A Flexible Policy Control Architecture for InterAS Routing - PowerPoint PPT Presentation

About This Presentation
Title:

A Flexible Policy Control Architecture for InterAS Routing

Description:

Routing information is spreading over the Internet in a hop-by ... Using tool such as Looking glass. Operators cannot adapt dynamically changing environment. ... – PowerPoint PPT presentation

Number of Views:59
Avg rating:3.0/5.0
Slides: 13
Provided by: networkinn
Learn more at: https://www.apnoms.org
Category:

less

Transcript and Presenter's Notes

Title: A Flexible Policy Control Architecture for InterAS Routing


1
A Flexible Policy Control Architecture for
Inter-AS Routing
  • Osamu Akashi, Kenichi Kourai, Kensuke Fukuda,
  • Toshio Hirotsu, Koji Sato, Mitsuru Maruyama,
  • Toshiharu Sugawara
  • NTT Network Innovation Laboratories
  • Tokyo, Japan
  • akashi,kourai,fukuda,hirotsu,koji,
    mitsuru,sugawara_at_t.onlab.ntt.co.jp

2
Background in inter-AS routing
  • Routing information is spreading over the
    Internet in a hop-by-hop manner using BGP-4.
  • Receive ? modify ? advertise
  • Each BGP entry includes AS path information that
    information traversed.
  • All ASes along with source AS to destination AS
    should set their routing tables as source AS
    intends.
  • Loss of connectivity, instable access, policy
    inconsistency

Route information
ISP
3
Problems of inter-AS routing
  • Difficulty in understanding the behavior
  • Routing information mutates as it spreads.
  • Independent administrative domain that has its
    own policy and routers are configured by hand.
  • Needs analysis by experts by hand
  • ex. Using tool such as Looking glass
  • Operators cannot adapt dynamically changing
    environment.
  • Policy is only represented by low level
    primitives, namely router configuration commands.
  • No inter-AS cooperative policy control scheme

Need a cooperative distributed solution
4
A policy inconsistency where local control is
insufficient
AS y
Case2 To ASy Use ISP A
ISP C
International ISP A (AS i)
Major transit AS (AS k)
AS x
Bottleneck!!
Decision point Bandwidth of links to ISP-A and
ISP-B seem the same
international ISP B (AS j)
AS z
Case1 To ASz Use ISP B
Narrow link
5
Our approach
  • Diagnosis for inter-AS routing anomalies
  • ENCORE4, 5
  • Multi-agent based, cooperative analysis
  • Flexible inter-AS policy control
  • AISLE
  • (Autonomous and Intelligent Self-control
    Environment)

Extended
6
Basic idea for controlling routing information
Modify BGP attributes based on environmental
changes
Routing information
Outgoing packets
Send policy requests on the preference base.
Routing information
Modify BGP attributes based on received requests
and its own policy.
Incoming packets
7
AISLE layer
ASx
Human operator
ASy
ASz
Policy
Inter-AS coordination
Intra-AS control
Intra-AS control
Intra-AS control
AISLE layer
Routing management
8
AISLE cooperation model
Modifies BGP attribute values for controlling
policy.
Controls traffic in the ASx
AS
agent
agent
BGP information exchange
AS x
AS
Defines autonomous actions
agent
Policy description Reaction against autonomously
observed result Cooperation with other agents
9
Structure of AISLE agent
agent
Agent In other AS
Cooperative action controller
Read status information
Communication / cooperation
Agent In other AS
Policy description
Policy control engine
Modify configuration commands
Configuration commands
BGP controller
iBGP session
Router
Exchanges modified BGP entry
eBGP session
10
Application
  • Adaptation of policy for dynamic parameters
  • Auto load balancing
  • On demand advertisement of backup route
  • Cooperative control
  • Preference control for incoming packets
  • Verification of routing policy
  • Defense against attacks

11
Application example
route-preference (ASi, ASj )
AS y
International ISP A (AS i)
agent
Major transit AS
agent
AS x
ISP C
Modifies policy For ASx, select route via ISP-A
international ISP B (AS j)
AS z
Case2 To ASy Use ISP A
Narrow link
12
Conclusion
  • AISLE inter-AS flexible policy control
    architecture
  • Multi-agent based implementation
  • Needs more experiment in real internet
    environments
  • Verification and feedback
Write a Comment
User Comments (0)
About PowerShow.com