2. AGENDA
Introduction
BitTorrent
Energy Efficient BitTorrent
MILP Model for Energy Efficient BitTorrent
Results
Future Work
3. Introduction
Energy Consumption of ICT networks.
Two Content Distribution Networks
◦ Client-Server
◦ Peer-to-Peer
IP Over WDM Networks
◦ IP layer
◦ Optical Layer
4. BitTorrent
Peer-to-Peer file sharing protocol
Used to distribute large amount of files
Accounts for approximately 20% of internet traffic
Why BitTorrent?
Bram cohen
BitTorrent Terminologies
◦ Torrent, Client, Peer, Tracker, Seeder, Leecher, Swarm
5. File sharing Strategies of Peers
Random first piece
Rarest First
Endgame mode
Choking/Unchoking
Optimistic Unchoking
6. Energy Efficient BitTorrent
Energy Efficient BitTorrent
we use number of methods to reduce Cross ISP traffic
◦ Bandwidth Throttling
◦ Use of Cache
◦ Gateway peer
NSFNET Network
◦ Locality
We consider 160,000 groups of downloader distributed randomly over the
NSFNET network nodes.
Homogenous system where all peers have same upload capacity of 1 Mbps
Tit for Tat ensures fairness among peers
7. MILP Model
Objective. Maximize
Subject To:
Download rate for leecher to its download capacity
Download rate in accordance with upload capacity
Peer Upload rate would not exceed peer upload capacity
Fairness Constraints, Tit-for-Tat
Time phase for each swarm to dowmload a different file
9. Results
Seeders Energy
15 9.6 %
20 18.3 %
25 28.2 %
30 39.4 %
35 52 %
40 66 %
45 82 %
Different Peer Sizes with Fix Number of Swarms
10. Results
Seeders Energy
15 49.7 %
20 49.4 %
25 49 %
30 48.5 %
35 47 %
40 42.7 %
45 32.6 %
Consumption With Different Swarms Sizes with Peer Sizes
11. Future Direction
This project can be extended up to an energy efficient BitTorrent for Hybrid
kind of network, which is the combination of client-server and peer-to-peer
network. In such systems, BitTorrent can assist a CDN (content delivery
networks) to distribute its contents to the end users if the content is popular,
leading to exerting less load on the CDN. Studying file sizes, popularity and
peers distribution among swarms in such systems can assist in
understanding the energy gains that can be achieved in the hybrid systems
by understanding the best time and network location in which the delivery
task is switched from CDN to P2P.