Sockets and Networking

1
Sockets and Networking
2
Buzzword Bingo

• The only thing that's cooler than saying that
  your program is multi-threaded is, of course,
  saying that it's "client-server". Bonus points
  if you can throw in something like "thin-client"
  or "n-tiered".
• What do these terms actually mean? You should
  probably know, on the off chance that you get the
  one in a million manager who knows what they're
  talking about -).

3
Client Server Architecture

• A client-server architecture is one where the
  application, or more correctly, the system,
  consists of two or more programs.
• Can you guess what those programs are?

4
The Client

• The client is the user front end to the system.
  It is what will run on the users desktop, or in
  the browser.

5
Thick VS Thin

• The client can be either a "thick" client, which
  means that most of the processing logic and
  computation is done by the client program,and
  summaries are passed back to the server.
• It can also be a "thin"client, which means that
  it is mostly an empty UI that passes back all
  input to the server for processing.

6
How To Choose

• In previous times, the thick client was the norm,
  because it distributed the load of processing
  across many desktops, rather than asking one
  server to do all the work.
• As servers became more powerful, the thin client
  came into ascendancy, because it was far easier
  to do updates if the processing logic changes,
  but the client-server API stays the same, there
  is no need to update umpteen desktops to update
  the logic.

7
The Server

• The server is what runs on the host system.
• There is normally only one server in a
  client-server app, although there may be load
  balanced servers, or fail over servers, that
  afford high availability.

8
Server Details

• The server is usually responsible for managing
  multiple clients, that connect to the server over
  a network.
• The server handles their requests, processes
  whatever needs to be processed, and sends back
  responses to the client.
• The server is normally in the same language as
  the clients, but this is not a requirement. You
  can have a server written in C, and clients
  written in Java that connect to it, as long as
  they are "talking" in the same language. After
  all, strings and ints are strings and ints.

9
So, Whats With The Tiers?

• A normal client server application could be
  considered to be "2-tiered", where the client is
  one tier, and the server a second.
• client ----------------------- server
• tier 1 tier 2

10
N-Tiered

• An N-tiered application is one where there are
  more steps in this pathway.
• N-tiered applications usually have some sort of
  collector, or filter, or translator acting as a
  proxy between the client and the server, or they
  have separated the processing on the server side
  from the data/mail servers, or some other
  configuration of a similar nature.

11
Communication

• So, how does the client talk to the server?
• The connection is typically made over a network,
  using some sort of network protocol to
  communicate between the two programs.

12
Examples

• The most ubiquitous example of client server
  programming today is the internet.
• The browser is a client, and it connects to a web
  server, and makes requests using HTTP.
• FTP clients, SSH clients, even your terminals,
  are client server applications.

13
How Do They Connect?

• Each computer has a unique address called an IP
  (Internet Protocol) address.
• This identifies a computer on a network.
• The address is a set of four numbers, from 0 -
  255, that will uniquely define a machine.
• For example the IP of armstrong is 128.205.32.3,
  and yeager is 128.205.36.9.
• This number is sort of like a telephone number,
  with each successive sequence more closely
  identifying the location of the computer.

14
DNS Names

• Most machines also have a DNS (Domain Name
  System) name, which is the user-friendly (or
  human friendly) way to get to computers.
• For example, you can get to armstrong by typing
  ssh armstrong.cse.buffalo.edu, as well as by
  typing ssh 128.205.32.3.
• The DNS name is simply a mapping from a human
  readable name to its unique IP address.

15
Port Numbers

• Each host machine also has a number of ports.
• Ports are sort of like frequencies on the radio,
  or extensions to a phone system each one
  provides a sub-address that you can talk to the
  host computer on.
• Some ports have standard designations for
  example, all internet web traffic comes to port
  80, and all e-mail goes out from port 25.
• There are standard ports for SSH, and Telnet, and
  most of the other standard services.

16
UNIX Ports

• On UNIX systems, all ports from 0-1023 are
  reserved for root level processes.
• However, higher ports are open for client server
  applications to access.

17
The Communications Channel

• In order for your client to talk to your server,
  it has to know what host the server is running
  on.
• It also needs to know what port that the server
  is listening to for requests from that type of
  client.
• The server needs to know what port to listen to.

18
The Server Starts The Process

• Once your server is up and running, it will open
  a "socket" to the port that you want it to listen
  to.
• It will wait until a client attaches to that
  socket, and forms a connection.
• You can then get an input and an output stream
  for the connection from that socket.
• These are the streams that will form the
  communication pathways between the client and the
  server.

19
WARNING!

• Any time you create a client server application,
  you create a potential security problem.
• Due to the nature of network communication, you
  have no idea where the input that someone sends
  to a server is coming from.
• It may be from your client, or it may be from
  some malicious hacker. The server is running as
  YOU, so anything that the server does in response
  to a request from a client (legit or otherwise)
  will happen to your account. This can be bad.

20
Network Security

• In Java, we can protect ourselves to some extent
  by passing arguments by java object streams.
• Random port assignment is another way to avoid
  some security holes.
• Beware anything especially that writes to the
  file system, does recursion based on input, etc.

21
Network Security

• It is a bad idea to have a server that accepts
  plain text commands like "delete filename" and
  then deletes things from your home directory,
  unless you also create some sort of security
  layer.
• Also - don't leave your server running unless
  necessary. This is asking for trouble.

22
Code Example

• See SimpleServer.java and SimpleClient.java

23
That Was Cool, But

• When the client disconnected, the server also
  died.
• What's up with that?
• The server was only set up to receive a single
  connection.
• See StayaliveServer.java

24
Better

• Ok, now we can connect and disconnect without
  killing the server.
• But what about communication back to the client?
• See TwoWayServer.java and TwoWayClient.java

25
But What About

• Cool! Now we have two way communication.
• But what about multiple clients?
• Well, what do we want to do with multiple
  clients?
• Do we want to process the client input
  simultaneously?
• What do we do when we want to do two (or more)
  things simultaneously in a program?
• Hint - no needles required for this one.
• See MultiServer.java

26
Not To Get Picky

• But what about asynchronous communication?
• With a tight loop between the send to the server
  and the response, the server can't send
  unsolicited messages to the clients.
• The main while loop in the client controls server
  send and receive.
• How do we set it up so that the client can
  receive messages back from the server?
• Hint This application is starting to look more
  like a tapestry...
• See MultiSocketClient.java and MultiSocketServer.j
  ava

27
Things To Think About

• How would you handle server shutdown?
• How about decoupling the servers tight loop for
  accepting clients?