Introduction

When a server is distributing a large file with $F$ bits to $N$ receivers at rate $u_s$

• Time required = $NF/u_s$

Receiving the data at the slowest receiver:

• slowest receiver has download rate $d_{\min} = \min(d_i)$
• Time required = $F/d_{\min}$

Thus, File Distribution Time:

We can speed up the file distribution by:

1. Increasing the server upload rate: requires higher link bandwidth, multiple servers
2. Have the receivers help: Receivers redistribute to other receivers

Upload time using all upload resources:

• total upload bandwidth = $u_s + \sum_i(u_i)$
• time required = $NF/\{u_s + \sum_i(u_i)\}$

Thus, File Distribution Time:

Challenges:

• Peers come and go
• Peers need to find each other
• Peers need to be willing to help each other

Finding Peers:

1. Central Directory- Napster:

Operation:

• JOIN: NO NEED
• PUBLISH: Update central directory on the available content
• SEARCH: Ask central directory for peers,
• FETCH: get file directly from peers

Comments:

• Servers directory is continually updated
• Always know which content is available
• P2P File transfer, No load on the server

Challenges: - Locating content is highly centralised. SPOF, Performance bottleneck

2. Query Flooding - GNutella:

Operation:

• JOIN: Contact a few nodes to become neighbours
• PUBLISH: NO Need
• SEARCH: Ask neighbours,
• FETCH: get file directly from peers

Comments:

• Fully Decentralised

Challenges:

• Search scope can be large may take long time

3. Hierarchical Overlay - Kazaa

Operation:

• JOIN: contact the super node
• PUBLISH: Update super node on the available content
• SEARCH: Queries flooded among super nodes
• FETCH: Get file directly from peers

Comments:

• Supernode selection favours nodes with high up-time
• Propagating query to a sub-node may take more time than for the super-node to answer itself

Read: BitTorrent