Grid research trend 2005

1
Grid research trend 2005

• (2005. 8. 19)
• Kyung-Lang Park

2
About this seminar

• What we have to learn in the seminar?
• Where the Grid is heading for and its research
  trend
• Role of MPI in the Grid
• Can it be a high-level programming language?
• What we have to do after the seminar?
• Write a report
• Define a framework and suggest the role of MPI in
  the framework

3
Reference

• Service-oriented Science I. Foster
• A Loosely Coupled Vision for computational Grids
  I M. L iorente et
• An Ecosystem of Grid Components
• The Role of the Globus Toolkit
• GGF 13 and 14 Summary
• Grid Book Chapter 4 and 17
• OGSA Version 1.0
• Some documents about OGSI and WSRF

4
Contents

• Summary of some papers
• Movement of GGF
• OGSA, OGSI, WSRF, and GTK
• Globus 4.0 and WSRF (Next week)

5
Service-oriented Science

• 2005 May, Science, Vol. 308, Ian Foster

6
Service-oriented Science

• The paper describes how service-oriented
  computing is used for science
• Service-oriented science refers to research
  enabled by distributed networks of interoperating
  services
• Job can be automated by having services access
  services
• Service provides machine-intended uniform
  interface
• Creating and Sharing services
• A variety of commercial and open-source Web
  services tools exist
• But, problems remain
• Interoperability
• Scale
• Management
• Quality control
• Incentives

7
Service-oriented Science

• Rethinking infrastructure Outsourcing
• Three required resources
• The domain-specific content
• The domain-independent software functions
• Physical resources
• Last two can be handed off to specialist
  providers
• Scientists can focus on what they are good at
  providing contents and advancing science
• SourceForge for science
• Approaches to scaling
• Cookie-cutter researcher create dedicated
  domain-specific infra
• BIRN, NEES, PlanetLab,
• Researchers develop service ecologies in which
  agreements on interfaces allow participants to
  provide content and function in any way they see
  fit
• ESG, myGrid, Energy Fusion Collaboratory
• Definition and deployment of general-purpose
  infrastructure that deliver discipline-independent
  resources or functions
• OGS (Open Science Grid), EGEE

8
Loosely-Couple Vision for Computational Grids

• 2005 June, IEEE DS Online
• I. M. L. lorente and Ruben S. Montero

9
Loosely coupled vision for computational Grids

• The coexistence of different infrastructure opens
  an interesting debate about their coordinated
  operation
• Three requirements
• Isnt subject to centralized control
• Based on standard, open, and general-purpose
  interfaces and protocols
• The interfaces and protocols provide some of
  quality of service
• The tendency ignores the first two requirements
  to get higher levels of QoS, but they are the key
  to grids success in our loosely coupled model

10
Loosely coupled vision for computational Grids

• It resembles the Internet
• Instead tailoring the core middleware to our need
  or homogenizing the underlying resource, strictly
  follow the end-to-end principle
• Certain required end-to-end functions can only be
  performed correctly by the end-systems themselves
• Globus follows the principle
• Four main characteristics of LC
• Scalability
• Autonomy of multiple administration domain
• Dynamism
• Heterogeneity

11
Loosely coupled vision for computational Grids

• Some existing projects
• EGEE (Enabling Grids for E-Science)
• IRISGrid
• NorduGrid
• Loosely coupled grids allow straightforward
  resource sharing
• resources are accessed and exploited through de
  facto standard protocols and interfaces
• Similar to the Internets early stages

12
Ecosystem of Grid Component

• http//www.globus.org

13
An Ecosystem of Grid Component

• In 90s, computer scientist enjoyed helping
  scientists do these
• Computation intensive applications
• Data intensive applications
• Distributed collaboration
• Some problems are listed
• System administrator cant agree on a uniform
  authentication system
• But you have to allow your users to authenticate
  once then use services on all systems
• Need to be able to offload work during peak times
• But the volume of work theyll accept changes
  from day-to-day
• You and your colleagues want to start sharing
• But no one is willing to submit the data to a
  centrally managed storage system
• You need to support 24 experimental teams
• But each team has its own coordination code and
  each facility has its own control and data
  acquisition mechanisms

14
An Ecosystem of Grid Component

• Generally, Grid pioneers found that
• It was too hard to keep track of authentication
  data across institutions
• Too hard to monitor system and application status
  across institutions
• Too easy to leave dangling resources lying
  around, reducing robustness
• Too many ways to submit jobs
• Too many ways to store, access, and ultimately
  manage distributed data
• They finally found reusable solutions is key
• Most solutions should be
• one or more generic solutions
  application-specific additions
• Eg.,) Myproxy, ganglia

15
A Healthy Grid Ecosystem

• The collection of generic components can be
  viewed as an ecosystem
• Healthy force includes competition, innovation,
  evolution, diversity, and natural selection
• Unhealthy forces includes hegemony, centralized
  control, top-down policy enforcement, redundancy,
  protected diversity, imaginary boundaries
• It is important to remember that your components
  occupy niches in the Grid ecosystem.
• There are a lot of niches available
• You must understand what the niche is
• Key requirement for success is usefulness

16
Grid architecture in a nutshell

• Trying force homogeneity on distributed groups of
  collaborators is futile
• We have to learn from Internet protocol
• The Open Grid Services Architecture extends the
  IP network services upward from physical
  transport of messages to a variety of services
  that have proven common in Grid applications.

17
OGSA, OGSI, WSRFand GTK
18
Key concepts

• Web services
• Standard interoperable technologies
• Grid services
• Extension of WS
• OGSA defines Grid architecture and Grid services
• OGSI specify Grid services
• GTK is an implementation of OGSI

19
Key diagram
20
OGSI to WSRF

• Open Grid Service Infrastructure
• Grid services are extensible Web services
• It provide stateful Web services by using
• Factory mechanism
• Service Data Elements
• It also provide some functionalities
• Notification
• Lifecycle management
• But, there are some drawbacks
• Does not work well with existing Web services
  tooling
• Too many stuff in one specification
• Too object oriented
• WSRF proposal is a refactoring of OGSI concept to
  align better with Web services

21
Key diagram
WSRF
OGSA v1.5
GT4
22
The role of Globus Toolkit

• Provide standard system components that can
  support a wide variety of highly customized
  applications and user interfaces without
  requiring a completely unique infrastructure to
  be developed for each application
• GTK doesnt provide a complete solution for Grid
  projects
• Collection of tools which can be solutions to
  sub-problems
• It provides
• Two very important SDK
• Web Services Core implementation
• Grid Security Infrastructure Implementation
• A Variety of Basic Grid Services
• Developer APIS
• Tools and Examples
• Currently GTK 4.0 is available

23
GGF Summary

• GGF pervasive adoption
• Objectives
• Community Satisfaction
• Standard Adoption
• Publish GGF Standards Roadmap
• Publish OGSA v 1.5
• Liaisons with standard organizations
• Operations Efficiency
• Upcoming documents
• A Roadmap for the OGSA
• OGSA Profile definition
• OGSA WSRF Basic Profile 1.0
• Job Submission description Language 1.0
• WS-Agreement Specification

24
Conclusion

• Keywords
• Service-oriented
• Loosely-coupled
• Ecosystem
• Generic reusable components application-specific
  additions
• Rethinking the MPICH-GX