This document discusses peer-to-peer (P2P) file sharing networks. It describes how early systems like Napster and Kazaa worked, comparing their centralized index models to the decentralized BitTorrent protocol. BitTorrent uses a tracker to coordinate file downloads between seeds (peers with the full file) and leeches, allowing transfers from multiple locations simultaneously for high speeds. While efficient, P2P networks face challenges around free-riding and availability of unpopular content.
2. Peer to peer
More users, better it is
Large amounts of data transfer
Encourages peers to upload
> a quarter of a billion monthly users*
3. Client-server architecture
Conventional method used for years
Direct connection between server & client
Large no. of dedicated servers
4. Download data at high speeds
Reach high transfer rates very quickly
No loss of files once uploaded
File availability does not depend on popularity
5. Single point of failure
Maintenance of servers
Servers require large bandwidth
Website crash eg: VTU website
6. Decentralized communications model
Nodes(peers) are interconnected
Content not on a central server
Content present on peer’s computer
No need of server for data transfer
Peers->provide and consume resources
7. Designed as P2P MP3 sharing service
Song downloaded from other peer
Server only used to connect peers
Redefined the internet
8. Peers run Napster utility
Central Index Server(CIS)
Peers connect to CIS
Peers inform CIS about files on their directory
When the user wants a file, he queries the CIS
CIS directs them to the computer that has it
Connection is established between peers
Transfer is initiated
9.
10. Transfer takes place from a single source
Speed limited to upload capacity of source
Requires only a subset of users to share
Encourage free-riding
Exerts too much load on original sources
Napster was shutdown in its original form
11. No CIS like the one used in Napster
Made use of FastTrack protocol
Super nodes and ordinary nodes
1 super node serves 60 to 150 ordinary nodes
12. Software comes with a list of super nodes
The client connects to one of the super nodes
A file request in passed through super node
Super node passes the query super nodes
The other super nodes pass it to ordinary nodes
The ordinary nodes ask other ordinary nodes
7 levels deep
If a file is found, transfer takes place
Super nodes do not take part in the transfer
13.
14.
15. Super node has to handle lot of traffic
Transfer happens between only 2 peers
Bandwidth available to receiver not utilised
Encourages free-riding
16. Similarities with Napster:
Users share their files with everyone else
Users run a software to connect to network
Differences with Napster:
No central database like Napster
Machines inform other machines about files
Achieved using distributed query approach
17. User types in the name of the required file
This machine requests other known machines
These machines search their directory
If not found, forward the request
This process may go 7 levels deep
A single search may cover 8000 machines
18.
19. The 8000 machines may not contain the file
Takes time for search results to appear
Bandwidth to handle requests from other users
Bandwidth available to receiver not utilised
25. Create a torrent
◦ Select files
◦ Choose a tracker
◦ Select saving directory
◦ Select piece size(better left untouched)
◦ Start seeding!!
26. Downloading files
◦ Download torrent file
◦ Client software communicates with a tracker to find
Other computers that have the complete file (seeds)
Those with a portion of the file
◦ Peers communicate with each other
◦ Download/upload starts from/to different peers
27. .torrent file
◦ Metadata about the required file
The URL of the tracker
Pieces <hash1, hash 2,…, hash n>
Piece length
Name of the file
Length of the file
28. Tit for tat approach
Optimistic unchoking
Random first piece
Rarest first
Endgame mode
30. Open source
Share large amount of data quickly
Discourages free-riding
More users, the better it is
Download takes place from multiple locations
Reduces burden on original distributors
Easy to download expensive software, movies
Organisations distribute legitimate software
31. Leechers may leave swarm after download
Unpopular content has no seeds
Takes time to reach high download speeds
No streaming playback