PArallel, Networking and Distributed Applications Research at UVic

1
PArallel, Networking and Distributed Applications
Research at UVic

• Steven Shelford and Jeff Hornsberger
• PANDA Lab
• University of Victoria

2
Agenda

• Optimal Admission Control - Steven
• Sensor Network Routing - Jeff

3
Optimal Admission Control

• Steven Shelford

4
The Problem

• Tech wreck, telecomm spending collapse
• Internet traffic continues to grow
• Too little of it attracts REVENUE! So,
• Got to find some new apps . . .
• Video on demand Blockbuster,
• videoconferencing especially since 9/11,
• Interactive video . . .

5
The Problem

• multimedia applications like
• movies
• high-quality videoconference
• Web/video/audio hybrid forms
• have stringent Quality of Service requirements
  (data rate, latency, BER, etc)
  

6
But . . .

• Current Internet cannot guarantee QoS
• IP datagram based
• Too-late binding of resources to datagrams

7
Optimal Admission Control

• Work at UVic

8
Dave Hudsons IdeaThe High Performance Optical
Overlay

• A second set of networks
• carries IP datagrams
• designed for guaranteed QoS
• I.e. connection-oriented
• augments the current Internet

9
The High Performance Optical Overlay
Admission controller
High Performance Optical Overlay
Client-side content switch
Internet
clients
10
The Admission Controller

• Guarantee QoS for every admitted session
• Admitting to optimize network revenue, while
  considering all link capacity constraints
  
• Mapped to a type of combinatorics problem the
  knapsack problem
  
• Only for small set of interconnected networks

11
Optimal Admission Control

• Future Work scalability, economics computer
  science

12
New Ideas

• Considering optimal admission between a large set
  of networks, owned by different entities
  
• Need new definition of optimal socially optimal
  economic term

13
Network Market Model

• Optimal link pricing by layered market of network
  components

14
Research Focus

• Assume bandwidth exchange per network or
  geographic area
  
• Purchase necessary bandwidth from various
  vendors
  
• Purchase multiple items (link bandwidths) to form
  end-to-end path
  
• Considering optimal admission, where optimal is
  socially optimal
  

15
How Much Functionality?

• Base functionality only, allow firms to offer
  extended functionality
  
• I.e.
• Purchase bandwidth on a path in short time
  intervals (say 5 seconds) if prices go too
  high, lose connection.
  
• Can purchase guaranteed long-term (2 hours for a
  movie) connections through insurance firms
  (i.e. ISPs) which will accept risk of fluctuating
  prices but at a cost

16
Summary

• Optimal admission necessary for new revenue
  producing services over the Internet
  
• Previous works focus on admission control in
  terms of knapsack problem for a small set of
  interconnected network
  
• New work focuses on economics for insight into
  scalable solution

17
Sensor Network Routing

• Jeff Hornsberger

18
Why Sensor Networks?

• Future ubiquitous computing applications

19
Types of Wireless Networks

• Infrastructure-based
• Infrastructure-less

20
What is a Sensor Network?

• Infrastructure-less, resource constrained,
  wireless network.

21
Sensor Networks

• Sensors form self-organizing, multi-hop ad hoc
  network.

Sink
Task Manager
User
Sensor Nodes
Sensor Field
22
Sensor Constraints

• Cost
• Size
• Power
• Processing Capability
• Bandwidth and Range
• Memory

23
Network Characteristics

• Location knowledge
• Network size
• Traffic patterns
• Node failure

24
Research Activities

• Moving data through the network.

25
Sensor Network Research

• Need for targeted research
• Network Layer
• Routing

26
Network Layer Routing

• End-to-end communication
• Scalability
• Sensor field lifetime

27
Routing Strategies

• Location knowledge
• Addressing
• Diversity injection
• Reduced state
• Energy and node conservation

28
Summary

• Basis for ubiquitous computing applications
• Targeted research required
• Routing must scale and prolong sensor field
  lifetime

29
Thanks

• Steven Shelford Admission Control
• Jeff Hornsberger Sensor Networks
• For other Multimedia Research within PANDA see
• http//panda.cs.uvic.ca