Advanced Java Programming

Unit One: Networking Gareth Lee

Department of Electrical and Electronic Engineering, University of Western Australia

Overview
Javas network support Addressing other machines Communicating using TCP/IP Communicating using UDP Broadcasting and multicasting

Recommended Reading
Java Network Programming, Elliotte Rusty Harold, OReilly and Associates, 1997, ISBN 1-56592-227-1 TCP/IP Network Administration, Second Edition, Craig Hunt, OReilly and Associates, 1997, ISBN 1-56592-322-7 The Java Developers connection: http://www.javasoft.com/jdc The Javadoc documentation

Network Programming
Mechanisms by which software running on two or more computational devices can exchange messages
Desktop Computers PDAs / Set Top Boxes / Mobile Telephones?

Java is a network centric programming language Java abstracts details of network implementation behind a standard API
Portable (and future proof) . . . but may be rather limiting

A programming model for network communications

Applications
Application Protocols

Java APIs HTTP/FTP/etc... 1:1 or 1:Many TCP, UDP Internet Protocol

Java & IP dont care

Unicast Broadcast Multicast

Streams Datagrams Addressing & Routing Physical Transport

Dealt with in this Unit One. Dealt with in this Unit Two.

Internet Protocol (IPv4)

Abstracts away the details of the physical network implementations (such as Ethernet, Token Ring, ATM, Sonet) All traffic uses the same rules to move from machine to machine
Easy to write programs Easy to build network hardware

Works with Datagrams: small discrete packets of data (rather like a letter)

Internet Protocol (IPv4)

A way of uniquely addressing machines using 32 bit addresses: giving 4 billion possible addresses (like a zip code) A system for numbering ports on each machine (like a post office box) Port numbers allow several services to operate from a machine at the same time

Common well known ports

Ports 20/21 File Transfer Protocol Port 23 Telnet Port 25 Simple Mail Transport Proto. Port 79 Finger Port 80 HTTP Port 110 POP3 (Post Office Protocol) All well known ports in the range 1..1023

Internet Protocol (IPv4)

The Internet consists of a large number of independent sub-networks (subnets) A mechanism for relaying datagrams from one network to another (routing) For routing to work each organisation must have a well known prefix (all UWA addresses start with the bytes 130.95) . . but we dont use addresses very efficiently and were running out fast (UWA doesnt need 65536 separate addresses)

Next Generation Internet

The solution is IPv6 which uses 128 bit addresses Improves services such as multicasting and secure communication Several addresses per m2 of the Earths surface (3.4 x 1038 to be precise) Not yet widely deployed by ISPs Perhaps widely deployed in 2-4 years Well written Java software should move to IPv6 without modification/recompilation
One benefit of abstracted APIs

IP Addresses and Java

Java has a class java.net.InetAddress which abstracts network addresses Serves three main purposes:
Encapsulates an address Performs name lookup (converting a host name into an IP address) Performs reverse lookup (converting the address into a host name)

java.net.InetAddress (1)
Abstraction of a network address Currently uses IPv4 (a 32 bit address) Will support other address formats in future Allows an address to be obtained from a host name and vice versa Is immutable (is a read-only object)
Create an InetAddress object with the address you need and throw it away when you have finished

java.net.InetAddress (2)
Static construction using a factory method
InetAddress getByName(String hostName) hostName can be host.domain.com.au, or hostName can be 130.95.72.134 InetAddress getLocalHost()

Some useful methods:

String getHostName() Gives you the host name (for example www.sun.com) String getHostAddress() Gives you the address (for example 192.18.97.241) InetAddress getLocalHost() InetAddress[] getAllByName(String hostName)

Using InetAddress objects

import java.net.InetAddress; import java.net.UnknownHostExcepion;

public static void main(String[] args) { try { InetAddress inet1 = InetAddress.getByName("asp.ee.uwa.edu.au"); System.out.println( "HostAddress=" + inet1.getHostAddress()); InetAddress inet2 = InetAddress.getByName("130.95.72.134"); System.out.println("HostName=" + inet2.getHostName()); if (inet1.equals(inet2)) System.out.println("Addresses are equal"); } catch (UnknownHostException uhe) { uhe.printStackTrace(); } }

Transmission Control Protocol

TCP is built on top of IP Provides the illusion of a continuous flow (or stream) of data between sender and receiver (rather like a telephone call) Splits up streams into strings of small datagrams which are sent in succession Contains an error recovery mechanism to recover datagrams which are lost These features make application development simpler and so it is widely used

Transmission Control Protocol

Used by FTP / Telnet / Finger and numerous other network applications Used by stream oriented servers such as HTTP (as we will see in unit 2) Can also be used to provide inter-process communications (IPC) between the applications on a single machine (such as a X-windows clients and servers)

Two types of TCP Socket

java.net.ServerSocket is used by servers

so that they can accept incoming TCP/IP connections

A server is a piece of software which advertises and then provides some service on request

java.net.Socket is used by clients who

wish to establish a connection to a (remote) server

A client is a piece of software (usually on a different machine) which makes use of some service

java.net.ServerSocket (1)
Listens on well-known port for incoming connections Creates a dynamically allocated port for each newly established connection Provides a Socket connected to the new port Maintains a queue to ensure that prospective clients are not lost

java.net.ServerSocket (2)
Construction:
ServerSocket(int port, int backlog) Allows up to backlog requests to queue waiting for the server to deal with them

Some useful methods:

Socket accept() Blocks waiting for a client to attempt to establish a connection void close() Called by the server when it is shutting down to ensure that any resources are deallocated

More details in the Javadoc (as always!)

java.net.Socket (1)
Provides access to TCP/IP streams Bi-directional communication between sender and receiver Can be used to connect to a remote address and port by using the constructor:
Socket(String remoteHost, int port)

Also used to accept an incoming connection (see ServerSocket)

java.net.Socket (2)
Can obtain access to input and output streams Input stream allows reception of data from the other party
InputSteam getInputStream()

Output stream allows dispatch of data to the other party

OutputStream getOutputStream()

How it all fits together

Client (sid)
s = new Socket (fred, 80)

Server (fred)
2037 80
ServerSocket ss. s = ss.accept()

2037

1583

Socket s

s.getInputStream() 2037 s.getOuputStream()

1583 s.getInputStream()

s.getOuputStream()

A sample TCP server

public static void main(String[] args) { try { ServerSocket agreedPort = new ServerSocket(AGREED_PORT_NUMBER, 5); while (isStillServing()) { Socket session = agreedPort.accept(); respond(session); session.close(); } agreedPort.close(); } catch // } catch // } }

(UnknownHostException uhe) { Very unlikely to occur

(IOException ioe) { May occur if the client misbehaves?

A sample TCP client

public static void main(String[] args) { try { InetAddress server = InetAddress.getByName(args[0]); Socket connection = new Socket(server, AGREED_PORT_NUMBER); makeRequestToServer(connection); getReplyFromServer(connection); connection.close(); } catch (UnknownHostException uhe) { // arg[0] is not a valid server name or IP address } catch (IOException ioe) { // The connection to the server failed somehow: // the server might have crashed mid sentence? } }

What are datagrams?

Datagrams are discrete packets of data Each is like a parcel that can be addressed and sent to an recipient anywhere on the Internet This is abstracted as the User Datagram Protocol (UDP) in RFC768 (August 1980) Most networks cannot guarantee reliable delivery of datagrams

Why use datagrams?

Good for sending data that can naturally be divided into small chunks Poor for (lossless) stream based communications Makes economical use of network bandwidth (up to 3 times the efficiency of TCP/IP for small messages) Datagrams can be locally broadcast or multicast (one-to-many communication)

Application using datagrams

UDP can be used for economical point-topoint communications over LANs
Unix NFS (Network File System) NIS (a.k.a. Yellow Pages)

Datagrams can be used for one-to-many communication:

Local network broadcasting; Multicasting (MBONE)

but there is no way to create one-to-many streams using TCP/IP

java.net.DatagramPacket (1)
DatagramPackets normally used as short lived envelopes for datagram messages:
Used to assemble messages before they are dispatched onto the network, or dismantle messages after they have been received

Has the following attributes:

Destination/source address Destination/source port number Data bytes constituting the message Length of message data bytes

java.net.DatagramPacket (2)
Construction:
DatagramPacket(byte[] data, int length)

Some useful methods:

void setAddress(InetAddress addr) InetAddress getAddress() void setPort(int port) int getPort()

DatagramPackets are not immutable so, in principle you can reuse then, but . . Experience has shown that they often misbehave when you do -- create a new one, use it once, throw it away!

java.net.DatagramSocket (1)
Used to represent a socket associated with a specific port on the local host Used to send or receive datagrams Note: there is no counterpart to java.net.ServerSocket! Just use a DatagramSocket with a agreed port number so others know which address and port to send their datagrams to

java.net.DatagramSocket (2)
Construction:
DatagramSocket(int port) Uses a specified port (used for receiving datagrams) DatagramSocket() Allocate any available port number (for sending)

Some useful methods:

void send(DatagramPacket fullPacket) Sends the full datagram out onto the network void receive(DatagramPacket emptyPacket) Waits until a datagram and fills in emptyPacket with the message

. . . and a few more in the Javadoc

sea.datagram.DatagramSender
This example sends datagrams to a specific host (anywhere on the Internet) The steps are as follows:
Create a new DatagramPacket Put some data which constitutes your message in the new DatagramPacket Set a destination address and port so that the network knows where to deliver the datagram Create a socket with a dynamically allocated port number (if you are just sending from it) Send the packet through the socket onto the network

sea.datagram.DatagramSender
byte[] data = This is the message.getBytes(); DatagramPacket packet = new DatagramPacket(data, data.length); // Create an address InetAddress destAddress = InetAddress.getByName(fred.domain.com); packet.setAddress(destAddress); packet.setPort(9876); DatagramSocket socket = new DatagramSocket(); socket.send(packet);

sea.datagram.DatagramReceiver
The steps are the reserve of sending:
Create an empty DatagramPacket (and allocate a buffer for the incoming data) Create a DatagramSocket on an agreed socket number to provide access to arrivals Use the socket to receive the datagram (the thread will block until a new datagram arrrives) Extract the data bytes which make up the message

sea.datagram.DatagramReceiver
// Create an empty packet with some buffer space byte[] data = new byte[1500]; DatagramPacket packet = new DatagramPacket(data, data.length); DatagramSocket socket = new DatagramSocket(9876);

// This call will block until a datagram arrives socket.receive(packet);

// Convert the bytes back into a String and print String message = new String(packet.getData(), 0, packet.getLength()); System.out.println("message is " + message); System.out.println("from " + packet.getAddress());

But its never quite that simple!

Several of the constructors/methods throw exceptions which I have omitted Each datagrams can only hold up to a maximum of 64KB of data . . . . but the underlying transport layer may split the message into smaller packets (for instance Ethernet uses about 1500 bytes) Always remember that UDP is an unreliable protocol: If any of the split datagrams are lost the whole message will be lost

Broadcasting
Broadcasting allows a single datagram to be sent to a group of listeners The group consists of all the computers within the local network The previous code examples can be used for broadcasting Just change the address: each network has a unique broadcast address

IP addresses revisited
Each 32 bit IP number consists of two components:
The network address The unique international address of the network The host address The unique address of a specific host in the net

There are three classes of network address denoted class A, B and C

Class A,B and C addresses

192 Class A 0... . 85 . 35 . 87

Class B
Class C

10...
110... Network Address Byte Host Address Byte

Broadcast addresses
CIIPS has a class C network which has the address 130.95.72 This portable computer has host address 134 within the CIIPS network Each network has a single host address which is set aside for broadcasts (either all one bits or all zero bits) The CIIPS network uses broadcast address 130.95.72.255 Broadcasts are never routed onto other networks

Multicasting (1)
Described in RFC1112 (August 1989) Multicasting allows distribution of a datagram to a group of listeners who are not within the local network Routers between networks need to pass multicast datagrams. . but many do not! The MBONE is a way of tunneling datagrams across the Internet between islands of multicast activity

Multicasting (2)
Multicasts are also sent to a special address (known as a group) Multicast groups need to be agreed in advance. They are not derived from a specific network/host address
Multicast groups identify a subject area (or stream of content) rather than a specific computer or network. They are more like a TV channel number than a telephone number.

The IETF has set aside addresses from 224.0.0.1 to 239.255.255.255 specifically for multicasting

Multicasting (3)
To send to (or receive from) a multicast group it is first necessary to register interest in the group
This results in an Internet Group Management Protocol (IGMP) message being sent to your router (RFCs 988/1112/2236)

Then a datagram is created, addressed to the group (and the chosen port) Java has a specialised socket for multicasting: java.net.MulticastSocket

Some multicast groups

224.0.0.1 All hosts within local subnet 224.0.1.7 Audio news multicast 224.0.1.12 Video from IETF meetings 224.0.1.20 Expts. within local subnet 224.0.1.25 NBC Professional News There are 268 million multicast addresses (in IPv4) with 65 thousand ports in each!

java.net.MulticastSocket
Subclass of java.net.DatagramSocket Constructed the same way Adds some extra methods:
void joinGroup(InetAddress mcastGroup) Enter the specifies group so that you can send or receive datagrams void leaveGroup(InetAddress mcastGroup) Leave a group that you previously joined void setTimeToLive(int ttl) Sets how far your datagrams will travel before routers ignore them int getTimeToLive()

sea.datagram.MulticastSender
Sending similar to the previous example. . . . .but must register with the multicast group and decide the longevity The steps involved are:
Create the MulticastSocket. Join the multicast group(s) (on startup). Create the DatagramPacket. Send the packet through the socket. Leave the multicast group (on exit).

sea.datagram.MulticastSender
InetAddress multicastGroup = InetAddress.getByName(multicastGroupAddr); MulticastSocket socket = new MulticastSocket(); socket.joinGroup(multicastGroup); socket.setTimeToLive(5); byte[] data = This is the message.getBytes(); DatagramPacket datagram = new DatagramPacket(data, data.length); datagram.setAddress(multicastGroup); datagram.setPort(9876); socket.send(datagram); socket.leaveGroup(multicastGroup);

sea.datagram.MulticastReceiver
The steps are:
Create a multicast socket on an agreed port. Join the multicast group (on startup). Create an empty datagram. Wait for datagram to be delivered on the socket. Unpack and use the datagram. Leave the multicast group (on exit).

sea.datagram.MulticastReceiver
InetAddress multicastGroup = InetAddress.getByName(multicastGroupAddr); MulticastSocket socket = new MulticastSocket(9876); socket.joinGroup(multicastGroup); byte[] data = new byte[1000]; DatagramPacket packet = new DatagramPacket(data, data.length); socket.receive(packet); String message = new String( packet.getData(), 0, packet.getLength()); socket.leaveGroup(multicastGroup);

Useful sources of information

The Internet Engineering Task Force (IETF) which is at http://www.ietf.org -- you will be able to download RFCs from here Multicasting try reading the HOWTO which is available from the URL: http://ftp.southcom.com.au/LDP/HOWTO/... Multicast-HOWTO.html

Homework
Read through the code samples to convince yourself you understand whats going on Sample code can be downloaded from http://ciips.ee.uwa.edu.au/~gareth If you can, run the examples

Comments, Suggestions. . .
How was the presentation paced? Was there enough (or too much) technical content? Any areas of particular interest? Comments regarding presentation style?