Scalable Conferencing

1
Scalable Conferencing
Ahmed Helmy University of Southern
California/ISI ahelmy_at_usc.edu http//catarina.usc.
edu/ahelmy
2
Outline

• Enabling interactive multimedia conferencing
• A new model for network protocols
• Multicast Routing
• Motivation and merits
• Cutting-edge technologies
• Reservation Protocols
• Designing multimedia networks using virtual
  environments (simulation and animation)

3
Multimedia Protocol Support
Telnet, FTP, http
Unicast Application
TCP
Reliable unicast
Application
Unicast Routing
DV (RIP), OSPF
Transport
Network
Data-Link
Audio, video, wb
Multicast Application
Physical
SRM, RTP
Multicast Transport
RSVP
Multicast Routing
DVMRP, PIM
Unicast Routing
QoS Support
WFQ, prio Q
4
Unicast vs. Multicast
S
S
R1
R1
R2
R2
R3
R3
R4
R4
Multiple unicasts
Multicast
5
Dense vs. Sparse Mode Multicast
S
R1
R2
R3
R4
Dense-Mode Multicast
6
Dense vs. Sparse Mode Multicast
S
R1
Root
R2
R3
R4
Sparse-Mode Multicast
7
Standard Multicast Protocols

• Distance Vector Multicast Routing Protocol
  (DVMRP)
• operates over the MBONE, but has limitations
• routing table explosion
• dense mode broadcast over the Internet
• Multicast Open Shortest Path First (MOSPF)
• limited to OSPF unicast domains
• Core-Based Tree (CBT) - uses sparse mode
• Protocol Independent Multicast (PIM)catarina.usc.
  edu/pim
• combines dense and sparse capabilities
• avoids routing table explosion
• Internet deployed over 4 years by Cisco

8
ReSerVation Protocol (RSVP)

• RSVP soft state setup protocol,
  receiver-initiated reservation
• Uses aggregation to reduce messages and state, to
  improve scalability
• Reservations controlled by administrative
  policies
• www.isi.edu/rsvp

9
RSVP Paradigm
Application
Policy Control
RSVP Daemon
Admission Control
Packet Scheduler
Packet Classifier
Data
10
Virtual Inter-Network Testbed (VINT)
RSVP
complexity (multi-protocol interaction)
Emulation Interface
scale
detail system issue

• Composable modular multi-protocol simulations
• Abstraction techniques combinable with detailed
  modules
• Graphical visualization of simulation and
  empirical data
• Methods and tools for systematic design
  evaluation (scenario generator)
• catarina.usc.edu/vint

11
Design Testing Framework
12
Problem Dimensions

• A Test Scenario may include
• Routed Topology
• Host Events
• Failures and network dynamics

Failures
Topology
Host Events
13
Scenario Generation
Complete/ formal
random
systematic

• Topology Generation
• random, transit-stub, N-level hierarchy (e.g.
  GT-ITM)
• Agent (protocol/app) generation
• spacial distribution (e.g. clustering,
  sparseness)
• temporal distribution
• traffic models, membership dynamics, connection
  start/end
• Fault Generation
• - loss models, crashes, network failures

FSM search
14
Simulation Framework
15
Future Work

• Automate Scenario and Topology Generation using
  search techniques
• Forward Search
• Fault-oriented backward search
• Achieve better completeness and coverage
• Extend the framework for performance analysis of
  end-to-end multicast protocols

16
Thank You Ahmed Helmy ahelmy_at_usc.edu http//catar
ina.usc.edu/ahelmy