22
GRID RESOURCE BROKER FOR SCHEDULING COMPONENT-BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

Embed Size (px)

Citation preview

Page 1: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

GRID RESOURCE BROKER FOR SCHEDULING COMPONENT-BASED APPLICATIONS ON DISTRIBUTED RESOURCES

Reporter : Yi-Wei Wu

Page 2: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

ABSTRACT

This paper presents design and implementation of seamless integration of two complex systems ProActive Gridbus Resource Broker

ProActive → Leverage the economy-based and data-intensive scheduling algorithm

Gridbus Broker → Use component-based distributed applications.

2

Page 3: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

OUTLINE

Introduction Background Knowledge Integration Challenges System Implementation Validation Remarks of the Integration Conclusion

3

Page 4: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

INTRODUCTION

Grid The job includes requirements and submit to a

Grid Return the result of the execution

What’s need? Dynamically vary the parameter space Grid as a unified platform Component-based application

4

Page 5: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

INTRODUCTION

ProActive Use functions for handling the heterogeneity and

dynamicity of Grid resources

This paper An integration of Gridbus Broker and ProActive

framework

5

Page 6: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

BACKGROUND KNOWLEDGE

ProActive Grid Scheduler Gridbus Broker Scheduling Infrastructure

6

Page 7: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

PROACTIVE GRID SCHEDULER

What is ProActive? An open source java

library (GPL2)

For parallel, dist-ribution application

7

Page 8: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

8

Page 9: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

PROACTIVE GRID SCHEDULER

9

Why Integration? lack of advanced scheduling algorithm

Page 10: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

GRIDBUS BROKDER SCHEDULING INFRASTRUCTURE

10

Page 11: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

INTEGRATION CHALLENGES

ProActive Can leverage the economy-based and data-

intensive scheduling algorithm

Gridbus Broker Can utilise the programming environment

especially the component-based programming concept

11

Page 12: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

INTEGRATION CHALLENGES

ProActive + Gridbus Broker = Challenge?1. Identify is on the potential impact of the

integration on both systems. There should be no or minimum impact on both

systems

1. The reuse of existing infrastructure and codebase provided by both systems.

Maxmise reusability and avoid modifying the existing source code

12

Page 13: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

SYSTEM IMPLEMENTATION

13

Replacing the existing scheduling infrastructure dynamically

Responsible for wrapping the ProActive terms into the Broker’s terms that can be scheduled and managed

Page 14: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

SYSTEM IMPLEMENTATION

14

Page 15: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

VALIDATION

The integration solution is valid unless the following two conditions are satified1. Legacy applications developed should work

without changing and recompiling the source code

2. The dependencies between each system should be minimized

3. Reuse the existing infrastructure at both sides without adding new features.

15

Page 16: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

VALIDATION

C3D application Java benchmark application Measure the performance of a 3D raytracer

renderer over several JVM

16

Page 17: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

VALIDATION

17

Page 18: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

REMARKS OF THE INTEGRATION

Impacts on ProActive Impacts on Gridbus Broker

18

Page 19: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

REMARKS OF THE INTEGRATION-IMPACTS ON

PROACTIVE

System level- Scheduler add a new overload start method

Application level- The client applications developed using the

ProActive framework

Last notable- Check how many extra dependencies

19

Page 20: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

REMARKS OF THE INTEGRATION-IMPACTS ON GRIDBUS

BROKER

The broker role as a middle-man who is responsible for matching job to heterogeneous resources.

Various types of runtime Grid middleware can be plugged into runtime environment of broker at runtime via the configuration

20

Page 21: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

CONCLUSION AND FUTURE WORK

1. A schedule component-based applications on global Grids by utilising the Gridbus resource broker

2. The solution follows the object-oriented design principles

3. Legacy application running without modifying and recompiling the source code

4. Focus on other types of runtime environments

21

Page 22: G RID R ESOURCE BROKER FOR SCHEDULING COMPONENT - BASED APPLICATIONS ON DISTRIBUTED RESOURCES Reporter : Yi-Wei Wu

THE ENDThanks for your attention

22