Design and Implementation of Service Data Aggregator for GAIS

1
Design and Implementation of Service Data
Aggregator for GAIS

• Dec 1 2003
• Kyungwoo Kang
• Cheonan University

2
Contents

• Introduction
• Grid Advanced Information Service
• Service Data Aggregator Architecture
• Conclusion

3
Introduction

• Requirements of Grid Information Service
• - A basis for configuration and adaptation in
  heterogeneous environments
• - Uniform and flexible access to static and
  dynamic information
• - Scalable and efficient access to information
• - Access to multiple information sources
• - Decentralized maintenance

4
Introduction

• Functionality of Information Service
• - Dynamic service data creation and management
  via information providers
• - Registration of Grid service instances
• - Aggregation of service data from multiple
  instances

5
Motivation

• How to collect service data for GAIS
• History information for analyzing and forecasting
  the performance of Grid
• Persistent Service Data

6
Project Objectives

• Analysis of Grid index service based on OGSA
  
• Design and implementation of an extended Service
  Data Aggregator using XML DB

7
GT3 Index Service
User
Service Data
Service Data
GSH
Index Service
Aggregator Mechanism(caching here)
Provider Mechanism
Registry Mechanism
Java Provider
Collective Layer
Existence Notification Message
Notification Message
Service Data
Service Data
Service Data
Grid Service
Grid Service
Grid Service
Resource Layer
8
Base GAIS
9
Service Data Aggregator

• Service Data
• - aggregated into any given service by using the
  ServiceDataAggregator class in service code
• ServiceDataAggregator class
• - provides a mechanism for handling subscription,
  notification and updating of locally stored
  copies of service data

10
Flow of Service Data
Service Data Collection Module
Grid Services 1
Service Data
Notification Sink
Grid Services 2
Service Data
Grid Services 3
Service Data

Xindice manager
Interpret
Xindice Server
Parse resource information
Lookup service name
trigger Collection operation
Repository
11
Functionality

• Support for Notification of SDE
• Classification of Service Data according to VO
• Maintain History Data
• Query Support

12
XML Database

• Xindice
• - store and index compressed XML documents
• - designed for managing many small to medium
  sized documents as a Native XML Database
• All data that goes into and out of is XML
• - XPath is used for a query language
• - The programming APIs support DOM and SAX

13
Comparison of Grid Information Service
14
Resource Information Category

• Single System
• VO Number, IP Address, CPU Load
• Cluster System
• VO Number, IP Address, Number of CPUs,
• CPU Load of each CPU
• Network
• Source/Destination System IP Address, Bandwidth,
• Latency, Packet Loss Rate
• Queue Status
• Queue Name, Time Limit, Number of Nodes,
• Number of jobs in Queue

15
(No Transcript)
16
(No Transcript)
17
Service Data Collection
18
System Architecture
19
Information Hierarchy
VO
Resource
Service
GridFtp
Duroc

Host
Cluster
Network
20
Sequence Diagram for Adding SDE
21
Sequence Diagram for retrieving SDE
22
Schedule
2
3
4
5
6
7
8
9
10
11
12
23
Conclusion

• Information Service
• XML Database (Xindice)
• Service Data Aggregator System Architecture
• Support for Notification of SDE
• Classification of Service Data according to VO
• Maintain History Data
• Query Support

24
Future Works

• Developing the ServiceData Broker
• Definition of Metadata about Service Data