Theories and Frameworks for Ubiquitous Computing Alan Dearle School of Computer Science University of St Andrews

1
Theories and Frameworks forUbiquitous
ComputingAlan DearleSchool of Computer
ScienceUniversity of St Andrews
2
The Gloss Vision

• In the Gloss project we set out to provide an
  infrastructure able to offer
• global,
• customised,
• contextual,
• pervasive and
• adaptive
• assistance to mobile users

3
A simple scenario

• User Bob likes ice cream, but only when the
  weather is hot and when he has spare time to eat
  it
• It is 20ºC in South Street at 16.30 on 25/6/2003
• Bob is on holiday in St Andrews from 20/6/2003 to
  27/6/2003
• Bob is Scottish
• Bob is in North Street at 16.45 on 25/6/2003
• Bob is on foot on 25/6/2003
• Janettas in Market Street sells ice cream, and
  is open between 9.00 and 17.00
• Bob knows Anna
• Anna is at coordinate 56.3397, -2.80753 at 16.15
  on 25/6/2003

4
The result..

• Service suggests to both Bob and Anna that they
  might wish to meet for an ice cream at Janettas
  at 16.55

5
This looks easy but isnt

• Performing correlation of information items
  requires the expression and detection of spatial,
  temporal and logical relationships
• It requires the provision of a generic global
  event service capable delivering events to
  appropriate locations at appropriate times
• Engineering of the matching computation is
  difficult
• partitioning of computation
• deployment of components
• dynamic incorporation and release of resources
• caching and replication for speed reliability
• adaptation to changing patterns of use
• Huge privacy issues (not discussed here)

6
Metaphors

• We developed a number of metaphors to help
  develop services, to inform design decisions and
  to understand the problem space
• Trails
• Radar
• Hearsay
• In Gloss we used these metaphors to describe a
  large number of common scenarios
• Can rephrase the engineering problem in terms of
  these metaphors

7
Trails

• Observational Trail
• Observations of people or artefacts moving
  through space and time
• Intentional Trail
• An unordered set of locations, regions or
  coordinates, with associated information, linked
  by theme
• Archetypal Trail -
• A route between two Wheres

8
Radar

• The Radar metaphor is concerned with the
  detection of artefacts andpeople of interest in
  some geographic vicinity
• It gives the user an overview beyond the
  immediate environment.
• Different forms exist - active and passive
• May incorporate an event horizon

9
Hearsay

• Intuitively Hearsay provides the means for the
  user to leave and obtain information specific to
  particular geo-spatial regions, circumstances and
  personal preferences
• Two services are required to implement hearsay
  placement and delivery

The equivalent to leaving Post-It notes for users
with a particular profile in a particular location
10
Architectural Requirements

• Three architectures are required to implement
  these services
• Local architectures to permit clients and servers
  to be constructed
• A global architecture to mediate the global flow
  of information between users, artefacts and
  service provision points
• A distributed matching engine capable of matching
  events and outputting new events

11
Local Architectures

• Design goals
• (try to) abstract over particularimplementation
  technologies
• Make componentsindependent of each other
• Allow components to beassembled into
  applications
• An assembly may be createdfrom a collection
  ofcomponents linked via pipesand busses
• Alternatively may be constructed using Web
  Services Middleware such as RAFDA

12
Global Architecture

• We need a scalable global architecture on which
  to deploy services and capable of routing
  messages to appropriate location(s)
• We are experimenting with a hierarchy of peers
  which partition the world into non-overlapping
  regions
• This fits well with the region-transition
  hypothesis the hierarchy gives a nested set of
  locations in which a profile may be stored
• Different cache policies at different levels in
  the hierarchy is likely to be beneficial

13
Example - Hearsay delivery
One technology mix.. demonstrated at Ubinet
Gothenburg
14
Global Architecture
15
Real P2P Spanning tree
16
Global Matching

• The problem
• Assimilating and filtering information from
  various sources and determining relevant matches
• Correlating data against some distributed rule
  base which will determine a meaningful action
• Very high volume of globally distributed items of
  information, distilling them down into a
  relatively small volume of meaningful events

17
Engineering Hurdles What we are doing

• Investigating P2P architectures to distribute
  events - Siena, Chord and Plaxton
• Investigated Active Middleware Amit IBM Haifa
• Adopted an active pipeapproach Keller et al
  2001
• Investigating component deployment mechanisms

18
Component Deployment Mechanisms

• Require mechanisms to deploy and evolve pipeline
  components
• Pipeline components deployed as code bundles that
  execute within security domain
• Infrastructure based on Cingal code-push
  technology
• Flexibility in initial deployment and in later
  incremental evolution
• Use P2P routing to leverage this deployment using
  Rafda and Asa components
• Use constraint languages to specify legal
  deployments

19
Research Challenges

• Provision of a generic global event service
• delivering events to appropriate locations at
  appropriate times
• Engineering of the matching computation
• partitioning of computation
• deployment of components
• dynamic incorporation and release of resources
• caching and replication for speed reliability
• techniques for describing and binding to data and
  events
• adaptation to changing patterns of use
• Ability to incorporate new devices and
  technologies incrementally