[Network] Application Layer

1 Principles of Network applications

1.1 Application Architectures

Client-server

Peer-to-peer

Hybird of C/s and peer

--skype

Client-server:find the address of the remote client

Client-client:direct communication

--instant messaging

similar with Skype

1.2 Application Requirements (from Transport Layer Services)

1. About Data Loss

2. Timing

3. Throughput

4. Security

2 Application-layer Protocols

2.1 Web and HTTP

1. Properties of HTTP

Uses TCP

Nonpersistent Http:one TCP Connection for one object transmission

RTT: Time for a small packet from client to server and back.

ResponseTime: RTT (TCP connection) + RTT (HTTP request) + file transmission

Persistent Http:multiple Objects on one TCP connection

HTTP is stateless:

Server maintains no information about past client requests.

2. HTTP Request messages:

Request Line:method Path Protocol

Header lines

Entity body

3. HTTP Response messages:

Status Line:protocol statuscode statusphrase

Header lines

Data

4. Cookies

As we have disscussed above, HTTP was stateless, so if we want to keep States of users, we must utilize the process of HTTP To get it.

Most of the Web Sites use cookies.

Four components:

1) Cookie header line in HTTP request message

2) Cookie header line in HTTP response message

3) cookie file kept on user's host, managed by user browser

4) Cookie file kept on server database of WEB sites

To sum up, HTTP is stateless for it does not store the information of user states. But we can store the cookie in user hosts and servers and let the HTTP messages carry states.

5. Web Cache

Why caching

1) Reduce response time

2) Reduce traffic

With HTTP:

Use conditional get:if-modified-since, not-modified

2.2 FTP

The

FTP seperate control and data connections, see following:

Out-of-band: Is the control and separation, when there is an emergency (control data need to transfer, control transmission With a higher priority)

1) Client build control connection, server build data connection

2) FTP is stateful

2.3 Email

1. SMTP: Reliably transfer email messages (TCP connection)

 

Compare with HTTP

Http:pull (GET); No such restriction; Each object encapsulated with its own response msg

Smtp:push (transfer); 7-bit ASCII; Multiple objects sent in multiple msg

Message format:7-bit ASCII

Header lines (if extention, add some lines to declare the file type)

Body

2. POP3

Store email locally.

3. IMAP

Store on mail server.

4. The difference of them

This is the page from the

quoted.

SMTP is the basis. POP3 (locally, stateless) and IMAP (on the server, Stateful-folder) is used to access the mails.

2.4 DNS

tld:top-level-domain com, cn, org ...

1. DNS Records

hostname is the name of a HOST-IBM

Domain is part of the

URL you can consider as.-IBM.com

alias name is url-www.ibm.com

canonical name is real name-servereast.backup2.ibm.com

2. DNS Protocol, Message

Query and Reply messages

Header:identification and Flags

some

fields

3. An example:inserting records into DNS

(Networkutopia.com,dns1.networkutopia.com, NS)

(dns1.networkutopia.com,212.212.212.1, A)

Then

others can find your Web site at IP 212.212.212.1

2.5 Peer

There is some properties of peer:

1) No always-on server

2) End systems directly communicate

3) Peers change IP

1. File Distribution

c/S is "n copies upload and n copies download", but peer is "one copy uploads and n copies download".

BitTorrent

Tracker:track peers participating in torrent. -An individual

Torrent:group of peers exchanging chunks of a file.-A group

The

processes of bit torrent consist of pulling chunks and sending chunks.

-When downloading chunks, the peer would request a list of chunks from its neighbors, and choose the rarest chunks F Irst. (start with the most scarce content)

-When uploading chunks, the peer would send chunks to four neighbors currently sending it chunks at highest RA Te. (reciprocate)

So one file is held by several peers in the network and each

peer can both download and upload. And there is some prescriptions for peers, and up.

2. Searching for information

Query Blood

Is

a method to find resource in a peer network. Because the graph of the peer network may is connected by each node. So a query must is broadcasted in the network to find a resource.

It may cause DoS attack.

The

TCP connection built in the process of blooding form a hierarchical searching graph.

3. Skype

almost similar with BitTorrent and Query blood.

End.

[Network] Application Layer