1 performance study of congestion price based adaptive service performance study of congestion price...
TRANSCRIPT
1
Performance Study of Performance Study of Congestion Price Based Congestion Price Based
Adaptive ServiceAdaptive Service
Xin Wang, Henning SchulzrinneXin Wang, Henning Schulzrinne
(Columbia University)
http://www.cs.columbia.edu/xinwang/http://www.cs.columbia.edu/xinwang/public/projects/RNAP.htmlpublic/projects/RNAP.html
2
MotivationMotivation
• Combine resource reservation Combine resource reservation with multimedia adaptive with multimedia adaptive serviceservice
• Pricing network services based Pricing network services based on level of service, usage, and on level of service, usage, and congestion - a natural and fair congestion - a natural and fair incentive for applications to incentive for applications to adapt their sending rates adapt their sending rates according to network according to network conditions.conditions.
• Our work: Our work: – resource commitment short resource commitment short
intervalinterval– trade-offs between blocking trade-offs between blocking
and raising prices in networkand raising prices in network
3
OutlineOutline
• Resource negotiation & Resource negotiation & RNAPRNAP
• Pricing strategyPricing strategy
• User adaptationUser adaptation
• Simulation modelSimulation model
• Results and discussionResults and discussion
• RNAP message RNAP message aggregationaggregation
• ConclusionConclusion
4
Resource Negotiation & Resource Negotiation & RNAPRNAP
• AssumptionAssumption: network provides a choice : network provides a choice of delivery services to user of delivery services to user – e.g. diff-serv, int-serv, best-effort, with e.g. diff-serv, int-serv, best-effort, with
different levels of QoS different levels of QoS – with a pricing structure (may be usage-with a pricing structure (may be usage-
sensitive) for each.sensitive) for each.
• RNAP (Resource Negotiation and RNAP (Resource Negotiation and Pricing)Pricing): a protocol through which the : a protocol through which the user and network (or two network user and network (or two network domains) negotiate network delivery domains) negotiate network delivery services.services.– Network -> UserNetwork -> User:: communicate availability of communicate availability of
services; price quotations and accumulated services; price quotations and accumulated chargescharges
– User -> NetworkUser -> Network: request/re-negotiate : request/re-negotiate specific services for user flows.specific services for user flows.
• Underlying MechanismUnderlying Mechanism: combine : combine network pricing with traffic engineeringnetwork pricing with traffic engineering
5
Resource Negotiation & Resource Negotiation & RNAP (cont’d.)RNAP (cont’d.)
• Who can use RNAP?Who can use RNAP?– Adaptive applications: adapt Adaptive applications: adapt
sending rate, choice of sending rate, choice of network servicesnetwork services
– Non-adaptive applications: Non-adaptive applications: take fixed price, or absorb take fixed price, or absorb price changeprice change
6
Protocol Architectures: Centralized Protocol Architectures: Centralized Architecture (RNAP-C)Architecture (RNAP-C)
NRN
S1
R1
NRN NRNHRN
HRN
Access Domain - BAccess Domain - A
Transit Domain
Internal Router
Edge Router
Host
RNAP Messages
NRN
HRN
Network Resource Negotiator
Host Resource Negotiator
Intra domain messages
Data
7
S1
R1
HRNHRN
Access Domain - B
Access Domain - A
Transit Domain
Protocol Architectures: Distributed Protocol Architectures: Distributed Architecture (RNAP-D)Architecture (RNAP-D)
Internal Router
Edge Router
Host
RNAP Messages
NRN
HRN
Intra domain messages
Data
8
RNAP messagesRNAP messages
Query
Quotation
Reserve
Commit
Quotation
Reserve
Commit
Close
Release
Query: User enquires about available services, prices
Quotation: Network specifies services supported, prices
Reserve: User requests service(s) for flow(s) (Flow Id-Service-Price triplets)
Commit: Network admits the service request at a specific price or denies it (Flow Id-Service-Status-Price)
Per
iodi
c re
-neg
otia
tion
Close: tears down negotiation session
Release: releases the resources
9
Pricing on Current InternetPricing on Current Internet
• Access rate dependent Access rate dependent charge (AC)charge (AC)
• Volume dependent Volume dependent charge (V)charge (V)
• AC + V AC-VAC + V AC-V• Usage based charging: Usage based charging:
time-based / volume-time-based / volume-basedbased
10
Pricing StrategyPricing Strategy
• Fixed pricingFixed pricing– Service class independent flat Service class independent flat
pricing pricing – Service class sensitive priority Service class sensitive priority
pricing pricing – Time dependent time of day pricingTime dependent time of day pricing– Time-dependent service class Time-dependent service class
sensitive priority pricingsensitive priority pricing
11
Pricing Strategy (cont’d.)Pricing Strategy (cont’d.)
• Congestion-based PricingCongestion-based Pricing– Usage chargeUsage charge: :
ppuu= = ff (service, demand, (service, demand,
destination, time_of_day, ...) destination, time_of_day, ...) ccuu(n)(n) = = ppuu x V (n) x V (n)
– Holding chargeHolding charge: : PPhh
ii = = ii x x ((ppuui i - p- puu i-1i-1) )
cchh (n) (n) = = pphh x R(n) x R(n)
x x – CongestionCongestion chargecharge: :
ppcc (n) (n) = min [{ = min [{ppcc (n-1) (n-1) + +
(D, S) x (D-S)/S,0 (D, S) x (D-S)/S,0 }}++, p, pmaxmax] ]
cccc(n)(n) = =
ppcc(n) x V(n) (n) x V(n)
12
User AdaptationUser Adaptation
• Based on perceived valueBased on perceived value
• Maximize total utility over Maximize total utility over
the total costthe total cost
• Constraint: Constraint:
budget, min QoS & budget, min QoS &
max QoSmax QoS
13
User Adaptation (cont’d.)User Adaptation (cont’d.)
• An example utility functionAn example utility function– U U ((xx) = ) = UU00 + + log log ((x / xx / xmm))
• Optimal user demand Optimal user demand – Without budget constraintWithout budget constraint: x: xjj = = jj / p / pjj
– With budget constraintWith budget constraint: x: xjj = (b x = (b x j j / Σ / Σll ll
)) / p / pj j
• AffordableAffordable resource is distributed resource is distributed proportionally among applications of proportionally among applications of the system, based on the user’s the system, based on the user’s preference and budget for each preference and budget for each application.application.
15
Simulation Model (cont’d.)Simulation Model (cont’d.)
• Parameter Set-upParameter Set-up– topology1: topology1: 4848 users users– topology 2: topology 2: 360360 users users– user requests: user requests: 6060 kb/s -- kb/s -- 160 160 kb/skb/s– targeted reservation rate: targeted reservation rate: 90%90%– price adjustment factor: price adjustment factor: σσ = = 0.060.06– price update threshold: price update threshold: θθ = = 0.050.05– negotiation period: negotiation period: 30 30 secondsseconds
– usage price: usage price: ppu u = = 0.230.23 cents/kb/min cents/kb/min
– average session length 10 minutes, average session length 10 minutes, exponential distributed. exponential distributed.
• Performance measuresPerformance measures– Bottleneck bandwidth utilizationBottleneck bandwidth utilization– User request blocking probabilityUser request blocking probability– Average and total user benefitAverage and total user benefit– Network revenueNetwork revenue– System priceSystem price– User chargeUser charge
16
Design of the ExperimentsDesign of the Experiments
• Performance comparison of Performance comparison of congestion-based pricing system congestion-based pricing system (CPA) with a fixed-price based system (CPA) with a fixed-price based system (FP) (FP)
• Effect of system control parameters: Effect of system control parameters: – target reservation ratetarget reservation rate– price adjustment stepprice adjustment step– price adjustment thresholdprice adjustment threshold
• Effect of user demand elasticityEffect of user demand elasticity
• Effect of session multiplexingEffect of session multiplexing
• Effect when part of users adaptEffect when part of users adapt
• Session adaptation and adaptive Session adaptation and adaptive reservationreservation
17
Bottleneck UtilizationBottleneck Utilization
Request blocking probabilityRequest blocking probability
18
Total network revenue ($/min)Total network revenue ($/min)
Total user benefit ($/min)Total user benefit ($/min)
21
Bottleneck utilizationBottleneck utilization
Request blocking probabilityRequest blocking probability
Effect of target utilization levelEffect of target utilization level
22
Effect of Price Adjustment StepEffect of Price Adjustment Step
Request blocking probabilityRequest blocking probability
Bottleneck utilizationBottleneck utilization
23
Effect of Price Adjustment ThresholdEffect of Price Adjustment ThresholdBottleneck utilizationBottleneck utilization
Request blocking probabilityRequest blocking probability
24
Effect of User Demand ElasticityEffect of User Demand ElasticityAverage user bandwidthAverage user bandwidth
Average user chargeAverage user charge
25
Effect of Multiplexing Between Effect of Multiplexing Between User SessionsUser Sessions
Request blocking probabilityRequest blocking probability
Total user benefitTotal user benefit
26
Adaptation by Part of User PopulationAdaptation by Part of User PopulationBottleneck utilizationBottleneck utilization
Request blocking probabilityRequest blocking probability
27
One-time Versus Ongoing Adaptation One-time Versus Ongoing Adaptation Bottleneck utilizationBottleneck utilization
Request blocking probabilityRequest blocking probability
29
RNAP Message Aggregation (cont’d)RNAP Message Aggregation (cont’d)
• Aggregate for senders sharing Aggregate for senders sharing the same destination networkthe same destination network– merged by source domainsmerged by source domains
– split for HRNs at destination net: split for HRNs at destination net:
border router border router
(RNAP-D) NRN (RNAP-C) (RNAP-D) NRN (RNAP-C)
• Two messages:Two messages:
– aggregated-resource message aggregated-resource message
reserves and collects price in the reserves and collects price in the
middle of networkmiddle of network
– original messages sent directly to original messages sent directly to
destination without visiting agents destination without visiting agents
in betweenin between
30
ConclusionsConclusions
• CPA gain over FPCPA gain over FP– Network availability, revenue, Network availability, revenue,
user perceived benefituser perceived benefit
– CPA congestion control is stable CPA congestion control is stable
and effectiveand effective
• Target reservation rate Target reservation rate
(utilization): (utilization):
– too high or too low utilization: too high or too low utilization:
User benefit User benefit
– Too low target rate: demand Too low target rate: demand
fluctuation is highfluctuation is high
– Too high target rate: high Too high target rate: high
blocking rate blocking rate
31
ConclusionsConclusions
• Effect of price scaling factor Effect of price scaling factor – , blocking rate, blocking rate too large under-utilization, too large under-utilization,
large dynamics large dynamics
• Effect of price adjustment Effect of price adjustment
threshold threshold – Too high, no meaningful Too high, no meaningful
adaptationadaptation
32
ConclusionsConclusions
• Demand elasticityDemand elasticity– Bandwidth sharing is proportional Bandwidth sharing is proportional
to user’s willingness to pay to user’s willingness to pay
• User adaptation by some users User adaptation by some users
still results in overall still results in overall
performance improvementperformance improvement