Link to home
Create AccountLog in
Avatar of srinitin

asked on

Serialize/Deserialize using Streams


I am using following streams to send objects on NIO sockets:

For sending data( Few code snippet):==>

ByteArrayOutputStream   bStream = new ByteArrayOutputStream();
ObjectOutputStream oStream = new ObjectOutputStream( bStream );

// serializedata is a object to be sent on socketchannel
oStream.writeObject (serializedata);

byteArray = bStream.toByteArray();

// objOutput is ByteBuffer


numWrite = sc.write(objOutput);

// Some extra logic to stop sender


For receiving data ( Few code snippet):==>

numRead = );
byteArray  = objInput.array();

ByteArrayInputStream bStream1 = new ByteArrayInputStream(byteArray, offset,limit - offset);

ObjectInputStream iStream = new ObjectInputStream( bStream1 );

serializeData = (SerializedData) iStream.readObject();


My program is working fine the only problem is that it is taking more time than old IO.

I want to use global Stream objects (rather than creating a new object for each message as used in code snippet above).

When i am trying to use global objects .. getting StreamCorruptedException.

Can anyone help me and let me know the reason for this exception and also the way to use global stream objects to speedup.?
Any other way to improve performance will be of great help.

Thanks in advance.
Avatar of CEHJ
Flag of United Kingdom of Great Britain and Northern Ireland image

Link to home
Create an account to see this answer
Signing up is free. No credit card required.
Create Account
Avatar of srinitin


Thanks CEHJ for quick input.
Does it means that for each message i have to use new objects of streams as i am already using in my sample program or any other better way?

Then what are the probable problem in my program and why it is slower than my old IO program.
It is running perfectly fine ( as expected) and i am not getting any error.

Say for sending 50000 msgs:
using old IO time taken : 32 secs
and with NIO , time taken : 49 secs.

Any help will be greatly appreciated.
Does the used buffer size matters? as for sending data i have used BUFFER_SIZE=10000 and for reading data i have used BUFFER_SIZE=35000.
How are you sending the messages using old IO? Can you post that code?
Sure Ajay.  Sharing some logic here.

The old IO using Socket in place of socketChannel:

Some code snippet==>

For sending:

 protected ObjectOutputStream  objOutput;

  objOutput = new ObjectOutputStream(sc.getOutputStream());

For receiving:

 protected ObjectInputStream   objInput;
 objInput = new ObjectInputStream(sc.getInputStream());

msg = (MessageImpl) objInput.readObject();

One extra logic which i have used in my NIO program on receiver side is as follows:
To handle the scenario when multiple messages receives at once

// limit: number of bytes received in buffer
// offset : received message size
offset =0;
while( offset < limit)
bStream = new ByteArrayInputStream(byteArray,offset,limit - offset);
ObjectInputStream iStream = new ObjectInputStream( bStream );
serializeData = (SerializedData) iStream.readObject();
System.out.println("Msg: " + serializeData.getData());

offset= limit-bStream.available();

It is working fine.
I am working in synchronous mode and without using selectors. ( due to limitation of my existing system).  Only want to take advantage of bytebuffers and socketchannel.

Just want to know what should be expected behavior of NIO over old IO in such scenario i.e sending/receiving bytearray using socketchannel in synchronous mode.

1.) For small size of messages
2.) For large size messages.

Is there always increase in speed or it may degrade the performance also?
My intent is just to sppedup the things. is it good to use NIO?
any inputs?
Certainly you can't expect NIO to deliver improved performance in all cases. Sometimes an enhancement will only appear in heavy operations.

To enhance your current code, try enlarging the send buffer on the Socket. Call allocateDirect to allocate the buffer and make sure the buffer size aligns to word boundaries

final int BUF_SIZE = 1 << 16; // 64K
Link to home
Create an account to see this answer
Signing up is free. No credit card required.
Create Account