Iterator itS = salaries.iterator();
while(itS.hasNext()) {
Salaries mySal = (Salaries) itS.next();
if(mySal.getToDate().toString().equals("9999-01-01")){
mySal.setSalary(mySal.getSalary() + 100);
}
}
se.saveOrUpdate(myEmp);
se.disconnect();
se.close();
Performance
ALERT

3
5
Employees myEmp = se.find(Employees.class,10001, LockModeType.PESSIMISTIC_WRITE);
● Hibernate: select employees0_.emp_no as emp_no1_3_0_... from employees.employees employees0_
where employees0_.emp_no=? for update
● Hibernate: select salarieses0_.emp_no as emp_no1_4_0_... from employees.salaries salarieses0_ where
salarieses0_.emp_no=?
● Hibernate: update employees.salaries set salary=?, to_date=? where emp_no=? and from_date=?

3
6
● Entities and sets (References)
● Great to have them (sets) but sometime they stab you in the back!
● What is Lazy and how to control it ... if you can!
● A few tests will follow:
○ Get (select) Employees Lazy/Not Lazy Hibernate versus simple code
○ Update a Salary value for a given employee

3
7
Entity and references the Lazy factor
Remember this?

Entity and References - the Lazy Factor
<set name="salarieses" table="salaries" lazy="true|false" fetch="select" >
<key>
<column name="emp_no" not-null="true" />
</key>
<one-to-many class="net.tc.employees.Salaries"/>
</set>
LAZY
List<Employees> employees = se.createQuery("from EmployeesSummary where emp_no >10000 and emp_no < 20020 " ).list();
logger.info("EmployeeSummary name = " + myEmp.getFirstName()+" "+ myEmp.getLastName());
Iterator it2 = myEmp.getTitleses().iterator(); ← HERE I retrieve the info
Iterator it3 = myEmp.getSalarieses().iterator(); ← HERE I retrieve the info
Iterator it4 = myEmp.getDeptEmps().iterator(); ← HERE I retrieve the info
}
NOT LAZY
List<Employees> employees = se.createQuery("from EmployeesSummary where emp_no >10000 and emp_no < 20020 " ).list(); ← HERE
logger.info("EmployeeSummary name = " + myEmp.getFirstName()+" "+ myEmp.getLastName());
Iterator it2 = myEmp.getTitleses().iterator();
Iterator it3 = myEmp.getSalarieses().iterator();
Iterator it4 = myEmp.getDeptEmps().iterator();
}

Lazy SQL
Hibernate: select ... from employees.employees employeess0_ where emp_no>10000 and emp_no<20020
19/05/15 16:34:57 INFO [PL20192]: EmployeeSummary name = Georgi Fucelli
Hibernate: select ... from employees.titles titleses0_ where titleses0_.emp_no=?
Hibernate: select ... from employees.salaries salarieses0_ where salarieses0_.emp_no=?
Hibernate: select ... from employees.dept_emp deptemps0_ where deptemps0_.emp_no=?
19/05/15 16:35:06 INFO [PL20192]: EmployeeSummary name = Bezalel Simmel
Not Lazy SQL
Hibernate: select ... from employees.employees employeess0_ where emp_no>10000 and emp_no<10003
Hibernate: select ... from employees.dept_manager deptmanage0_ where deptmanage0_.emp_no=?
Hibernate: select ... from employees.dept_manager deptmanage0_ where deptmanage0_.emp_no=?
19/05/15 16:40:54 INFO [PL20192]: EmployeeSummary name = Georgi Fucelli
19/05/15 16:41:01 INFO [PL20192]: EmployeeSummary name = Bezalel Simmel

4
0

4
1
Best way to handle it … and be safe.

4
2
Best way to handle it … and be safe. WAIT
WAIT
WAIT
WHY FKs???

DBA HR Complains
Your application has overwritten HR data
● Data out of order
● Missing consistency
● Unexpected results

4
4
Lock Me Please
TRX 1
logger.info("I am the 1st TRX ");
EmployeesSummary myEmp =
se.find(EmployeesSummary.class,10001,
LockModeType.PESSIMISTIC_WRITE);
logger.info("1TRX Employee name Before = "
+ myEmp.getFirstName()+" "
+ myEmp.getLastName()
+" " + mySal.getSalary());
mySal.setSalary(mySal.getSalary() + 1000);
logger.info("1TRX Employee name After = "
+ myEmp.getLastName() +" " + mySal.getSalary());
}
}
logger.info("Another Transaction is modifying the same value ");
se.disconnect();
se.close();
logger.info("1TRX COmplete");
TRX 2
logger.info("I am the 2nd TRX ");
EmployeesSummary myEmp = se.find(EmployeesSummary.class,10001);
logger.info("2TRX Employee name Before = "
+ myEmp.getLastName() +" "
+ mySal.getSalary());
mySal.setSalary(mySal.getSalary() - 400);
logger.info("2TRX Employee name After = "
+ myEmp.getLastName() +" "
+ mySal.getSalary());
}
}
se.disconnect();
se.close();
logger.info("2TRX COmplete");

4
6
Lock Me Please
I am the 1st TRX
Hibernate: select ... from employees.employees employees0_ where employees0_.emp_no=? for update
Hibernate: update employees.salaries set salary=?, to_date=? where emp_no=? and from_date=?
I am the 2nd TRX
Hibernate: select ... from employees.employees employees0_ where employees0_.emp_no=?
Hibernate: update employees.salaries set salary=?, to_date=? where emp_no=? and from_date=?

4
7
Lock Me Please - WHY?
Because the select for update is on the main object -> Employees, but I am actually updating a Salaries
record so ... no lock
Hibernate: select employees0_.emp_no as emp_no1_3_0_... from employees.employees employees0_
where employees0_.emp_no=? for update
If using lock also on the second TRX
19/05/14 12:16:57 ERROR [org.hibernate.engine.jdbc.spi.SqlExceptionHelper2]: Lock wait timeout
exceeded; try restarting transaction
Exception in thread "main" javax.persistence.PessimisticLockException: could not extract ResultSet
The right thing to do?
- If you want to operate by parent object be sure you lock them also in the other TRX
- DO NOT operate by parent and instead operate by sub (Salaries) and be sure to lock it

4
8
A Can of Worms
This is actually even more complicated, and a real...

5
0
What Can We Do Better?
A few thing to keep in mind:
● Select by pager (yes you can do it)
● Use reduced object when you don't need all columns (use different classes)
● Use Lazy
● Use LockModeType.PESSIMISTIC_WRITE only when needed, but use it
● Use high JDBC fetch size in Hibernate conf
● Leave sequence to DB
● Do not detach / attach entities lightly
● Release connection after transaction
● Use connection Pool with Hibernate
● Do not use Hibernate unless you really need to
● Apply JDBC optimizations

Accessing Data Through Hibernate; What DBAs Should Tell Developers and Vice Versa

Accessing Data Through Hibernate; What DBAs Should Tell Developers and Vice Versa

More Related Content

What's hot (20)

Similar to Accessing Data Through Hibernate; What DBAs Should Tell Developers and Vice Versa (20)

More from Marco Tusa (20)

Recently uploaded (20)

Accessing Data Through Hibernate; What DBAs Should Tell Developers and Vice Versa