SAHARA Project

1
SAHARA Project

• Mohave Group
• Kevin Lai, Z. Morley Mao, Bhaskaran Raman

2
Motivation

• Mutually distrusting and competitive service
  providers
• Tragedy of the commons / prisoner's dilemma
  effect
• competitiveness prevents maximization of utility
  of provided services
• How to provide more sophisticated services
  without global cooperation?
• e.g. mobility, multicast, Qos, composed services?

3
Solutions/Design Principles

• Overlay
• Build sophisticated services on top of lowest
  common denominator model of services
• Make minimal assumptions about underlying network
  services
• Must infer service characteristics through
  measurement
• Confederation
• Stitch intra-domain functionality to form
  inter-domain service
• limited scope agreement to compose/exchange
  services
• May enable full translation of capabilities
• Full specification of service characteristics,
  but must verify.

4
Which is better model?

• May depend on specific service
• Evaluation
• Deployability provides more incentive and less
  cost to deploy
• Utility maximizes overall system utility
• Scalability scales up to global economy

5
What we should have for describing SAHARA
Scenarios
Goals
Specific issues/projects
Design Principles
Solutions/Mechanisms
6
Scenarios

• Mobile user in foreign domain
• Coverage by providers A B
• As network overused, Bs underutilized
• Can network resources be shared across providers?
• User wishes to access email over cell-phone
• Need to choose nearby text-to-speech engine
• Web-access
• Two CDNs, one far away, one nearby
• Can the users web-pages be made available on the
  nearby CDN?

7
What we should have for describing SAHARA
Scenarios
Goals
Specific issues/projects
Design Principles
Solutions/Mechanisms
8
Design Principles

• Utility-based resource allocation
• Auctions, Congestion pricing
• Trust but verify
• Confederated composition
• Middleware (measurement) infrastructure
• Overlay composition

9
Problem/Issue Mechanism Principle
Wireless bandwidth allocation WaveLAN (Jimmy)
Utility-based resource allocation
Congestion pricing
10
Problem/Issue Mechanism Principle
Content placement in CDNs (Yan Chen)
Middleware meas. layer
Internet host clustering Monitors in each
cluster Periodic measurements
11
Problem/Issue Mechanism Principle
Wide-Area Service Composition (Bhaskar)
Middleware meas. Layer for overlay composition
Measurement betn. overlay peers Link-state-based
propagation
12
Problem/Issue Mechanism Principle
Always Best Connected (Machi)
Utility-based resource allocation?
Auctions? Congestion pricing?
13
Problem/Issue Mechanism Principle
Inter-provider Connectivity
Trust but verify
BGP protocol verification mechanism? (Lakshmi)
14
Problem/Issue Mechanism Principle
Authorization framework for Service Composition
(Suzuki)
Authorization control infrastructure Authorization
control protocol
Design Principle?
15
Problem/Issue Mechanism Principle
VoIP Gateway Selection (Matt)
Design principle?
Maintain per-gateway routes Route exchanges?
16
Problem/Issue Mechanism Principle
Efficient bandwidth allocn. in overlay networks
(Weidong)
Design principle?
Failure prob. model Centralized computation?
17
Problem/Issue Mechanism Principle
Broadcast Inter-Networking Inter-domain Multicast
(Mukund)
Design principle?
Gateways for inter-operation
18
(No Transcript)