1. Overlay Networking
Overview

2. Outline
Overview
Overlay multicast
Overlay QoS
Overview

3. Overlay Networking
Overview

4. Overlay Networking Q: What new services/applications enabled?
Q: What are the new technical challenges?
Overview

5. What new services/applications enabled?
Large-scale file sharing
Multicast
Routing
End-to-end commutations
QoS
Security
Network Integration
Overview

6. Why Overlay Multicast? Why not IP Multicast?
Potential benefits of Overlay Multicast
Concerns of Overlay Multicast
Overview

7. Routers with multicast support
IP Multicast
Gatech
Stanford
CMU
Berkeley
Routers with multicast support
No duplicate packets
Highly efficient bandwidth usage
Key Architectural Decision: Add support for multicast in IP layer
Overview

8. Key Concerns with IP Multicast
Scalability with number of groups
Routers maintain per-group state
Analogous to per-flow state for QoS guarantees
Aggregation of multicast addresses is complicated
Supporting higher level functionality is difficult
IP Multicast: best-effort multi-point delivery service
End systems responsible for handling higher level functionality
Reliability and congestion control for IP Multicast complicated
Inter-domain routing is hard.
Deployment is difficult and slow
ISP’s reluctant to turn on IP Multicast
Overview

9. Overlay Multicast Overlay Tree Gatech Stan1 Stanford Stan2 Berk1
CMU
Gatech
Stan1
Stanford
Stan2
Berk1
Berkeley
Overlay Tree
Berk2
Use a few unicast sessions (fewer than pure unicast).
End systems do the forwarding.
Routers do not participate in multicast.
Stan1
Gatech
Stan2
CMU
Berk1
Berk2
Overview

10. Potential Benefits Scalability (number of sessions in the network)
Routers do not maintain per-group state
End systems do, but they participate in very few groups
Easier to deploy
Potentially simplifies support for higher level functionality
Leverage computation and storage of end systems
For example, for buffering packets, transcoding, ACK aggregation
Leverage solutions for unicast congestion control and reliability
Higher-level functionality:
Many proposals for reliability require end-systems to organize themselves into a tree for e.g. ACK aggregation.
Alternatively, proposals for a smarter router able to aggregate ACKs and do other things useful for transport layer functionality.
ESM is very conducive to either approach
End-hosts already form a tree.
Easier to deploy fancier (app specific) functionality in other end hosts who participate in the application.
Overview

11. Concerns of Overlay Multicast
Performance
Self-organization
Overview

12. Performance Concerns Delay from CMU to Berk1 increases Gatech
Stan1
Stan2
Berk2
Gatech
Berk1
Delay from CMU to
Berk1 increases
CMU
Gatech
Stan1
Stan2
Berk1
Berk2
Duplicate Packets:
Bandwidth Wastage
Basic Point: There is going to be a performance hit associated with ESM. The question is: how big is the hit and is it worth taking?
Goal: Build a “good” overlay tree.
Overview

13. What is an efficient overlay tree?
The delay between the source and receivers is small
Ideally,
The number of redundant packets on any physical link is low
Heuristic we use:
Every member in the tree has a small degree
Degree chosen to reflect bandwidth of connection to Internet
CMU
CMU
CMU
Stan2
Stan2
Stan2
Small degree is good for two reasons:
Fewer duplicate packets leave the node.
Physical link bandwidth is split between fewer overlay links.
Trick: want end-systems to to self-organize into a “good” tree in a distributed way.
Stan1
Stan1
Stan1
Berk1
Gatech
Berk1
Gatech
Berk1
Gatech
Berk2
Berk2
Berk2
High latency
High degree (unicast)
“Efficient” overlay
Overview

14. Performance Metrics Delay between members Delay from CMU to
Berk2
CMU
Stan1
Stan2
Gatech
Berk1
Delay from CMU to
Berk1 increases
Delay is a measure of the quality of tree from members’ point of view
Stress is a measure of quality from the network’s point of view.
Note: We expect both metrics to be worse than pure multicast, but better than pure unicast
Berk2
Overview

15. Performance Metrics (Cont.)
Stress, defined as the number of identical copies of a packet that traverse a physical link
Stan1
Stress = 2
CMU
Stan2
Berk1
Delay is a measure of the quality of tree from members’ point of view
Stress is a measure of quality from the network’s point of view.
Note: We expect both metrics to be worse than pure multicast, but better than pure unicast
Berk2
Overhead
Overview

16. Why is self-organization hard?
Dynamic changes in group membership
Members join and leave dynamically
Members may die
Limited knowledge of network conditions
Members do not know delay to each other when they join
Overlay must self-improve as more information available
Dynamic changes in network conditions
Delay between members may vary over time due to congestion
Overview

17. A Whole Picture on Overlay QoS
Overlay OoS
Routing
End control
Benefits
Flexibility
Concerns
No fully control
QoS guarantees is difficult to achieve.
Overview

18. Conclusions Overlay network is more flexible
Overlay network can enable many “new” networking applications
Overlay network can provide better QoS, but still without guarantees
Overview