Parquet Java Serialization is very slow

Hi,
I am doing a POC to compare different data formats and its performance in terms of serialization/deserialization speed, storage size, compatibility between different language etc. 
When I try to serialize a simple java object to parquet file,  it takes _\*6-7 seconds\*_ vs same object's serialization to JSON is **_100 milliseconds._**

Could you help me to resolve this issue?

+**My Configuration and code snippet:**
**Gradle dependencies**
dependencies

{ compile group: 'org.springframework.boot', name: 'spring-boot-starter' compile group: 'org.projectlombok', name: 'lombok', version: '1.16.6' compile group: 'com.amazonaws', name: 'aws-java-sdk-bundle', version: '1.11.271' compile group: 'org.apache.parquet', name: 'parquet-avro', version: '1.10.0' compile group: 'org.apache.hadoop', name: 'hadoop-common', version: '3.1.1' compile group: 'org.apache.hadoop', name: 'hadoop-aws', version: '3.1.1' compile group: 'org.apache.hadoop', name: 'hadoop-client', version: '3.1.1' compile group: 'joda-time', name: 'joda-time' compile group: 'com.fasterxml.jackson.core', name: 'jackson-databind', version: '2.6.5' compile group: 'com.fasterxml.jackson.datatype', name: 'jackson-datatype-joda', version: '2.6.5' }

**Code snippet:**+

public void serialize(List inputDataToSerialize, CompressionCodecName compressionCodecName) throws IOException {

Path path = new Path("s3a://parquetpoc/data_"+compressionCodecName+".parquet");
Path path1 = new Path("/Downloads/data_"+compressionCodecName+".parquet");
Class clazz = inputDataToSerialize.get(0).getClass();

try (ParquetWriter writer = **AvroParquetWriter.**builder(path1)
.withSchema(ReflectData.AllowNull.get().getSchema(clazz)) // generate nullable fields
.withDataModel(ReflectData.get())
.withConf(parquetConfiguration)
.withCompressionCodec(compressionCodecName)
.withWriteMode(OVERWRITE)
.withWriterVersion(ParquetProperties.WriterVersion.PARQUET_2_0)
.build()) {

for (D input : inputDataToSerialize)

{ writer.write(input); }

}
}

+**Model Used:**
@Data
public class Employee

{ //private UUID id; private String name; private int age; private Address address; }

@Data
public class Address

{ private String streetName; private String city; private Zip zip; }

@Data
public class Zip

{ private int zip; private int ext; }

 

private List **getInputDataToSerialize**(){
Address address = new Address();
address.setStreetName("Murry Ridge Dr");
address.setCity("Murrysville");
Zip zip = new Zip();
zip.setZip(15668);
zip.setExt(1234);

address.setZip(zip);

List employees = new ArrayList<>();

IntStream.range(0, 100000).forEach(i->{
Employee employee = new Employee();
// employee.setId(UUID.randomUUID());
employee.setAge(20);
employee.setName("Test"+i);
employee.setAddress(address);
employees.add(employee);
});
return employees;
}

**Note:**
**I have tried to save the data into local file system as well as AWS S3, but both are having same result - very slow.**

**Reporter**: [Felix Kizhakkel Jose](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=FelixKJose) / @FelixKJose

_{**Note**: *This issue was originally created as [PARQUET-1680](https://issues.apache.org/jira/browse/PARQUET-1680). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

Commencez par reproduire le benchmark autour de AvroParquetWriter.builder, ReflectData.AllowNull.get().getSchema(clazz) et writer.write en utilisant le modèle Employee fourni et l’entrée de 100,000 enregistrements. Comparez la génération du schéma, la construction du writer et l’écriture sur le chemin local et le chemin S3 ; le travail est considéré comme terminé lorsque la source dominante de latence est identifiée et qu’une amélioration mesurée est démontrée.

Rédigé par le modèle d'indexation à partir du texte de l'issue.