13 Application Layer

Applications are the entities that communicate with each other to exchange services

To make any use of the Internet, application programs should run on the two endpoints of a network connection

A socket is one end of an inter-process communication channel.

Client Server Model¶

Many-to-One

Server	Client
Run all the time (i.e. infinite) Provide service to any client Usually specialize in providing certain type of service, e.g. Mail Listen to a well-known port and passively open connection.	Run when needed, then terminate (i.e. finite) Actively open TCP or UDP connection with Server’s socket

Client needs to know existence & address of server
However, the server does not need to know the existence or address of the client prior to the connection
Once a connection is established, both sides can send and receive information

Both client and server will construct a socket
The process to establish a socket on the client side is different from the process to establish a socket on the server side

Every node in the network acts alike

No central point of failure
Scalability
Since every peer is alike, it is possible to add more peers to the system and scale to larger networks

HyperText Transfer Protocol

Used for Client-Server model

Uses TCP
“Stateless”
A ‘state’ is information kept in memory of a host, server or router to reflect past events: such as routing tables, data structures or database entries
HTTP server maintains no information about past client requests
Protocols that maintain “state” are complex!
- history (state) is maintained
- if server/client crashes, views of “state” may be inconsistent, must be reconciled
- state is added via ‘cookies’

	Non-Persistent HTTP	Persistent HTTP without Pipelining	Persistent HTTP with Pipelining
Max no of objects sent over TCP connection	1	Multiple	Multiple
Used in HTTP Version	HTTP/1.0		HTTP/1.1
Categories
Steps		Server leaves connection open after sending response Subsequent HTTP messages between same client/server sent over open connection Client issues new request only when previous response has been received One RTT for each referenced object	Server leaves connection open after sending response Subsequent HTTP messages between same client/server sent over open connection Client sends requests as soon as it encounters a referenced object 1 RTT for all referenced objects
Total Response Time (\(n =\) no of objects)	\(n (2 \ \text{RTT} + \text{Transmit Time})\) - one RTT to initiate TCP connection - one RTT for HTTP request & first few bytes of HTTP response to return - file transmission time	\(n (1 \ \text{RTT} + \text{Transmit Time})\)	\(1 \ \text{RTT} + n(\text{Transmit Time})\)
Disadvantages	- requires 2 RTTs per object - OS overhead for each TCP connection - browsers often open parallel TCP connections to fetch referenced object