Olympus: A HighLevel Programming Model for Pervasive Computing environments

1
Olympus A High-Level Programming Model for
Pervasive Computing environments

• Anand Ranganathan, Shiva Chetan, Jalal Al-Muhtadi
  
• University of Illonois At Urban-Champaign
• Supercomputing Lab.
• Joo-kyoung Park (05. 7. 26)

2
Contents

• Introduction
• Active Spaces and Gaia
• The Olympus Programming Model
• Virtual Entities
• High-level Operators
• The Discovery Process
• Ontology hierarchies of entities
• Ontological description of entities
• Developer-specified constraints
• Semantic Matching of classes
• Space-level policies
• Utility function
• High-level Operators
• Example Programs
• Related work
• Conclusion and future works

3
Introduction

• Characteristic of Active Space
• Device-rich, Context-aware environment
• Very highly dynamic
• Problem of programming Active Space
• The resource and context can change rapidly
• Developer can not expect to know how various
  tasks are performed
• Program must be able to adapt the changing
  context and resource
• Can solve High-level abstraction
• Propose Olympus Programming model
• Abstract entity and operator

4
Active Space and Gaia

• Active Space
• A physically-bounded collection such as a room,
  devices, objects, users, services and application
• Gaia
• Meta-operation system
• Manages resources of an active space
• Contains a set of core service that manages the
  resource collection
• Provide a programming interface to application
  developer

5
The Olympus Programming Model

• Programming model
• Defined as an abstract machine
• Provide certain data type and operation to high
  level
• Require implementation for each these operation
  to low level
• Virtual Entity
• Abstracted entity
• Olympus allows a user to program and Active Space
  in terms of virtual entity
• Virtual entity associated with class defined in
  an ontology
• Eight basic type of entities
• Application, ApplicationComponent, Device,
  Service, Person, PhysicalObject, Location,
  ActiveSpace
• High-level Operator
• Olympus can provide Active Space operator in
  their high-level program

6
The Discovery Process(1/2)

• Choose appropriate instance for the virtual
  entity
• There are various type of constraints
• Constraints of the value of variable specified
  high-level program
• Constraints listed in ontology
• Policies specified by a Space Admin
• User policies for the user using the program

7
The Discovery Process(2/2)

• Discovering suitable classes of entities
• Quries ontology server to find semantically
  similar to the virtual entity class
• Checking class-level constraints on the similar
  classes
• Filters the list of classes depending on whether
  they satisfy class-level constraints (Used Prolog
  reason engine)
• Discovering entity instances in the current
  space
• Quries Space Repository to get instances of the
  classes that are available in the space
• Checking instance-level constraints on the
  similar classes
• Check if it satisfies instance-level constraints
  specified in the program or policies (Used Prolog
  reason engine)
• Choosing the best among the possible instances

8
Ontological hierarchies of entities(1/3)
Discovery process

• ApplicationComponent hierarchy
• Built using an extension of the
  Model-view-Controller framework
• Model
• Implements the logic of application and export to
  access it
• Presentation
• Provide output interface
• Adapter
• Map controller input to method call on model
• Controller
• Act as input interface to application
• Coordinator
• Manage the composite of the other components of
  the framework

9
Ontological hierarchies of entities(2/3)
Discovery process

• ApplicationComponent hierarchy can allow a class
  to have multiple parents
• Help in identifying how similar any two entity are

10
Ontological hierarchies of entities(3/3)
Discovery process

• Device and other Entity type hierarchy
• Similar to the ApplicationComponent hierarchy

11
Ontological Description of entity
Discovery process

• Each class is associated with a description in
  OWL
• OWL describe various semantic property such as
  the task it can perform, the class of device that
  can host it
• All application Component classes have an OWL
  description
• Define relations between different concept
• Ontologies are initially created by an admin of
  space and can extend application developer

12
Developer-specified constraints
Discovery process

• Developer can specify constraints that the class
  and instance of virtual entity
• Take the form of triple (ltentitygtltpropertygtltvaluegt
  )

13
Semantic Matching of classes(1/2)
Discovery process

• Semantic similarity of entity class
• Based on the principle that entity can be
  substituted by another if they are semantically
  similar (EX. Powerpoint -gt acrobet)
• Matching rule
• C1 is equivalent to C2, with similarity-level 0
• C1 is a sub-concept of C2, with similarity-level
  1
• C1 is a sub-concept of C2 or C1 is a sub-concept
  of a super-concept of C2 whose intersection with
  C2 is satisfiable, with similarity-level i2,
  where I is the number of nodes in the path in toe
  ontology hiererchy graph from C2 the relevant
  super-concept of C2

14
Semantic Matching of classes(2/2)
Discovery process

• Example
• similarity-level 0 PowerPointViewer
• similarity-level 1 None
• similarity-level 2 AcrobetReder, GhostScript
• similarity-level 3 WindowsMediaPlayer, Paint
• similarity-level 4 WinAmp, GaiaSpeechEngine

15
Space-level Policies
Discovery process

• Written by the administrator of the space
• Policies are written in the form of rules in
  Prolog
• Specify constraints on the classes and instances
  of entities
• Class level constraints
• Instance level constraints

16
Utility Function
Discovery process

• Employ a multi-dimensional utility function in
  order to choose best entity instance
• Location of the entity
• Task supported by the entity
• State of the entity
• Context of the space

17
High-level Operators(1/2)

• Operators can express many common Active Space
• Operators act on a target entity and can have
  zero or more argument
• Event operator generate event when the associated
  condition become true

18
High-level Operators(2/2)
19
Example Program

• Consist of two parts
• Virtual entities are declared
• Developer specifies the type of the entity and
  its properties of constraints
• Use high-level operators on the entity
• Developer can iterate through the list and
  perform operations on instance

20
Related work

• Task-computing model
• Allows a user to specify a behavior as a set of
  task that need to be completed
• System determine the way the task are to be
  composed
• Operator graph model
• Use programming model where service to be
  composed are specified as description and
  interaction
• Other discovery mechanisms
• Jini Lookup, CORBA, UDDI are used for discovering
  entity
• But specify constraints based on name-value pair

21
Conclusion

• Relieves developers from task of discovering
  proper entity for performing task
• Provide abstraction for common Active Space
• Provide a way of integrating policies and a
  utility model into the discovery process