Dynamic Layout of Distributed Applications in FarGo

1
Dynamic Layout of Distributed Applications in
FarGo

• Ophir Holder, Israel Ben-Shaul, Hovav Gazit
• Presented by Hossein Tajalli
• November, 2007

2
Mobility and Embedded Systems

• One of the challenges
• System Parameters Change by Time
• Bandwidth
• Machine Load
• Availability
• We need to change the layout of the distributed
  system dynamically.
• Objects Moves Around
• Locational Relationships Change

3
Table of Contents

• Challenges of Dynamic Layout Implementation
• FarGo
• FarGo Structure
• Extended Research on FarGo
• Evaluation
• Summary and Conclusions

4
Challenges of Dynamic Layout (1)

• Location Relationship of Objects
• Location Transparent Design
• Programming Scalability
• Unreliable Application
• Performance
• Lack of Shared Memory
• Location Based Design
• System Scalability
• Complicated Programming
• E.g. CORBA, RMI

5
Challenges of Dynamic Layout(2)

• Granularity of minimal entity
• Fine-Granularity (objects)
• Flexible
• Unneeded Overhead
• Lack of Shared Memory
• New Programming Language
• e.g. Distributed Oz
• Coarse Granularity (Processes)
• Clear Separation of Local and Remote
• Efficiency Problem
• e.g. Telescript, AgentTCL

6
FarGo

• Since 1997
• Introduced in 1999
• Provides Dynamic Layout Capability in Run Time
• Programming of Applications Logic Separated from
  Programming of the Layout
• Extension of Java
• Enables Scalable Applications
• Provides Monitoring Facilities

7
FarGo Architecture

• FarGo contains two unique Layers
• Complet and Complet References
• Monitor

8
Complete, Inter-Complete Reference

• Complet
• Basic Moveable Unit
• In a mid-grained approach
• Complet Reference
• Connect Complets to one another
• Can Stretch gtgt become remote
• Can Shrink gtgt become local
• Adopt to Layout Changes

9
Complet References Types

• Link
• Pull
• Duplicate
• Stamp
• Bi-Directional Pull

pull
a
b
10
Monitor

• Monitoring API
• Dynamic layout algorithms within application
• Monitors system parameters (network bandwidth,
  machine load ) and system events
• Raises events
• Dynamic Changes Happen
• Event-Based Scripting Language
• Dynamic layout algorithms Outside of application
• Even after deployment
• Event Action Style
• Graphical Monitoring tool
• Analyze and manual Control

11
Extended Research (1)

• Multi-Threading Model for FarGo
• Presented in 2001
• Good threading model contributes to FarGo
• Complet-Based thread partitioning
• Transparency

12
Extended Research (2)

• FarGo-DA
• Presented in 2002
• Disconnect Aware Application
• Resource limitations
• Expensive Online Connectivity
• Frequent Disconnect Conditions

13
Extended Research (3)

• FarGo.Net
• Proves that the FarGo concept and design can be
  implemented simpler in .NET

14
Evaluation

• Strengths
• Good presentation of the concept
• Stating what other systems do
• Good example
• Weaknesses
• They have not presented any simulation to show
  the performance of the system or to compare it to
  other approaches
• Relationship to the Embedded system
• Well related to the concept of embedded systems

15
Summary Conclusions

• FarGo Provides Dynamic Layout Capability
• Separated Logic Programming
• Separated Layout Programming
• Monitoring Capability
• System Scalability and Programming Scalability at
  the Same Time
• Easy for Programmers ( Extension to Java)
• Special Multi-Threading Model
• Extended to FarGo-DA

16
Resources

• Ophir Holder, Israel Ben-Shaul, Hovav Gazit
  Dynamic Layout of Distributed Applications in
  FarGo., ICSE 1999,p 163-173
• Israel Ben-Shaul, Hovav Gazit, Yoad Gidron, Ophir
  Holder, Boris Lavva FarGo A System for Mobile
  Component-Based Application Development. ICSE
  1999, p 658-659.
• Miki Abu, Israel Ben-Shaul A Multi-threading
  Model for Distributed Mobile Objects and Its
  Realization in FarGo., ICDCS 2001, p 313-321.
• Yaron Weinsberg, Israel Ben-Shaul A programming
  model and system for disconnected-aware
  applications on resource-constrained devices.,
  ICSE 2002, p 374-384.

17
Thanks For Your Attention!

• Questions and Comments