Best-Effort Multimedia Networking Outline

1

• Scalable Network Architectures
• for Providing
• Per-flow Service Guarantees
• Jasleen Kaur
• Department of Computer Science
• University of North Carolina at Chapel Hill

2
The trend richer network services

• Basic Internet service providing is commoditized
• Last decade network connectivity
• Next decade value-added services
• Value-added services
• Quality of Service, Virtual Private Networks,
• Intrusion detection, Transcoding services

Focus providing QoS guarantees in networks
3
The opportunity QoS

• New applications with stringent timeliness
  requirements
• Live and on-demand video streaming, real-time
  stock quote
• VPNs for mission-critical enterprise applications
• Requirements

• Delay guarantees upper bound on network delay
• Throughput guarantees sustained throughput even
  at short time-scales
• Fairness guarantees throughput in proportion to
  reserved rate

Need to provide per-flow network service
guarantees
4
The challenge growth

• Link capacities are increasing rapidly (double
  every year)
• Less time available to routers for per-packet
  processing

Capacity Per-packet Time
100 Mbps Ethernet 38 ?s
2.45 Gbps (OC48) 1.5 ?s
9.6 Gbps (OC192) 0.38 ?s

• Internet traffic demands are increasing at
  similar rate

• Requirements
• Minimize of instructions, memory accesses,
  amount of memory
• Utilize resources efficiently

Networks need to be scalable and efficient
5
Requirements summary

• A network architecture should
• Provide per-flow guarantees on delay, throughput,
  fairness
• Scale to high capacity links
• Use efficiently available resources

Design network architectures that meet these
requirements
6
Outline

• State of the art
• Research directions and methodology
• Core-stateless Guaranteed Services networks
• Scalability evaluation
• Summary
• Current research directions

7
Network model
8
State of the art

• FIFO networks
• Are simple and scalable
• - Do not provide service guarantees in
  presence of bursty traffic

• Integrated Services (IntServ) networks
  Shenker95
• Provide per-flow guarantees use
  sophisticated scheduling algorithms
• - Do not scale require per-flow state and
  packet classification

• Differentiated Services (DiffServ) networks
  Nichols97
• Are scalable only per-aggregate processing
  in core routers
• - Do not provide per-flow guarantees within an
  aggregate

Architecture Per-flow Guarantees Scalability Efficiency
FIFO X X
DiffServ X X
IntServ X X
9
Two research directions

• Can scalable mechanisms be added to enable FIFO
  networks to provide per-flow service guarantees?
• Can complexity of IntServ mechanisms be
  eliminated, while retaining per-flow guarantees?

• Performance of FIFO networks with CBR
  traffic-shaping NOSSDAV-99
• Analytical model heavy-tails at high
  utilization in large-scale networks
• Simulations heavy-tails even at moderate
  utilization and small networks

Network architectures that provide per-flow
service guarantees without maintaining or using
per-flow state in core routers
10
Core-stateless networks

• Core routers do not maintain per-flow state
• Scalable no state maintenance or classification
  complexity
• Edge routers maintain state
• Scalable small number of flows and low-speed
  links

Edge Routers
Core Routers
11
Core-stateless schemes
Type of service guarantees in core-stateless
schemes
Statistical
Deterministic

• CSFQ Stoica98, RFQ Cao00,
• CHOKe Pan00, TUF Clerget01
• Approximate fairness over long time-scales
• No guarantees for short-lived flows

• CJVC Stoica99
• End-to-end delay guarantees
• Non work-conserving

12
Research methodology

• Careful blend of theory and practice

• Theory
• Understand end-to-end guarantees in core-stateful
  networks
• Design core-stateless networks to provide similar
  guarantees

• First tight lower bound on end-to-end fairness

Exactly same delay guarantees Throughput
guarantees within an additive constant Fairness
guarantees even better

• Practice
• Design, implement and evaluate
• Scalability of edge and core routers
• Feasibility of deploying the core-stateless
  network

13
Delay guarantees are fundamental

• Theorem 1 (throughput ? delay)
• A network that provides throughput guarantees
  also provides delay guarantees

Theorem 2 (fairness ? throughput) A network
that provides fairness guarantees also provides
throughput guarantees
A network that does not provide delay
guarantees, can not provide throughput or
fairness guarantees
14
Guaranteed Rate (GR) scheduling algorithms

• GR Algorithms
• Class of algorithms that provide delay guarantees
  to flows
• Basic operation
• Reserve a rate for each flow
• Associate with packet k, a Guaranteed Rate Clock
  GRC(k) value
• GRC(k) Transmission deadline for packet based on
  reserved rate
• Scheduling algorithm belongs to class GR if it
  guarantees transmission of packet k by GRC(k) ?
• Examples
• Virtual Clock, Delay-EDD, SCFQ, SFQ, WF2Q,

15
Virtual Clock need for per-flow state

• Assign a transmission deadline (VC) to packet k
• EAT(k) max VC(k-1), AT(k)
• VC(k) EAT(k) lk/r
• Transmit packets in increasing order of their VC
  values
• If ?flow r ? C, packet gets transmitted by VC(k)
  lmax/C
• End-to-end delay bound f(upper bound on VC(k)
  at last node)

Delay bound f(upper bound on transmission
deadline)
Transmission deadline of packet k f(state of
packet k-1) ? Need to maintain state of previous
packet!
How to compute deadlines without maintaining
state?
16
Key insight

• Ingress router does maintain per-flow state
• ? can compute upper bounds on deadlines for all
  nodes

• Using upper bounds on deadlines results in same
  network delay guarantee

17
Core-stateless Guaranteed Rate networks

• Ingress router maintains per-flow state
• Computes and encodes deadlines for all nodes
• Core routers do not maintain per-flow state
• Use deadline encoded by ingress router

1
2
Core routers
Ingress router
18
CSGR properties
Theorem End-to-end delay guarantee of a CSGR
network is same as corresponding GR network

• Salient features
• Methodology for deriving core-stateless version
  of any GR network
• Leads to design of work-conserving core-stateless
  networks
• Core-stateless Delay-EDD decouples delay and
  rate guarantees
• Same bound on end-to-end delay as core-stateful
  version
• Simple computations
• Caveat
• Do not preserve short time-scale throughput or
  fairness guarantees
• Flows that use idle capacity to send at more
  than their reserved rate accumulate debit and
  may be penalized in the future !

19
CSGS networks properties

• CSGR Infocom-01 Delay
• Provide exactly same delay guarantees as
  core-stateful networks
• CSGT Infocom-03 Throughput
• Provide throughput guarantees within an additive
  constant of core-stateful networks
• First work-conserving core-stateless network that
  provides deterministic throughput guarantees
• CSGF IWQoS-03 Fairness
• Provide better fairness guarantees than
  core-stateful networks
• First work-conserving core-stateless network that
  provides deterministic fairness guarantees

20
Research methodology

• Careful blend of theory and practice

• Theory
• Understand end-to-end guarantees in core-stateful
  networks
• Design core-stateless networks to provide similar
  guarantees

• First tight lower bound on end-to-end fairness

Exactly same delay guarantees Throughput
guarantees within an additive constant Fairness
guarantees even better

• Practice
• Design, implement and evaluate
• Scalability of edge and core routers
• Feasibility of deploying the core-stateless
  network

21
Scalability evaluation of network architectures

• Constraints in high-speed routers
• Time Per-packet processing time budget is
  limited
• Space Total fast-path memory is limited
• Key question
• What are the performance limits of routers in
  different network architectures?
• Specific values depend on router platform !

Our Approach Implement a CSGS, FIFO, and
IntServ router on common platform and measure
relative performance
22
Router throughput in different architectures
Source routing core-stateless architecture ? A
network architecture that provides end-to-end
per-flow service guarantees with scalability
close to conventional IP routers
23
Summary

• Goal design network architectures that provide
  per-flow guarantees, are scalable, and efficient
• FIFO inadequate if premium traffic occupies a
  large fraction of capacity NOSSDAV-99
• Core-stateless networks theory
• First end-to-end fairness analysis of fair
  queuing networks RTSS-02
• Design of core-stateless networks
• Exactly same delay guarantees
  Infocom-01
• Throughput guarantees within a constant
  Infocom-03
• Fairness guarantees even better
  IWQoS-03
• Core-stateless networks practice
• Routers in core-stateless networks, with source
  routing, have performance similar to conventional
  IP routers

24
Some challenges and open questions

• CSGS networks still require modifications to all
  routers
• Is it possible to provide end-to-end service
  guarantees using mechanisms instantiated only at
  the edges of a network?
• Zhang-Sigcomm02 Throughput of many TCP flows
  is limited due to default parameter settings !
• How suitable for todays Internet are
  traditional end-host mechanisms for flow control?
• Does congestion occur at all? If so, where does
  it occur?
• At end-hosts? At the edge? At the core?

25
Variability in TCP round-trip times

• Max, median, and min RTTs may differ by several
  orders of magnitude within individual TCP
  connections !!

26
Current research directions

• Detecting congestion
• Where does congestion occur?
• What mechanisms help detect it quickly and
  non-intrusively?
• How to design a large-scale, distributed
  congestion-monitoring infrastructure?
• Designing edge-based services
• Designing end-host flow control mechanisms

• Efficacy of overlay-based alternate path routing
• Availability of parallel bandwidth

• Does the single-bottleneck assumption hold?
• Does traditional flow control work well in high
  bandwidth networks?

27
More details being made available at

• URL http//www.cs.unc.edu/jasleen/
• Email jasleen_at_cs.unc.edu