Xin Wang

1
Scalable Network Architecture Measurements
for Multicast and Adaptive QoS

• Xin Wang
• Adviser Henning Schulzrinne
• Internet Real -Time Laboratory
• Columbia University
• http//www.cs.columbia.edu/xinwang

2
Scope of this Talk

• Main work
• Resource Negotiation, Pricing, and QoS for
  Adaptive Multimedia Applications
• Other work
• Measurements and Analysis of LDAP Performance
• IP Multicast Fault Recovery in PIM over OSPF

3
Resource Negotiation, Pricing and QoS
for Adaptive Multimedia Applications
4
Outline

• Introduction
• A Resource Negotiation And Pricing
  protocol RNAP
• Pricing models
• User adaptation models
• Test-bed demonstration of Resource Negotiation
  Framework
• Simulation and discussion of Resource Negotiation
  Framework
• Conclusion and future work

Resource Negotiation Framework
5
Is Simple Over-Provisioning Enough?

• Current Internet
• Growth of new IP services and applications with
  different bandwidth and quality of service
  requirements
• Revenue from the traditional connectivity
  services is declining
• New services present opportunities and challenges
  
• Even though average bandwidth utilization is low,
  congestion can happen access links get congested
  frequently
• Wireless bandwidth is even more scarce
• Bandwidth prices are not dropping rapidly
• No intrinsic upper limit on bandwidth use

Option - manage the existing bandwidth better,
with a service model which uses bandwidth
efficiently.
6
More Efficient Service Models

• Quality of Service (QoS)
• Condition the network to provide predictability
  to an application even during high user demand
• Provide multiple levels of services
• Problems signaling to facilitate service
  negotiation pricing scheme to support
  differentiated services
• Application adaptation
• Source rate adaptation based on network
  conditions - congestion control and efficient
  bandwidth utilization
• Problems
• How will adaptive applications work with
  QoS-assured services? How to motivate an
  application to adapt?

7
Design Goals

• Develop an efficient service model which
  combines QoS assurance with user rate adaptation
• Increase service value to the users through
  greater choices over price and quality, reduced
  blocking rate, and expected QoS
• Reduce network provision complexity, improve
  network efficiency and increase revenue to the
  providers allow network operator to create
  different trade-offs between blocking admissions
  and raising congestion prices

8
Existing Work

• Adaptive Internet Multimedia Applications
• Sender-driven, receiver-driven, or
  transcoder-based
• Problems
• Fairness
• non-adaptive applications benefit at the expense
  of adaptive applications
• resource allocation different for different
  adaptation schemes
• TCP friendly adaptation lead to abrupt changes in
  quality
• buffering smoothes the abrupt change at the cost
  of high delay
• No mechanism for rate adaptation in QoS-enhanced
  environment
• No motivation for users to adapt

9
Existing Work (contd)

• Signaling for Resource Reservation
• RSVP, YESSIR, SIBBS (Simple Inter-domain
  Bandwidth Broker Signaling protocol)
• Problems
• Restricted to either sender-driven or
  receiver-driven
• No support for service selection and negotiation
• No support for short-term resource commitment and
  dynamic resource negotiation
• No support for pricing and billing

10
Existing Work (contd)

• Pricing and Billing in the Network
• Total user benefit maximization based on welfare
  theory
• problems
• rely on a centralized optimization process
• assume knowledge of users utility functions
• Pricing for congestion control
• Kelly99
• packets are statistically marked
• control is applied at the packet level
• not scaling, and not clear delay effect on
  theoretical results
• Cocchi93, Lazar97 rely on well-defined source
  model
• Fulp98 restricted to single class, control only
  at service setup time

11
Existing Work (contd)

• Pricing in Multi-service environment
• Qdlyzko99 relies on user behavior, without
  service assurance
• Kumaran99 restricted to special admission
  control algorithm
• Bandwidth Auction
• problems
• signaling bursts, set-up delay, uncertainty of
  connection availability, user response to
  fluctuations in price
• Signaling Support for Pricing and Charging
• Very limited work in this area
• Karsten98
• estimated the cost for user resource requests
• no price quotation
• restricted to IntServ

12
Thesis Contributions

• Propose a Resource Negotiation And Pricing
  protocol RNAP
• Supports short-term resource commitment for
  better response to user demand and network
  conditions, and more efficient resource usage
  allows for service predictability
• Includes mechanisms for price and charge
  collation, auction bids and results distribution
• Designed to be scalable and reliable can be
  embedded in other protocols, or implemented
  independently
• Allows multi-party negotiation
• Develop a demand-sensitive pricing model for
  multiple services
• Services priced to reflect the cost and long-term
  user demand
• Allows for congestion pricing to motivate user
  adaptation
• Develop reference user adaptation models

13
Outline

• Introduction
• A Resource Negotiation And Pricing
  protocol RNAP
• Pricing models
• User adaptation models
• Test-bed demonstration of Resource Negotiation
  Framework
• Simulation and discussion of Resource Negotiation
  Framework
• Conclusion and future work

Resource Negotiation Framework
14
Protocol Architectures Centralized(RNAP-C)
Host Resource Negotiator
RNAP Messages
Network Resource Negotiator
NRN
NRN
NRN
HRN
HRN
Access Domain - A
Edge Router
Access Domain - B
Internal Router
Intra-domain messages
Transit Domain
15
Protocol Architectures Distributed (RNAP-D)
Local Resource Negotiator
RNAP Messages
HRN
LRN
LRN
LRN
LRN
LRN
LRN
LRN
LRN
HRN
LRN
LRN
LRN
Access Domain - A
LRN
LRN
Edge Router
Access Domain - B
Internal Router
Transit Domain
16
RNAP Messages
Query Inquires about available services, prices
Query
Quotation
Quotation Specifies service availability,
accumulates service statistics,
prices
Reserve
Commit
Reserve Requests services and resources,
Modifies earlier requests
Periodic negotiation
Quotation
Commit Confirms the service request at a
specific price or denies it.
Reserve
Commit
Close Tears down negotiation session
Close
Release Releases the resources
Release
17
Outline

• Introduction
• A Resource Negotiation And Pricing
  protocol RNAP
• Pricing models
• User adaptation
• Test-bed demonstration of Resource Negotiation
  Framework
• Simulation and discussion of Resource Negotiation
  Framework
• Conclusion and future work

Resource Negotiation Framework
18
Two Volume-based Pricing Strategies

• Fixed-Price (FP) fixed unit volume price
• During congestion higher blocking rate OR higher
  dropping rate and delay
• Congestion-dependent-Price (CP) FP
  congestion-sensitive price component
• During congestion users have options to maintain
  service by paying more OR reducing sending rate
  OR switching to lower service class
• Overall reduced rate of service blocking, packet
  dropping and delay

19
Proposed Pricing Strategies

• Holding price and charge based on cost of
  blocking other users by holding bandwidth even
  without sending data
• phj ? j (pu j - pu j-1) , chij (n) ph j r
  ij (n)? j
• Usage price and charge maximize the providers
  profit, constrained by resource availability
• max Sl x j (pu1 , pu2 , , puJ ) puj - f(C),
  s.t. r (x (pu2 , pu2 , , puJ )) ? R
  
  
• cuij (n) pu j v ij (n)
• Congestion price and charge drive demand to
  supply level (tatonnement process or auction)

20
Usage Price for Differentiated Service

• Usage price based on cost of class bandwidth
• lower target load (higher QoS) -gt higher per-unit
  bandwidth price
• Parameters
  
• pbasic basic rate for fully used bandwidth
  
• ? j expected load ratio of class j
• xij effective bandwidth consumption of
  application i
• Aj constant elasticity demand parameter
• Price for class j puj pbasic / ? j
• Demand of class j xj ( puj ) Aj / puj
• Effective bandwidth consumption xe j ( puj )
  Aj / ( puj ? j )
• Network maximizes profit
• max Sl (Aj / pu j ) pu j - f (C), puj
  pbasic / ? j , s. t. Sl Aj / ( pu j ? j ) ? C
• Hence pbasic Sl Aj / C , puj Sl Aj /(C? j)

21
Congestion Price

• Tatonnement process
• Congestion charge proportional to excess demand
  relative to target utilization
• pc j (n) min pcj (n-1) ? j (Dj, Sj) x
  (Dj-Sj)/Sj,0 , pmaxj
• M-bid auction Model
• User bids (bandwidth, price) for a number of
  bandwidths, bids obtained by sampling utility
  function.
• Congestion price charge highest rejected bid
  price
• Advantages
• reduces uncertainty since user can express
  multiple preferences
• periodic auctions enable congestion control
• more users served during high demand selects
  lower bandwidth (higher price per unit bandwidth)
  bids of elastic users
• inter-auction admission reduces set-up delay
  
  

22
Outline

• Introduction
• A Resource Negotiation And Pricing
  protocol RNAP
• Pricing models
• User adaptation models
• Test-bed demonstration of Resource Negotiation
  Framework
• Simulation and discussion of Resource Negotiation
  Framework
• Conclusion and future work

Resource Negotiation Framework
23
Rate Adaptation of Multimedia System

• Gain optimal perceptual value of the system based
  on the network conditions and user profile
• Utility function users preference or
  willingness to pay
• Example utility functions
• Utility is a function of bandwidth at fixed QoS
• U (x) U0 ? log (x / xm)
• U0 perceived (opportunity) value at minimum
  bandwidth
• ? sensitivity of the utility to bandwidth
• Function of both bandwidth and QoS
• U (x) U0 ? log (x / xm) - kd d - kl l , for x
  ? xm
• kd sensitivity to delay
• kl sensitivity to loss

24
Rate-Adaptation Models

• Model1 User adaptation under tatonnement process
• Optimize perceived surplus of the multimedia
  system subject to budget and application
  requirements
• U Si Ui (xi (Tspec, Rspec)
• max Sl Ui (xi ) - Ci (xi) , s. t. Sl Ci (xi)
  ? b , xmini ? xi ? xmaxi
• Determine optimal Tspec and Rspec
• With the example utility functions, resource
  request of application i
• Without budget constraint x i ?i / pi
• With budget constraint x i bi / pi, with b
  i b (? i / Sl ? k)
• Model2 User adaptation under M-bid auction
• Adapt rate based on allocated bandwidth/QoS

25
Outline

• Introduction
• A Resource Negotiation And Pricing
  protocol RNAP
• Pricing models
• User adaptation
• Test-bed demonstration of Resource Negotiation
  Framework
• Simulation and discussion of Resource Negotiation
  Framework
• Conclusion and future work

Resource Negotiation Framework
26
Testbed Architecture

• Demonstrate functionality and performance
  improvement
• blocking rate, loss, delay, price stability,
  perceived media quality
• Host
• HRN negotiates for a system
• Host processes (HRN, VIC, RAT) communicate
  through Mbus
• Network
• Router FreeBSD 3.4 ALTQ 2.2, CBQ extended for
  DiffServ
• NRN (1) Process RNAP messages (2) Admission
  control, monitor statistics, compute price (3)
  At edge, dynamically configure the conditioners
  and form charge
• Inter-entity signaling RNAP

VIC
RAT
Mbus
HRN
RNAP
NRN
27
Outline

• Introduction
• A Resource Negotiation And Pricing
  protocol RNAP
• Pricing models
• User adaptation
• Test-bed demonstration of Resource Negotiation
  Framework
• Simulation and discussion of Resource Negotiation
  Framework
• Conclusion and future work

Resource Negotiation Framework
28
Simulation Design

• Performance comparison fixed price policy (FP)
  vs. congestion price based adaptive service (CPA)
  
• loss, delay, blocking rate, user benefit,
  network revenue, stability
• Three groups of experiments effect of traffic
  load, admission control, and load balance between
  classes
• Weighted Round Robin (WRR) scheduler
• Three classes EF, AF, BE
• EF load threshold 40, delay bound 2 ms, loss
  bound 10-6
• AF load threshold 60, delay bound 5 ms, loss
  bound 10-4
• BE load threshold 90,delay bound 100 ms,loss
  bound 10-2
• Sources mix of on-off traffic and Pareto on-off
  traffic

29
Delay and Loss
CPA or admission control maintains the traffic
load at the targeted level, limits the packet
delay and loss
30
Service Price
Combining admission control with CPA results in
better control of price dynamics.
31
Network revenue and user benefit
CPA provides significantly higher revenue and
user benefit FP with admission control has
stable revenue due to blocking FP without
admission control has much lower user benefit
CPA
FP
32
Blocking rate
Coupled with user adaptation, the blocking rate
of CPA is up to 30 times smaller than that of FP.
33
Load Balance Between Classes
Even when a small portion of users (15) select
other service classes, the performance of the
over-loaded class is greatly improved.
Migrating to EF and BE
34
Other Results

• Users with different demand elasticity share
  bandwidth proportional to their willingness to
  pay
• Even a small proportion of adaptive users (e.g
  25) results in a significant performance
  improvement for the entire user population (18
  improvement)
• Performance of CPA further improves as the
  network scales and more connections share the
  resources
• Both M-bid auction and tatonnement process can be
  used to calculate the congestion price auction
  gives higher perceived user benefit and network
  utilization at cost of implementation complexity
  and setup delay

35
Conclusions

• Proposed a dynamic resource negotiation framework
  consisting of A Resource Negotiation And Pricing
  protocol (RNAP) , a rate and QoS adaptation
  model, and a pricing model
• RNAP
• dynamic service negotiation
• support for pricing and charging
• Pricing models
• based on resources consumed by service class and
  long-term user demand
• congestion-sensitive component to motivate user
  demand adaptation
• novel M-bid auction model for congestion pricing
• Performance
• Effectively restricts load to targeted level and
  meet service assurance
• Provide lower blocking rate, higher user
  satisfaction and network revenue
• Admission control and inter-service class
  adaptation give further improvements in blocking
  rate and price stability

36
Future Work

• Interaction of short-term resource negotiation
  with longer-term network provision
• A light-weight resource management protocol
• Cost distribution in QoS-enhanced multicast
  network
• Pricing and service negotiation in the presence
  of alternative data paths or competing networks
• User valuation models for different QoS
• Resource provisioning in wireless environment

37
Measurements and Analysis of
LDAP Performance

38
Problems and Contributions

• Lightweight Directory Access Protocol (LDAP)
  widely used, but little study of performance
• Related work Mindcraft98 treated LDAP server as
  black box, did not study the influence of system
  components
• Thesis contributions
• Developed a benchmark tool to analyze the
  performance of LDAP
• Provided guidelines for the configuration of
  LDAP client and server
• Suggested schemes for LDAP performance
  improvement
• Is the first effort to address LDAP performance
  issues and configuration issues
• LDAP usage context management of network
  resources, pricing policies

39
Experiments

• Experimental setup
• Hardware server- dual Ultra-2 processors, 200
  MHz CPUs, 256 Mb memory Clients- Ultra1, 170 MHz
  CPU, 128 MB memory 10 Mb/s Ethernet
• LDAP server OpenLDAP 1.2, Berkeley DB 2.4.14
• Search filter interface address, and
  corresponding policy object
• Default parameters 10,000 entries, entry size
  488 bytes
• General results
• response latency 8 ms up to 105 requests/second
• Maximum throughput 140 requests/second
• 5 ms processing latency - 36 from backend, 64
  from front end
• Connect time dominates at high load, and limits
  the throughput

40
Scalability of the Performance

• Scaling with Directory Size - determined by
  back-end processing
• In memory operation, 10,000 -gt 50,000 processing
  time increases 60, throughput reduces 21.
• Out-of-memory, 50,000 -gt100,000 processing time
  increases another 87, and throughput reduces
  23.
• Scaling with Entry Size (488 -gt4880 bytes)
• In-memory, mainly increase in front-end
  processing, i.e., time for ASN.1 encoding .
  Processing time increases 8 ms, 88 due to ASN.1
  encoding, and throughput reduces 30.
  
• Out-of-memory, throughput reduces 70, mainly due
  to increased data transfer time.

41
Performance Improvement

• Disabling TCP Nagle algorithm
• Reduces latency about 50 ms
• Entry Caching
• For 10,000 entry directory, caching all entries
  gives 40 improvement in processing time, 25
  improvement in throughput
• CPU
• For in-memory operation, dual processors improve
  performance by 40
• Connection Re-use
• 60 performance gain when connection left open

42
IP Multicast Fault Recovery
in PIM over OSPF

43
Problems and Contributions

• Many IP multicast applications require high
  availability, especially mission-critical
  real-time data
• A lot of work on reliable multicast, but little
  work on multicast fault tolerance
• Study failure recovery in a complete
  architecture IGMP OSPF (unicast) PIM
  (multicast)
• Focus the interplay of underlying protocols the
  interactions of failure recovery, between
  routers, links, WAN and LAN
• Method quantitative analysis simulation over
  OPNET study failure recovery and implementation
  issues on test-bed using Cisco routers

44
General Observations

• Channel recovery time
• Dominated by unicast table re-construction time.
• Protocol control loads
• PIM-DM control load increases proportionally with
  the redundancy factor and decreases inversely
  with the percentage of receivers
• Below certain interval threshold, OSPF load
  increases proportionally as OSPF Hello interval
  decreases
• Neither PIM nor OSPF has high control traffic
  during failure recovery.

45
Suggestions for Enhancement

• Fast recovery from Dedicated Router (DR) failure
• Reduce Hello-Holdtime to detect neighbor failure
  faster
• Backup DR
• IGMP group information caching in all LAN routers
  (reloading group membership information leads to
  minutes delay)
• Fast recovery from last-hop router failure
• DR records the last-hop router address, actively
  recover, instead of waiting for an IGMP report
  to reactivate its oif to the LAN (up to minutes
  delay)
• Use interrupts instead of polling to reduce
  recovering delay

46
Some References

• X. Wang, H. Schulzrinne, Auction or Tatonnement
  - Finding Congestion Prices for Adaptive
  Applications, submitted.
• X. Wang, H. Schulzrinne, Pricing Network
  Resources for Adaptive Applications in a
  Differentiated Services Network, In Proceeding
  of INFOCOM'2001, April 22-26, Anchorage,
  Alaska.
• X. Wang, H. Schulzrinne, An Integrated Resource
  Negotiation, Pricing, and QoS Adaptation
  Framework for Multimedia Applications, IEEE
  JSAC, vol. 18, 2000. Special Issue on Internet
  QoS.
• X. Wang, H. Schulzrinne, Comparison of Adaptive
  Internet Multimedia Applications, IEICE
  Transactions on Communications, Vol. E82-B, No.
  6, pp. 806--818, June 1999.
• X. Wang, H. Schulzrinne, D. Kandlur, D. Verma,
  Measurement and Analysis of LDAP Performance,
  International Conference on Measurement and
  Modeling of Computer Systems (ACM
  SIGMETRICS'2000).
• X. Wang, H. Schulzrinne, C. Yu, P. Stirpe, W. Wu,
  IP Multicast Fault Recovery in PIM over OSPF,
  In 8th International Conference on Network
  Protocols (ICNP'00), 2000. Also appears at ACM
  SIGMETRICS2000 as short paper.

47
Questions and Answers
Thanks !
48
Message Aggregation (RNAP-D)
Turn on router alert
Edge Routers
Sink-tree-based aggregation
49
Message Aggregation (RNAP-D)
Turn off router alert
Sink-tree-based aggregation
50
Block Negotiation (Network-Network)
Aggregated resources are added/removed in large
blocks to minimize negotiation overhead and
reduce network dynamics
Bandwidth
time
51
Example of M-bid Auction

• Total capacity 70, congestion price is 2

Bid Bandwidth
Bid Selection
Bid Price
Bidder
1
10
5
2
10
4
1
15
4
3
20
3.5
2
25
3
Cutoff
2
30
3
Congestion Price
52
Rate Adaptation of Multimedia System

• Gain optimal perceptual value of the system based
  on the network conditions and user profile
• Utility function users preference or
  willingness to pay

Cost
U1
U2
Utility/cost/budget
U3
Budget
Bandwidth
53
Simulation Architecture
Topology 1 (60 users)
Topology 2 (360 users)