EMERGING TRENDS IN DISTRIBUTED SYSTEM

PREPARED BY: G.S.MISHRA

UNIT-2

PEER-2-PEER SYSTEMS

WHAT IS PEER TO PEER?

Analogous to a telephone conversation

Two people of equal status communicate

A point to point connection

Definition:

P2P is a class of applications that takes advantage of resources e.g. storage, cycles, content, human presence, available at the edges of the Internet.

P2P CONTD…

e.g. in Gnutella, there are two key differences compared to client/server based systems:

• A peer can act as both a client and a server

• The network is completely decentralized and has

no central point of control.

Peers in a Gnutella network are typically

connected to three or four other nodes and to

search the network a query is broadcast

throughout the network.

HISTORY OF P2P

The Internet started as a peer-to-peer system.

The goal of the original ARPANET was to share

computing resources around the USA.

Its challenge was to connect a set of distributed

resources, using different network connectivity,

within one common network architecture.

The first hosts on the ARPANET were several US

universities, e.g., the University College of Los

Angeles, Santa Barbara, SRI and University of Utah.

CONTD….

These were already independent computing sites

with equal status and the ARPANET connected

them as such, not in a master/slave or client/server

relationship but rather as equal computing peers.

From the late 1960s until 1994, the Internet had

one model of connectivity.

Machines were assumed to be always switched on,

always connected, and assigned permanent IP

addresses.

CONTD…

With the invention of Mosaic, another model began

to emerge in the form of users connecting to the

Internet from dial-up modems.

This created a second class of connectivity

because PCs would enter and leave the network

frequently and unpredictably.

Because ISPs began to run out of IP addresses,

they began to assign IP addresses dynamically for

each session, giving each PC a different, possibly

masked, IP address.

CONTD…

This transient nature and instability prevented PCs

from being assigned permanent DNS entries, and

therefore prevented most PC users from hosting

any data or network-facing applications locally.

For a few years, treating PCs as clients worked

well. Over time though, as hardware and software

improved, the unused resources that existed

behind this veil of second-class connectivity

started to look like something worth getting at.

CONTD…

Given the vast array of available processors

mentioned earlier, the software community is

starting to take P2P applications very seriously.

Most importantly, P2P research is concerned in

addressing some of the main difficulties of current

distributed computing: scalability, reliability,

interoperability.

BINDING OF PEERS

Within today’s Internet, we rely on fixed IP addresses.

When a user types an address into his/her Web

browser (such as http://www.google.com/), the Web

server address is translated into the IP address (e.g.,

168.127.47.8) by a domain name server (DNS).

The Internet protocol (IP) then makes a routing

decision based on the IP Address.

If DNS is unavailable then typing http://168.127.47.8/

into a browser would be equivalent since the Web

page is permanently bound to the IP address.

THE PROCESS WHEREBY AN INTERNET ADDRESS IS CONVERTED INTO THE IP ADDRESS FOR LOCATING A WEB PAGE ON THE INTERNET

CONTD…

The above example shows Early Binding. Early bindings form a simple architecture very similar

to an address book on a mobile phone e.g., the person’s name is statically bound to his/her

telephone number.

If a Web site changed its IP address several times a

day then static binding will become impractical.

Often devices do not have a fixed address as they are

hidden behind Network Address Translation (NAT)

systems

therefore need a late binding of their addresses with

their network identifier.

A P2P ENVIRONMENT: DEVICES ARE CONNECTED BEHIND NATS AND FIREWALLS

MODERN DEFINITION: P2P

P2P is a class of applications that takes advantage of

resources e.g. storage, cycles, content, human

presence, available at the edges of the Internet

A peer can act as both a client and a server (they call

these servents i.e. server and client in Gnutella.)

The network is completely decentralized and has no

central point of control.

Peers in a Gnutella network are typically connected to

three or four other nodes and to search the network a

query is broadcast throughout the network.

SOCIAL IMPACTS OF P2P

Vaidhyanathan:

what we call P2P communicative networks actually

reflect and amplify - revise and extend - an old ideology

or cultural habit.

Electronic peer-to-peer systems like Gnutella merely

simulates other, more familiar forms of unmediated,

unsensorable, irresponsible, troublesome speech;

for example, anti-royal gossip before the French

revolution, trading cassette tapes among youth

subcultures as punk or rap, or the illicit Islamist cassette

tapes through the streets and bazaars of Cairo.

RAINSFORD : “INFORMATION FEUDALISM”

The current push for control over intellectual

property rights has bred a situation analogous to

the feudal agricultural system in the medieval

period.

In effect, songwriters and scientists work for

corporate feudal lords, licensing their own

inventions in exchange for a living and the right to

‘till the lands’ of the information society.

TRUE P2P Within P2P, there are three categories of systems Centralized systems: where every peer connects to

a server which coordinates and manages communication. Some examples here include the CPU sharing applications, e.g., SETI@Home

Brokered systems: where peers connect to a server in order to discover other peers, but then manage the communication themselves (e.g., Napster).

Decentralized systems: where peers run independently without the need for centralized services. Here, the discovery is decentralized and the communication takes place between the peers. Peers do not need a known centralized service for them to operate, e.g., Gnutella, Freenet

THE P2P ENVIRONMENT

Peers are: extremely transient (they are

continually disappearing and reappearing)

connections are often multi-hop (i.e., packets

travel via several intermediaries before they

reach their destination)

peers reside in hostile environments (i.e., they

live behind NAT routing systems and

firewalls).

THE P2P ENVIRONMENT Various devices used to partition a network Hubs: A hub is a repeater that works at the physical

(lowest) layer of OSI. A hub takes data that comes into a port and sends it to the other ports in the hub.

Switches and Bridges: These are pretty similar. Both operate at the Data Link layer (just above Physical) and both can filter data so that only the appropriate segment or host receives a transmission.

Routers: These work at the Network layer of OSI (above Data Link) and operate on the IP address. Like switches and bridges, they filter by only forwarding packets destined for remote networks thus minimizing traffic, but are significantly more complex than any other networking device

NAT SYSTEMS

A network address translation system allows a single

device, such as a router, to act as an agent between the

Internet (public network) and a local (private) network.

i.e. only a single, unique IP address is required to

represent an entire group of computers.

The internal network is usually a LAN; commonly

referred to as the stub domain.

A stub domain is a LAN that uses IP addresses internally.

Any internal computers that use unregistered IP

addresses must use NAT to communicate with the rest of

the world.

CONTD…

two types of NAT translation, static or dynamic

Static NAT involves mapping an unregistered IP

address to a registered IP address on a one-to-one

basis.

Particularly useful when a device needs to be

accessible from outside the network

e.g. the computer with the IP address of 192.168.0.0

will always translate to 131.251.45.110

CONTD…

Dynamic NAT, on the other hand, maps an

unregistered IP address to a registered IP address

from a group of local dynamically allocatable IP

addresses,

i.e., the stub domain computers will be allocated an

address from a specified range of addresses, e.g.,

192.168.0.0 to 192.168.0.50

A NAT SYSTEM CAN BE ALLOCATE DYNAMIC ADDRESS OR TRANSLATE FROM FIXED STUB DOMAIN ADDRESS TO OUTSIDE ONES.

FIREWALLS

A firewall is a system designed to prevent

unauthorized access to or from a private network.

All messages entering or leaving the computer

system pass through the firewall, which examines

each message and blocks those that do not meet

the specified security criteria.

Specifically, firewalls are implemented by blocking

certain ports, thereby disabling certain types of

services that operate on those ports.

A FIREWALL BLOCKS TRAFFIC TO AND FROM SPECIFIED PORTS

P2P OVERLAY NETWORKS

P2P implementations frequently involve the creation of

overlay networks with a structure that is completely

independent of that of the underlying network of

connected devices.

The purpose of overlay networks is that they abstract

the complicated connectivity of a P2P network to a

higher-level programmatical view of the peers that make

up the network.

For example, within Jxta, a virtual network overlay sits

on top of the physical devices and is organized into

transient or persistent relationships, which they call peer

groups.

CONTD…

connections are represented through the use of

virtual pipes

Virtual pipes simply define the endpoints of the

connection and leave it to the underlying

mechanisms to implement the appropriate

behaviour for that environment

e.g., for TCP, a fixed point-to-point connection is

created for the pipe but for UDP pipes this is not

required and therefore the pipe remains

connectionless.

AN ILLUSTRATION OF THE NOTION OF AN OVERLAY NETWORK.

P2P EXAMPLE APPLICATIONS MP3 File Sharing with Napster Napster the famous MP3 file sharing program, was

launched in 1999. It had a revolutionary impact on the Internet due to

its infamous reputation for sharing illegal MP3 files and its unique design

i.e., after the initial centralized Napster search, clients connected to each other and exchanged data directly from one system’s disk to another.

Napster is P2P because the Napster peers bypass DNS and because once the Napster server resolves the IP address of the PCs hosting a particular song, it shifts control of the file transfers to the nodes.

However, Napster is an example of brokered P2P for the same reasons.

THE NAPSTER SCENARIO FOR PROVIDING A DISTRIBUTED FILE SYSTEM FOR MUSIC FILES.

INSTANT MESSAGING WITH ICQ

One of the most popular instant messaging programs

ICQ notifies users when their friends come online and

allows them to send messages to each other.

Apart from its instant messaging capabilities it allows

users to exchange files.

ICQ is a hybrid of the decentralized and client/server

architectures

It uses a central server to monitor the users that are

currently on line and to notify interested parties when

new users connect to the network.

CONTD…

All other communication between users is

conducted between the users directly.

Therefore, this employs a brokered P2P architecture,

similar to Napster, having a central database of

users for lookup purposes only, with communication

taking place independently of this central authority.

ICQ SCENARIO USES A BROKERED APPROACH USING A CENTRAL DATABASE TO STORE USER’S INFORMATION. TO THE RIGHT, THE CURRENT ICQ USER INTERFACE IS GIVEN.

FILE SHARING WITH GNUTELLA Gnutella is a ‘true P2P’ system. It does not rely on central

control for lookup, organization and communication.

A GnuCache as a lookup server for a list of Gnutella nodes

Another method: e.g., use newsgroups to get lists of

nodes, Web sites, etc.

The node joins the network by connecting initially to one

Gnutella node, which can be any node on the network

making it generally easy to join in a decentralized fashion.

Once it has joined the node discovers other nodes through

the first node by issuing ping and receiving pong

descriptors from peers accepting connections.

CONTD…

Gnutella nodes typically connect to three nodes and

then search by broadcasting their search request to

all connected neighbours, as illustrated here.

Each neighbour repeats this search request to

his/her neighbours and so on, which is known as

flooding the network.

Here, User D has the required file so User A

connects directly to User D and downloads the file

using this point-to-point connection.

GNUTELLA DECENTRALIZED APPROACH. THERE ARE TWO ASPECTS TO DISCOVERY: JOINING THE NETWORK AND THEN DISCOVERING OTHER PEERS.