Seserv workshop manos dramitinos - tussle analysis from etics project
1. Tussle Analysis for FP7
Project ETICS Case Studies
Manos Dramitinos [mdramit@aueb.gr]
Athens University of Economics and Business
SESERV Workshop, Athens, Greece
January 31st, 2012
2. Introduction
• Internet: Multiple technologies and stakeholders,
conflicting interests
– Choices, Goals, Strategies, Policies
– Dynamic ecosystem
• “Design for tussle”
• SESERV has defined a systematic approach for the
analysis and assessment of socio-economic tussles
in the Internet
– ETICS case study
5. The ETICS Approach (I)
• Support performance-sensitive inter-carrier services through
network interconnections of assuredquality
– Technically: support automated E2E ASQ
– Economically: Market enabler for services
6. The ETICS Approach (II)
• ASQ products: Novel IC products with assured performance in terms of
business and technical attributes, described in SLA
– Variants of ASQ goods offered by the ETICS “community”
• “Bundling”, stitching and nesting
• Technology agnostic, not tied to a certain business model
Business Information
Customer SP
E6
E7 E7
Edge E1 Edge E2 Transit E3 Transit
NSP NSP NSP NSP
E4 E4
Transport E5 Transport
External NSP NSP
actor
ETICS Provider
ETICS requirements and specification scope
7. Distributed Pull Model
Customer buys from ISP-1 an E2E service to Content Provider and expects a certain quality
Customer Content
H Provider
F ISP-3
A D
B C ISP-2
G
ISP-1 E
SLA among ISP 1 and ISP-2 for a path between C & H
SLA among ISP 2 and
ISP 3 for a path between F & H
8. Case Study A:
ASQ Goods and ISPs Competition
congestion! Content
Provider
Peering
ISP 2 ISP 1
Peering
Transit
ISP 3
What are the necessary
business conditions for QoS-
aware interconnection?
Allowing the control of major parameters of ASQ interconnection is
important for promoting collaboration that is mutually beneficial
9. Tussle Evolution
What if an ASQ good is ISP-1 feels unfair
ISP-2 used by ISP-2 to bypass
strategies/policies
the Best-Effort peering
Stakeholders’
Tussle outcome
link for all traffic?
What if ISP-1 stopped
offering that ASQ? ?
What if ISPs could control major
properties of ASQ goods? ?
Support for
best-effort
connectivity
ISP-1 only ISP-2 feels unfair
Functionality: Network Service composition
time
Traffic is
strategies/policies
ISP-2 optimized
Stakeholders’
Tussle outcome
selfishly
ISPs perform
traffic Stable
engineering Introduction of ASQ routing
for optimizing goods make routing
network more stable and simpler Traffic is
usage
ISP-1 optimized
Functionality: Routing & Traffic Engineering selfishly
Functionality Stable outcome Evolves
Legend
Initial state Unstable outcome Affects
11. Case Study B: SLA Monitoring
incentives for backup ASQ provisioning
• SLA Monitoring checks conformance of service delivered to
the contract terms
– Required also if all the ETICS community actors are trusted
• Backup capacity: needed to deal with network failures
– …and avoid SLA violations in the ETICS context because either new
path is not good enough or because traffic arrived from a different
ingress point in the ISP network
– Incentivized by monitoring due to penalties for violations
• … or free-riding
• Tussle for responsibility: What technology decisions would
lead to (un-)fair allocation of SLA violation penalties?
– Three candidate schemes examined by ETICS, a centralized and two
distributed (coordinated sampling and active flow technology)
12. Distributed Hierarchical Monitoring
• Each ISP collects raw data from probes (BRs)
• Data sampling to keep the operational cost low (E2E)
• Monitoring data stored per ISP at proxies
• If SLA violation, a collector queries the proxies and checks
the validity of SLA
ETICS collector
Content
Customer
H Provider
F ISP-3
A D router/
B C ISP-2 X probe
G
proxy
ISP-1 E
collector
13. Tussle evolution for ETICS Network
Service Delivery functionality
Transit What if (sampled) monitored packets Transit ISPs
ISP Destination ISP are known in advance ? contribute less
under provisions
strategies/policies
to SLA penalties
backup ASQ goods ?
Stakeholders’
Tussle outcome
What if Broker
signals to all ISPs which Fair
Broker
Introduction
packets to probe during ? penalties
service provisioning?
of inter-domain ASQ
Source goods with no Source &
ISP adequate monitoring Destination ISPs
Dest. of individual ISPs Functionality: Network Service delivery contribute less to
ISP SLA penalties
14. Conclusions - References
• Tussle analysis for two sample ETICS cases
• More interesting cases to be investigated
• Useful insight for the market and technology configuration
– Crucial for the adoption of new technologies
• References:
– www.seserv.org
– www.ict-etics.eu
– C. Kalogiros, C. Courcoubetis, G. D. Stamoulis, M. Dramitinos, O.
Dugeon, “Internet Interconnection Assured Quality Services: Issues
and Strategic Impact”, Submitted to Future Network & Mobile
Summit 2012
Hinweis der Redaktion
Transit & Destination ISPs could forward probing packets preferentially if sample packets were known in advance