Grid Computing

1
Grid Computing Tuple Space

• Presented by
• Nelson Chu

2
What is Grid?

• Computing Grid is a very large distributed
  computing network, Virtual Organizations (VO)
• Grid is an infrastructure to collaborate many
  diverse computing resources in a network to
  accomplish a task(s).

3
What is Grid?

• Classification
• Computational Grid (eg. WestGrid)
• Desktop Grid (eg. SETI_at_HOME)
• Data Grid (eg. Earth Systems Grid)
• Global Grid Forum (GGF), Globus, etc.
• Open Grid Services Architecture - OGSA
• Web Service Resource Framework - WSRF (OGSI)

4
Grid Architecture
Fabric Physical resources
Resource Connectivity Protocols secured access
Collective Services brokering, monitoring
User Applications/Tools
5

• Fabric layer resource management
• e.g. computational resources, storage resources,
  and network resources.
• Connectivity layer Communication and
  authentication
• Collective layer global-naming, brokering,
  scheduling, workload and workflow management, and
  data replication.
• Application layer user defined tools and
  application

6
OGSA Architecture
Architect Service
7

• OGSA provides a service oriented architecture to
  simplify grid services development and to build a
  transparent grid structure
• Physical resources (for example storage,
  network, and computer) and logical resources
  services
• Web services (including WSRF)
• OGSA architect services (such as security
  management, data management, workflow management,
  file system management, etc.)
• Grid applications

Architect Service
8
Research Opportunities Grid

• Security
• Multi-institutional users and heterogeneous
  resources
• Authentication authorization mechanisms
• Different local name spaces, credentials, and/or
  accounts
• Trust mechanisms domain-based, user-based,
  service-based, network-based, and scheduler-based
• Fault Tolerance
• Detect and notify a fault occurrence
• Handle any failure

9
Research Opportunities Grid

• Scheduling
• Workflows in a service-based environment
• Split and distribute jobs to resources
• Time, cost, performance, completeness, fairness,
  effectiveness, and security policies.
• Resource discovery
• System selection
• Job submission resources reservation and
  supporting operation
• Monitoring and cleaning up

10
Current Research Trends

• Scheduling, Security, Fault Tolerance
• Replica location services
• Metadata catalog services
• Reliable file transfer services
• Globus Toolkit 3 / 4 Replica Catalog, GridFTP,
  and (GASS) Global Access to Secondary Storage .
• Globus, IBM, Oracle 10g, Sun, etc.

11
Research Opportunities Data

• Data Versioning
• Global name space data characterization,
  heterogeneous systems
• Data Security
• Fault tolerance recovery
• New challenges inter-domains replication, data
  synchronization, replica registration, querying,
  and meta-catalog.

12
Grid Environment

• Resources collaboration
• Loosely coupling
• Time, space, destination
• Communication allows significant degree of
  interoperability, synchronization, and
  extensibility
• Simple, flexible, and powerful mechanism

13
Grid System

• Open issues data synchronization, fault
  tolerance, persistence, concurrency,
  transparency, etc.
• Tuple Space paradigm
• Simple operations provide powerful inter-process
  mechanism for communication and synchronization.
• Flexible for function expansion ordered tuple,
  leasing,

14
Tuple Space

• Developed in a parallel and distributed
  programming system, Linda, at Yale U
• Basic element Tuple
• Tuple a vector of typed values / fields /
  variables
• Operations Writing, Reading, Consuming
• Matching

15
Tuple Space

• TS Vs DBS

Producers
Take B wait/abort
SpacesOne
A
Read
C
Write
Consumers
Processes
SpacesTwo
16
Tuple Space

• Globally shared memory space (Communication
  buffer)
• Tuples are associatively addressed by matching
• Loosely coupling
• Mechanism for inter-process communication and
  synchronization
• Simple, flexible, powerful,
• gt Grid Resource Collaboration

17
Tuple Space on the Grid

• Service Registration
• Service Discovery
• Service Lookup
• Job Scheduler
• Data Mgmt

Consumers
Tuple Spaces (Services / Tasks / Data)
Producers / Processes
18
Tuple Space on the Grid with Java Technology

• Look up Service
• Discovery Mgmt
• Leasing
• Plug and Work

• Resource Mgmt (GRAM)
• Security Feature
• Fault Tolerant
• Data Mgmt

Grid Enabled

• Resources / Services Collaboration

19
Grid Job Scheduling

• Eg) WestGrid,

• Centralized System
• Load balance
• Synchronization
• Fault Tolerance
• Policy
• Meta-Scheduler

Grid I Scheduler/ Resources Broker
Resource
Resource
Meta-Scheduler
Grid II
20
Tuple Spaces Based Gird Job Scheduler

• Fault Tolerance Management
• Leasing Structure for Dynamic Participation
• On-Demand Service
• Automatic Load Balance
• Global Name Space
• Security model ?

21
Gird Data Management System with Tuple Spaces
22
References

• Ian Foster, Carl Kesselman, and Steven Tuecke,
  The Anatomy of the Grid Enabling Scalable
  Virtual Organizations, Intl J. Supercomputer
  Applications, 2001
• Ian Foster, Carl Kesselman, Jeffrey M. Nick, and
  Steven Tuecke, The Physiology of the Grid An
  Open Grid Services Architecture for Distributed
  Systems Integration, http//www.globus.org, 2002
• Gregory T. Byrd, and Praveenkumar Cheruvu, Tuple
  Space Computing on the Grid, North Carolina
  State University
• K.A. Hawick, H.A. James, and L.H. Pritchard,
  Tuple-Space Based Middleware for Distributed
  Computing, Technical Report DHPC-128, University
  of Wales
• Globus, http//www.globus.org
• IBM, www.ibm.com/grid

23
Discussions

• Thank you !

24
Grid DBS Vs Distributed DBS

• Grid DB System is a Distributed DB System
• Inter-domains, Policy, Security
• Large amount of data intensive users/applications
  access to large amounts of data (terabytes or
  petabytes)
• Distributed data population (production,
  replication)
• Query processing techniques
• Secured and efficient data transport protocol
• Dynamic environment and data
• Grid (loosely coupled) DDBS (highly coupled)