Service-oriented Resource Broker for QoS-Guaranteed in Grid Computing System Yichao Yang, Jin Wu,...

12
Service-oriented Resource Broker for QoS-Guaranteed in Grid Computing System Yichao Yang, Jin Wu, Lei Lang, Yanbo Zhou and Zhili Sun Centre for communication systems research University of Surrey Guildford, United Kingdom

Transcript of Service-oriented Resource Broker for QoS-Guaranteed in Grid Computing System Yichao Yang, Jin Wu,...

Service-oriented Resource Broker for QoS-Guaranteed in Grid Computing System

Yichao Yang, Jin Wu, Lei Lang, Yanbo Zhou and Zhili Sun

Centre for communication systems researchUniversity of SurreyGuildford, United Kingdom

Grid Computing What is Grid Computing?

A type of parallel and distributed system that enables the sharing, selection, and aggregation of geographically distributed resources depending on their availability, capability, cost, and user Quality of Service requirements for solving large-scale problems/applications.

Challenges of Grid Computing To deliver quality of service:

• The grid system performance efficiency depend on the how good QoS is provided for different user’s different requirement based on the factor of functions, cost, deadline, budget etc.

• Enhance real-time & QoS capability in grid middleware to meet the demand of scientific applications.

Coordinate resources:• It is necessary to have a resource scheduling that provides

matching users’ need and resource availability.

Problems of Grid QoS

How to provide guaranteed bulk data transmission for user’s QoS requirement.

The existing grid computing solution does not consider networks as a key resource to provide efficient services for grid apps.

Grid QoS Problem Solution

Propose an end-to-end QoS framework to satisfy grid user’s QoS requirements.

Introduce a service-oriented resource broker which can provide guaranteed QoS for grid application over large-scale grid network and optimal resource selection for resource sharing.

Implement cost-based resource selection algorithm which can effectively utilize potential combined resources and achieve high user request success rate in dynamic service-oriented Grid.

End-to-End QoS Framework

Data Resource

Data Resource

Trans Service

Data Service

Data Service

Trans Service

Trans Service

Broker

End-to-End QoS

Grid Message Flow

Data Flow

User User submit application request

Display all resources Grid

User access resources data

Broker makes resource reservation

Interaction Diagram

Broker architecture

Grid information

service

Resource Requirement Interpreter

Resource discovering unit

User resource requirements

Resource coupler

Resource reservation unit

Network resource index

update

update

List of resource combination

Resource Gatekeeper A

...

Reservation

Reservation confirmation

Service-oriented resource broker

Abstract resourcerepresentation

Resource availability info

Index services

Resource combination ranker

Topology discoverer

Resource priority list

Resouces

discovering

register

Internal call

External call

Existing component

Internal unit

Existing machanism

Discovery interface

Reservation interface

Simulation Network Topology

Resource 3

Resource 2

Resource 1

Broker

User

Grid Information

Service (GIS)

R1

R2

R3

R4

R51Mbps

1Mbps

Simulation Result (1/2)

Simulation Result (2/2)

Conclusion End-to-End QoS framework has

been defined Fully utilize combined resource to support user’s

requirements Implement cost-based resource selection

algorithm in Grid Resource Broker Discussion of Simulation Result

The higher number of the task, the more processing time saved.

The broker reservation scheduling has reduced the total processing time by 15-30% and reduced the average cost of processing by 4% when compared to non broker reservation scheduling.

Q&A