A TMObased Object Group Model to Structuring Replicated RealTime Objects for Distributed RealTime Ap

1
A TMO-based Object Group Model to Structuring
Replicated Real-Time Objects for Distributed
Real-Time Applications

• Chang-Sun Shin, Su-Chong Joo, and Young-Sik Jeong
• School of Electrical, Electronic and Information
  Engineering, Wonkwang University, Korea

2
Content

• Introduction and Related Works
• TMO-based Object Group Model
• TMO Scheme
• Structure of TMO-based Object Group Model
• Dynamic Object Selection and Binding Service
  Strategy
• Dynamic Object Selection and Binding Strategy
  form Replicated TMOs
• Real-Time Service Strategy
• Conclusions

3
1. Introduction

• The distributed systems
• Physically extended forward the wide-area
  distributed real-time object computing
  environment.
• The distributed services
• Supporting the ubiquitous systems as a logical
  single view system.
• Logically reconfigured to the appropriate
  service-dependent groups.
• In this paper, our object group
• Supporting a logical application reconfiguration,
  for using the replicated resources

4
Related Works

• The Telecommunications Information Networking
  Architecture-Consortium(TINA-C)s TINA
• The distributed applications can logically be
  configured to groups as units of associated
  objects.
• Have not defined the real-time services in a
  distributed environment yet.
• The Real-Time Special Interest Group(RT-SIG)s RT
  CORBA
• Having the real-time extensibility for adding the
  real-time property to standard CORBA
  specification.
• Depending on the special system and/or the OSs
  for the real-time services
• Extending the ORB itself that is the core of
  CORBA.
• The Real-Time Object Group(RTOG) model
• Being developed as a framework supporting the
  distributed real-time services and the object
  group management
• Based on the TINAs object group concept.
• Using the Time-triggered and Message-triggered
  Objects(TMOs).

5
In this paper,

• The TMO Object Group(TMOOG) model
• Extending exiting RTOG model to the TMOOG model
• Defining the concepts of the TMO and the
  structure of our model.
• Designing the functions and interactions of the
  group components
• To verify the correct executions of the model
• Defining the dynamic object selection and binding
  and real-time strategies.
• These strategies are implemented in a Dynamic
  Binder object and a Scheduler object respectively
• The binding priority algorithm
• Earliest Deadline First(EDF) algorithm
• Finally, from the numerical execution results,
• Showed whether our TMOOG model could support the
  dynamic object selection and binding service and
  real-time scheduling service.

6
2. TMO-based Object Group Model

• TMO Scheme
• Real-time servicing object with the real-time
  constraints itself.
• The existing service object
• Not able to define the real-time constraints in
  its own data structure.
• Having the Spontaneous Method (SpM) that can be
  spontaneously triggered by the defined absolute
  time
• TMO structure scheme is designed by UCI DREAM
  LAB.
• ODS(Object Data Store)
• EAC(Environment Access Capability)
• AAC(Autonomous Activation Condition)
• SpM(Spontaneous Method)
• SvM(Service Method)

7
2.1 TMO Scheme

• The main features of the TMO
• Distributed computing component.
• Having two types of methods, SpMs and the SvMs.
• which are clearly separated from the existing
  object.
• Triggered by the BCC(Basic Concurrency
  Constraints).
• The deadlines are supplied to the all methods.
• TMO Scheme cannot support
• Checking the security for the object access.
• Managing the duplicated TMOs with the same
  property.
• The distributed scheduling service of several
  TMOs globally.
• For solving these problems, we suggested the
  TMO-based Object Group model.

8
2.2 Structure of TMO-based Object Group Model

• Using the object group concepts on COTS
  middlewares.
• Supporting the the object management service for
  dynamic object selection and binding service
• How to manage the selection and the binding among
  objects
• Supporting the Real-time scheduling service.
• How to support distributed global real-time
  scheduling service

9
Structure of TMO-based Object Group Model
TMO-based Object Group
SecurityObject
InformationRepository Object
SecurityObject
InformationRepository Object
SchedulerObject
SchedulerObject
DynamicBinder Object
DynamicBinder Object
RTMObject
RTMObject


Replicated TMOs
Sub-TMO Object Group


Replicated TMOs
other group
other group
COTS MIDDLEWARE
COMMUNICATION NETWORK
10
Components of Object Management Services

• Group Manager(GM) object
• Responsible for managing all of objects in an
  object group.
• Returning the reference of TMO requested from a
  client finally.
• Security object
• Checking the access rights of an object
• Referring the Access Control List(ACL).
• Information Repository object
• Storing information about all TMOs existing in an
  object group.
• Dynamic Binder object
• Selecting an appropriate object of the service
  objects that will be invoked by clients.
• Adopt the binding priority algorithm considering
  systems load information and the request
  deadline(RD).

11
Components of Real-Time Scheduling Services

• Real-Time Manager(RTM) object
• Calculating service deadline(SD) of the TMO
  object, when given a request deadline of a
  client.
• Scheduler object
• Assigning the task priorities of each TMO to the
  requesting tasks.
• Adopting the EDF algorithm.
• TMOs

12
3. Dynamic Object Selection and Binding Service
Strategy

• Service procedures and component functionalities
• The client requests a desiring TMOs reference to
  the Group Manager(GM) object.
• The GM object firstly checks the access right for
  the client via the Security object.
• The GM object sends a clients information to the
  Information Repository object.
• The Information Repository object sends
  replicated TMOs references to the Dynamic Binder
  object.
• Obtaining the reference of an appropriate object.
  
• The GM object receives the TMOs reference from
  the Information Repository object, and returns
  selected ones reference to a client.

13
The Binding Priority Algorithm

• The Dynamic Binder object
• Calculating the binding priority of each
  replicated TMO.
• Considering the load information of systems that
  TMOs are located on and the request deadline
  information of clients task.
• where request_deadline clients request
  deadline, CPU_utilization CPU utilization
  ratios, c rate constant(0.01)

14
The Binding Priority Algorithm
15
An executing example of the Binding Priority
Algorithm

• Replicated TMOs(TMO1 and TMO2) are being in a
  TMOOG.
• Each TMOs ready queue managing in the Dynamic
  Binder object is nulls.
• The CPU utilization ratios of systems in which
  TMO1 and TMO2 are located be the 10 and 11
  respectively and tasks are periodic every 1sec.
• The client1(c1)s request arrived to TMOOG
• The Request Deadline(RD) 100009 and the
  Client's Invocation Time(CIT) 090059, the GM
  objects time 100000
• Returning reference is the TMO1s reference since
  the CPU utilization of TMO1 is lower than TMO2s.
• The client2(c2)s request arrived to TMOOG
• The RD 100011, the CIT 100000, and the GM
  objects time 100001
• The TMO2s binding priority 0.1909, the TMO1s
  BP 0.1526
• These situations will be occurred repeatedly and
  continuously in TMOOG.

16
An Executing Example of the Binding Priority
Algorithm

• TMOs ready queue for object selection and binding

17
An Executing Example of the Binding Priority
Algorithm
18
4. Real-Time Service Strategy

• Service procedures and component functionalities
• A client tries to bind the selected TMO with
  Real-time Information(RI Client_Name, CIT, RD).
• TMO passes the received RI to the RTM object.
• The RTM object calculates the Service Deadline
• SDRD-Transfer Time(TT)
• The RTM object invokes the Scheduler object to
  decide the task priority of client request.
• The Scheduler object schedules its task priority
  using the EDF algorithm
• The Scheduler object requests the service to TMO
  via the RTM object.
• The TMO executes the real-time service and
  returns its executing result to a client.
• At the same time, the TMO informs the completion
  of service to the RTM object for executing the
  next requested service from another client.

19
4. Real-Time Service Strategy

• The EDF algorithm
• The highest priority task is defined a client
  request with minimum SD.
• The following simple expression uses for making a
  condition decision of the task priority(TP), when
  given SDs of two tasks.
• if SDi lt SDj then TPi gt TPj

20
An Execution Example of the Real-Time Service
Strategy

• The clients(c1,c3,c5,c6,c8) that will be bound
  with the TMO1
• The requests have been arriving to the TMO1
  sporadically and sequentially.
• In initial state, the TMO1 waits for executing
  clients requests.
• When the client1(c1)s request is arrived at the
  TMO1,
• The TMO1 executes the client request immediately.
  
• The TMO1 receives a new client3(c3)'s service
  request during the TMO1s execution,
• This request is passed to the Scheduler object
  and then stored in Scheduler object's ready
  queue.
• These situations will be occurred repeatedly and
  continuously.

21
An Execution Example of the Real-Time Service
Strategy

• Task Priority List in TMO1
• The clients requests are arrived in order of c3,
  c5, c6 and c8 to TMO1
• But, the c3, c5, c8, and c6 will be sequentially
  executed on the TMO1 in accordance with the
  non-preemptive mechanism.

22
An Execution Example of the Real-Time Service
Strategy
23
An Example of TMOOG Supporting Services
Clients Requests
Real-Time Scheduling Service
Client1
Client8
Dynamic Object Selection and Binding Service
Service Results
Client2
Client6
Client5
Client1
Client3
Client3
Client6
Client8
Client5
Client4
Client6
Client8
Client1
Client3
Client5
Client2
Client4
Client7
Client6
Client7
Client4
Client2
Client4
Client7
Client7
Client8
System Load Information Request Deadline
Request Deadline,Service Deadline, Transfer Time
etc.
Binding Priority Algorithm
EDF Algorithm
24
5. Conclusions

• The modern computing environments
• Changing to the distributed real-time object
  computing environments.
• Growing the real-time service requirements.
• Have to be efficiently managed and grouped with
  distributed objects with real-time properties
• For this reason, we proposed
• The TMO Object Group model
• Provide the object group management and the
  TMO-based real-time services on COTS middleware.
• Can adopt dynamic object selection and binding
  and real-time scheduling strategies for
  supporting reconfiguration of distributed
  real-time applications on the distributed
  environment.

25
5. Conclusions

• For achieving our goals,
• Described the concepts of the TMOs, the structure
  of the TMOOG, and the designing functions and
  interactions among the components.
• From the numerical executing results
• Whether our model could support
• The dynamic object selection and binding service
  of replicated TMOs,
• The real-time scheduling service for clients from
  the selected TMO.
• In future, for applying this model
• Having a plan to develop a prototypical platform
  that can adopt various dynamic object selection
  and binding strategies and real-time scheduling
  strategies in a TMOOG or between/among TMOOGs.