Você está na página 1de 21

ECE 595 Computer Network Systems

Peer-to-Peer Multicast (Or Overlay/End System Multicast)

IP Multicast
Gatech Stanford

Smart Network
No duplicate packets Highly efficient bandwidth usage Trick: Make the routers smart !

Berkeley

Significance of IP Multicast
First major functionality added to routers since original design Per-group state implies:
Routing tables with potentially millions of entries! Todays routers have only tens of thousands of entries in routing tables though there are millions of end systems! Potential scaling concerns

Per-group (per-flow) based solutions typically find slow acceptance in the Internet

Other concerns with IP Multicast


Slow deployment
Very difficult to change network infrastructure and routers

Difficult to support higher layer functionality


IP Multicast: Best effort delivery Model complicates support for reliability and congestion control

Overlay Multicast
Gatech Stan1 Stanford Stan2

Dumb Network
Berkeley

Berk1

Overlay Tree
Gatech
Stan1 Stan2 Purdue

Berk2

Berk1 Berk2
5

Overlay Multicast: Benefits


No per-group state in routers Easy to deploy: No change to network infrastructure Can simplify support for congestion control etc. Leverage computation and storage (e.g. Transcoding) Unicast congestion control
Stan-Modem

Purdue
Stan-LAN

Berk1

Gatech

Berk2

Overlay Performance
Even a well-designed overlay cannot be as efficient as IP Mulitcast But performance penalty can be kept low Trade-off some performance for other benefits Duplicate Packets: Bandwidth Wastage Gatech Stanford

Dumb Network
Berkeley

Increased Delay

Key Issues
What is an efficient overlay tree?
Good performance to application: high bandwidth, low delay Good network performance: reasonable clustering of receivers

How to construct efficient overlays?


Distributed self-organizing protocols

Inefficient Overlay Trees


Purdue Stan2 Stan1-Modem Berk1 Berk2 Gatech Berk1 Berk2 Gatech Stan2 Stan1-Modem Purdue

High latency

-Poor network usage -Potential congestion near Purdue


Stan-LAN Purdue

Poor bandwidth to members

Stan-Modem Berk1 Berk2


9

Gatech

Efficient overlay trees


Purdue

Stan-Modem

Purdue

Berk2

Stan-LAN Berk1 Berk2

Berk1 Stan-LAN Stan-Modem

Gatech

Gatech

10

Self-Organizing Protocols
Construct efficient overlays in a distributed fashion
Members may have limited knowledge of the Internet Adapt to dynamic join/leave/death of members Adapt to network dynamics and congestion

Active and evolving area of current research


Provide intuitive example, rather than details

11

Example

Purdue Berk2 Berk1 Stan-Modem

Stan-Lan joins

12

Example

Purdue Berk2 Berk1 Stan-Modem

Stan-Lan

13

Example

Purdue Berk2 Berk1 Stan-Modem

Bad Bw Perf! Switch!


Stan-Lan

14

Example

Purdue Berk2 Berk1 Stan-Modem

Stan-Lan

15

Example

Purdue Berk2 Berk1 Stan-Modem

More clustering if Stan-Modem moves to Stan-Lan


Stan-Lan

16

Example

Purdue Berk2 Berk1 Stan-Modem

Stan-Lan

17

Example

Purdue Berk2 Berk1 Stan-Modem

Stan-Modem is disconnected: Back to Berk1

18

Snapshot from Real Broadcast


U.S. East Coast U.S. Central U.S. West Coast Europe Asia Unknown
Source

19

Mesh-Based Streaming
So far we discussed tree-based approach to streaming Others have looked at mesh-based approaches Mesh-based:
Similar to BitTorrent:
Each node maintains a set of partners Exchange data availability info with partners Pull data from partner if do not possess it already

Key difference:
Which data is pulled from partners

Mesh-Based Streaming Vs. BitTorrent


Clever Scheduling Algorithm regarding which segments to fetch

Você também pode gostar