Overview of Pervasive Computing Systems: Smart Spaces and Operational Issues

1
Overview of Pervasive Computing Systems Smart
Spaces and Operational Issues

• Volatile Systems
• Smart Spaces and Context-Awareness
• Association and Directory Services
• Sensing and In-network processing
• How are the functions supported and what are the
  system operation problems?

2
Pervasive Computing Systems

• You are asked to design a pervasive computing
  application, i.e., the aware home system
• How to start??
• What are the components?
• What is the system architecture?
• What are the networks?
• What are the devices?
• How to collect and maintain the system status?
• What are the performance objectives in addition
  to providing the required functions of the
  system?
• How to process the sensor data/user queries using
  limited resources?
• You need to know the characteristics of the
  networks, devices, development tools, and the
  efficient algorithms and techniques for
  processing the data items and information in the
  system to the requests and queries from the users

3
Development of Pervasive Applications

• Pervasive computing systems are distributed
  systems
• What make a pervasive application different from
  a conventional distributed system?
• Application characteristics and requirements
  real-time, proactive, dynamic, etc.
• Network heterogeneous network, unstable and
  varying quality of services, dynamic workload
• Devices limited resources, sized, processing
  power and energy supply
• Backend server is similar to many distributed
  systems
• Needs to maintain a database for maintaining
  system status and location databases
• The database may be distributed for maintaining
  the information of different locations
• Since the database system may need to handle
  real-time data, it may be a real-time database
  system (main-memory and priority-cognitive
  scheduling)
• Development Tools
• Depends on client devices (TinyOS, WinCE, J2ME..)

4
Volatile Systems

• The set of users, hardware, and software in a
  pervasive computing system is highly dynamic and
  changes unpredictably
• Why volatile? (systems and applications)
• When a user carrying a sensor device enters a
  room (smart space), the sensors in the room
  should identify the user (a new object)
• How to identify the new object? Detection of the
  presence of a new object and getting the ID
  (universal identifier) carrying by the new
  object. Intruder or owner? Then
• Forms of volatility (system issues and more than
  users)
• Failures of devices and communication links
• Connection and topology (1 to 1, 1 to many and
  many to many)
• The connection may be intermittent (why?)
• Changes in performance of connection such as
  bandwidth and connection points
• The creation and destruction of an association
  (more than network) logical connection
  relationships between software components
  resident on the devices (i.e., by bluetooth or
  infra red)

5
A room responding to a user wearing an active
badge
2. Infrared sensor detects users ID
Users ID
1. User enters room wearing
3. Display responds
Hello Roy
an active badge
to user
Infrared
Fr. Dollimore
Active badge is a radio emitting identifier
(emitting a unique ID) What to be displayed?
(Depending on who is the user) What need to be
performed when the user leaves the room?
Note that you may build a smart space in the wild
and open area. How to define the scope of the
space?
6
Smart Spaces and Objects

• Smart space
• Dynamic environment and adaptive services
• Smart space is any physical place (environment)
  with embedded services (enter -gt detection -gt
  services delivery)
• Capture of new objects and maintenance of the
  space (including the objects) status
• Services are provided to the objects in the space
• The objects in space may also provide services to
  other objects
• Smart objects
• The object is smart (you do not need to submit
  commands or instructions)
• Spontaneously join the space
• Proactive services
• Understand what you need when you are inside the
  smart space
• How? Based on some pre-defined rules. Detection
  -gt Event occurred -gt generation an action
  (response)
• Why smart spaces and smart objects are feasible
  nowadays?
• Small size, processing capability, memory and
  energy issues

7
Context Awareness of Smart Objects/Services

• The objects have their own functions
• Different contexts generate different actions
  (services)
• Context of an entity/object is an aspect of its
  physical (logical) circumstances (i.e., location,
  time, status, temperature, identity, etc.) of
  relevance to system behavior
• It could be the presence of an entity or complex
  conditions
• Different actions for different situations
• They may have sensors and talk with others
• Location/situation/context aware
• Can remember pertinent events
• They have a memory and state
• State transition
• Evaluation problem
• Each state change triggers a rule evaluation

8
Context Awareness of Smart Objects/Services

• Using sensors for determining contextual values
• Location, velocity orientation navigation, GPS
  etc
• Ambient conditions light intensity, sound
  intensity, etc.
• Presence RFID, infra-red, active badge
• Are responsive/proactive
• Communicate with environment
• Networked with other smart objects spontaneously
• What are the main differences and similarities
  between a computer and a smart object?

9
Adaptations

• Context aware adaptation (services)
• Devices are heterogeneous in a pervasive
  computing system, i.e., have different
  capabilities and characteristics
• Display of web pages/videos
• Transcoding may need to be performed on the fly
  (by the proxy or by the source before
  transmission)
• Adaptation to resource allocation (performance)
• The number of devices and workload in a smart
  space could be highly dynamic
• To maintain the quality of services (proactive
  with real-time responses), it needs to adapt to
  the changing workload
• How admission control with reservation, and
  prioritization (scheduling)

10
Smart Christmas Tree

• You are asked to build a smart Christmas tree
• Installed with sensors to detect whether any
  object is moving towards it (sound sensor)
• Once a new object is detected, the lamps on the
  tree will be turned on to show the welcome sign
  for 10 seconds
• When a loud noise is detected, it will be
  disabled for 10 min
• If it has been activated continuously for 10
  times within 5 min, it will be disabled for 5 min
• How to implement the system
• Install multiple Mica Motes on a Christmas tree
  and they can communicate with each other using
  radio signals
• Your application in each Mica Mote captures
  sensor data generated from the sound sensor
  periodically (every second)
• If the sound level (sensor data) is higher than a
  threshold value, turn them on
• A simple database is maintained at each Mica Mote
  to record its current status and also the last
  received sensor data
• Problems the data from the sensors may contain
  errors

11
Association (Group Management)

• Multiple objects
• Multiple objects co-exist in the same smart
  space. Some of them are service providers and the
  others are users
• Multiple objects co-exist in the same smart space
  for improving the correctness of detected sensor
  data (i.e., multiple temperature sensors are
  installed in the same room. What is the room
  temperature? Do they give the same reading?
• Association at least one of a given pair of
  components communicates with another within a
  smart space (cooperatively perform a task
  (provide services))
• After association gt Interoperation the
  interaction during association
• Association is spontaneous (without user
  intervention)
• Network bootrapping
• Communication takes place over a local network
  within the smart space
• The device acquires an address (ID and a name) on
  the local network
• Who determine the assignment and manage the
  network

12
Association Problem
Maintenance
N6
N7
N6
N7
N1
N1
N2
N3
N2
N3
N4
N4
N5
N5
time t1
time t2
good link weak link
13
Connection and Communication Problem

• Multiple objects within a smart space
• How are they connected?
• What are the communication links?
• A network with dynamic configuration
• Mobile Ad hoc network (MANET)
• Limited bandwidth
• Probability of communication collision
• Limited energy
• Real-time control (communication delay and
  response delay)

14
Routing Problem

• Highly dynamic network topology
• Device mobility plus varying channel quality
• Separation and merging of networks possible
• Asymmetric connections possible

N6
N7
N1
N6
N3
N7
N2
N1
N2
N3
N4
N5
N4
N5
time t2
time t1
good link weak link
15
Association Problems

• Scale and scope
• Scale the number of objects within a space. More
  objects higher communication cost
• Scope space boundary
• I.e. consider the assignment of mobile stations
  to a base station in a cellular network
• Physical Vs. logical space
• The object physically locates within a space may
  be assigned to another space that is far from its
  current location
• How are they connected? Using broadcast messages
• Consider the problem of broadcasting a message to
  all the members within a group (room)
• In a smart space, disconnection is common and the
  network bandwidth and quality is varying

16
Smart Space 1
Smart Space 2
Smart Space 3
17
Physical Vs Logical Smart Space
integration of heterogeneous fixed andmobile
networks with varyingtransmission characteristics
regional
vertical handoff
metropolitan area
Fr. Schiller
horizontal handoff
campus-based
in-car, in-house, personal area
Why do we need to perform a handoff operation for
a mobile (or even a stationary) user?
18
Centralized Vs. Distributed Management

• Centralized approach use a powerful server to
  manage the space status and connection
  information
• Distributed approach multiple devices (service
  providers) manage the information
• Comparisons
• Problems in distributed computing
• Perform operations at device level because of
  limited bandwidth
• Due to the dynamic properties of the smart space
  and objects, a lot of updates are needed to be
  generated
• A distributed approach can make the management of
  objects to be localized and adaptive to the
  changing systems status (in networking
  processing). But, the communication overhead
  could be very heavy
• A hierarchical approach multiple levels with
  different types of coordinators may be used

19
Centralized Approach
20
Distributed Approach
Peer-to-peer and clustering problems
21
Discovery Services

• When a client enters a smart space,
• The system needs to update its location (and
  system status)
• The object needs to determine how to join it
  (registration) and what are the services provided
  in the space
• I.e., A client enters a hotel carrying a notebook
  and wants to print a document
• A discovery service is a directory service in
  which services in a smart space are registered
  and looked up by their attributes
• Types of services device directory and service
  directory
• Device directory records the names and addresses
  of co-present devices
• Select one of the devices for connection and to
  require for services
• Service directory records what are the services
  provided in the smart space
• Submit a request (query) for a particular type of
  service
• The smart space will select one of the devices to
  provide the service to the requesting client

22
Discovery Services

• Directory query operations
• Registration and de-registration of services
• Look-up for services
• Directory discovery problems
• Directory data required by a client are dynamic
  and generated at run-time (as a function of the
  context of the client)
• May be no infrastructure (distributed Vs.
  centralized) to hold the directory server
• The services registered in a directory are
  dynamic too
• The new object may be a service provider
• The protocols used for accessing the directory
  need to be energy sensitive
• Note that a service may disappear spontaneously.
  Why?
• Using leasing a lease is temporary allocation of
  a service by a server to a client. It can be
  renewed by a further request (refresh) from the
  client before the deadline expires

23
The interface to a discovery service
Fr. Dollimore
24
Directory Server Vs. Serverless

• For systems with a fixed infrastructure and the
  directory server is a powerful robust machine
• Directory server maintains a set of description
  of services
• The client issues a multicast request to locate
  the directory server
• The directory server responds with its unicast
  address
• They communicate point-to-point with each other
• If no pre-defined powerful directory server, the
  directory server may be elected from the group of
  devices within the space
• Problems disappear of the elected server
  (re-election). The cost for re-election depends
  on the degree of volatility of the system

25
Directory Server Vs. Serverless

• No fixed infrastructure and no powerful machine
  to be acted as a server
• The participating devices collaborate with each
  other to implement the service directory. How?
• Event synchronization problem all changes have
  to be propagate to all the member before any
  change can be updated (S1-gtS2-gtS3)
• Push model services multicast their description
  regularly. Clients listen for the multicasts and
  run their queries against them
• Pull model clients multicast their queries.
  Devices providing services run the queries
  against their descriptions, and only response
  with any description that match. Clients repeat
  their queries periodically if there is no
  response

26
Push Vs. Pull

• Consideration energy and bandwidth consumption
• Every time a device issues a multicast message,
  bandwidth is consumed and all listening clients
  expend energy receiving the message
• In a pure push model, devices need to multicast
  their services even there is no client
• In a pure pull model, a client can discover
  services as soon as it appears. But, the client
  may receive multiple responses from different
  devices

27
Jinis Discovery System

• Java based system for mobile and pervasive
  computing systems
• Components lookup services (discovery services),
  Jini services and Jini clients
• A Jini service provides services
• The lookup stores services
• Jini clients request services
• Lookup service allows Jini services to register
  the services they offer
• A Jini service may be registered with one or more
  lookup services
• Jini clients request services that match their
  requirements
• If a match is found, the Jini client downloads an
  object that provides access to the service from
  the lookup service

28
Jinis Discovery System

• When a Jini client or service starts up, it sends
  a request to a well-known IP multicast address
• Any lookup service that receives the request
  sends its address enabling the requester to
  perform a remote invocation to look up or
  register a service with it
• The client requires a lookup service in the
  finance group so it multicasts a request with
  that group name
• Only one lookup is bound to the group name and
  that service responds including its address
• The client communicates directly using RMI to
  locate all services of type printing
• Only one printing service has registered with the
  lookup service
• The client then uses the printing directly
• Problem subnet area Vs. smart space (a room)
• Possible solutions human input space ID, sensors
  to read the space ID

29
Service discovery in Jini
1. finance lookup service
admin
Printing
service
Client
admin
Client
Lookup
service
Network
2. Here I am .....
4. Use printing
service
admin, finance
Lookup
3. Request printing
service
Printing
Corporate
infoservice
service
finance
Fr. Dollimore
30
In-network Processing

• In-network processing process a query inside the
  network (close to the place where its required
  data are generated
• Physically distributed source of data and
  distributed processing
• Processing within the network (device level),
  i.e., close to the source for generating the data
  that are required by the request
• I.e. You submit a query to ask the traffic
  conditions in Nathan Road. It is assumed that we
  have traffic sensors on the roads
• Multiple data access The query may be divided
  into sub-queries to be processed at different
  nodes (distributed query processing) if it
  requires access to multiple data objects
• How to divide the query into sub-queries? Based
  on the locations of their required data
• It is assumed that the sources for generating the
  data are known
• A query may be represented as tree structure

31
In-network Processing

• If the sources are unknown
• Direct diffusion The query may be specified with
  its interests (an attribute-value pair). Then,
  the query is diffused into the network to search
  for the matching to select the nodes
• The sink propagates interests to its neighboring
  nodes
• Any node that receives an interest store a record
  of it along with information that needed to pass
  back to the sink node. Then propagate the
  interest to its neighboring nodes
• A source is the node that match an interest.
  There may be several source nodes
• Problems redundant propagation and may flood the
  whole system
• Solution
• Defining a gradient (a direction/value pair) to
  control the flow. If the value is below the
  threshold, the flow will not follow the direction
  (What is the assumption?)
• Adding a filter in the intermediate nodes to
  eliminate duplication

32
Directed diffusion
sink
sink
sink
source
source
source
source
source
source
C. Data delivery
B. Gradients set up
A. Interest propagation
Fr. Schiller
33
Sensor Data Collection
34
Sensing and Sensor Data Collection

• Multiple levels and integration
• Abstraction of sensor data
• Temperature reading from a sensor is the
  temperature of a room?
• Different sensors in the room may have different
  readings
• Presence of an intruder (a new physical and
  moving object and unknown ID)
• Errors
• Some of the sensors may have a very different
  reading from the others
• Context is dynamic
• The readings from the sensors may change quickly
• Update and stream data processing problem

35
Examples

• Context widget
• Present an abstraction of some type of context
  attribute while hiding the complexity of the
  actual sensors used
• Constructed from distributed components
• Generators acquire raw data from sensors such as
  from floor pressure sensors and provide that data
  to widgets
• Widgets use the services interpreters, which
  abstract contextual attributes from the generated
  low-level data to high level data such as the
  identity of a person
• The server provides further levels of
  abstractions by collecting, storing and
  interpreting contextual attributes from other
  widgets
• Active Badge Sensors
• For detecting presence of an identity carrying an
  active badge (location management system)
• Active badge sensors are deployed at fixed
  locations throughout the whole building
• For a telephone receptionist, when receiving an
  incoming call, searching the location management
  system, it can identify the location of the
  called party

36
A PersonFinder widget constructed using
IdentityPresence widgets
A PersonFinder widget for a building is
constructed from the IndetifyPresence widgets for
each room using footstep interpretation from
floor pressure readings or from face recognition
from vide capture
37
Reference

• Distributed Systems Concepts and Design by
  Dollimore, Ch16 (except 16.3, 16.5)