MIRO:

1

• MIRO
• Multi-path Interdomain Routing
• Wen Xu and Jennifer Rexford
• Department of Computer Science
• Princeton University
• Hong Dan

2
OUTLINE

• Motivation
• Protocol Design
• Performance Evaluation
• Discussion
• Related Work
• Conclusion

3
Motivation

• User demands
• BGP limits
• Source routing limits
• Overlay Network limits

4
Interdomain Routing(BGP)

• Destination-based
• Single-path routing
• Path-vector protocol
• Local-policy based

5
Source Routing

• Packets carry a list of hops in the path
• Limited control for intermediate ASes
• Scalability
• Efficiency and stability

6
Overlay Network

• Data-plane overhead
• Limited control
• Probing Overhead

7
Observations

• Single-path route is not good enough
• BGP is effective for a large portion of traffic
• End user don't need to control which path is
  taken
• BGP doesn't advertise all candidate routes
• AS select routes based on business
• AS can't influence other AS's choice easily

8
MIRO's feature

• AS-level path selection
• Negotiation for alternate routes
• Policy-driven export of alternate routes
• Tunnels to direct traffic on alternate paths

9
Some Application of MIRO

• Avoiding a specific AS for security or
  performance
• Achieving higher performance
• Load balancing for incoming traffic

10
Protocol Design

• AS-Level Path-Vector Protocol
• Pull-based Route Retrieval
• Bilateral Negotiation Between Ases
• Selective Export of Extra Routes
• Tunnels for Forwarding Data Packets

11
Implementation

• Intra-AS Architecture
• Data Plane Packet Encapsulation
• Control Plane Tunnel Management

12
Intra-AS Architecture
13
Data-panel Packet Encapsulation

• IP-in-IP encapsulation
• IP Address of the Egress Links
• IP Address of the Egress Routers
• One Reserved IP Address for All Tunnels

14
IP Address of the Egress Links
15
IP Address of the Egress Routers
16
One Reserved IP Address for All Tunnels
17
Control Plane Tunnel Management
18
Performance Evaluation

• Evaluation Methodology
• Exposing the Underlying Path Diversity
• 2 Application
• Avoid an AS in Default Path
• Controlling Incoming Traffic

19
Evaluation Methodology

• AS-level simulation
• Based on business relationships
• Inferring AS relationships
• Instances of 2000,2003,2005
• Policies
• Strict Policy(/s)
• Respect Export Policy(/e)
• Most flexible Policy(/a)

20
Exposing the Underlying Path Diversity
21
Avoiding an AS in Default Path

• Success rate of different policies

22
Avoiding an AS in Default Path

• Avoiding State Explosions

23
Avoiding an AS in Default Path

• Incremental Deployment

24
Controlling Incoming Traffic
25
ROUTING POLICIES

• Policy Configuration in Current Internet
• Multi-path Routing Policies
• Negotiation Rules
• Requestingwhen to trigger negotiation
• Requestingwhom to negotiate with
• Respondingwhether to allow negotiation
• Filter Rules
• Route filtering
• Route preference and cost

26
DISSCUSSTION

• Route Convergence
• Routing Loops
• Route Aggregation and Security

27
RELATED WORK

• Source Routing
• Overlay Networks
• Nimrod
• Nexit
• HLP