Grid Computing Sun Grid Engine SGE

1
Grid ComputingSun Grid Engine (SGE)
Manoj Katwal
2
Outline

• Grid and Cluster Terminology
• What is a Cluster vs. a Grid?
• Types of Grids
• Benefits of Grid Computing
• Areas of concern
• What is Sun Grid Engine?
• Example N1 Grid Network Setup

Do you know?
Grid computing market will reach 12 billion by
the end of 2007 (IDC)
3
Outline, Cont.

• Managing Jobs and Resources with a Grid Sun N1
  Grid Engine
• Concepts in Sun Grid Engine
• Job Submission
• Queues
• QMON
• Useful commands
• References

Do you know?
Grid computing market will reach 12 billion by
the end of 2007 (IDC)
4
Grid and Cluster Terminology

• A GRID is a collection of computing resources
  that perform tasks. A grid appears to users as a
  large system that provides a point(s) of access
  to powerful distributed resources.
• Resources Computer hardware, data,
  applications,networking etc.
• Cluster
• Tightly coupled homogenous computers, resources
• Redundant interconnections that appears as single
  system with high availability (HB)
• Not truly distributed system

5
Grid and Cluster Terminology, Cont.

• Grid
• Loosely coupled heterogeneous computers or
  resources
• Resources can come and go
• High scalability dynamically resources can be
  added and removed
• It is distributed virtually everywhere
• High Performance Computing (HPC)

Do you know?
The current version of Sun Grid Engine is
downloadable for free.
6
What is a Cluster vs. a Grid?

• Gentzsch, Grid Computing A New Technology for
  the Advanced Web

7
Types of Grids
8
Benefits of Grid Computing

• Better utilization of resources
• Increased user productivity
• Faster time-to-solution of complex scientific,
  engineering, and enterprise computational tasks

Do you know?
You can register to http//www.network.com and
get 200 CPU hours free of charge. No obligation
9
Areas of Concern

• Users concern about Grid operation and Management
  include
• Confusion about what grid is and how it works
• Cross-vendor interoperability
• CPU and application usage billing and accounting
• Security and user authentication
• Standardization
• Global enterprise grids across international
  boundaries

10
What is Sun Grid Engine?

• Open source Grid management software accepts
  jobs submitted by users
• Software driving job distribution to compute
  nodes in a grid
• By job means Unix like jobs (batches) or parallel
  programs (MPICH- MPI Interface)
• Enable scheduling and prioritizing of jobs
• Management of jobs submit, suspend, abort
• Accounting and reporting (Arco, web based)
• Almost all operating systems are supported

11
Example N1 Grid Network Setup
12 billions dollars
12
Managing Jobs and Resources with a Grid
13
Concepts in Sun Grid Engine

• Hosts Four types
• Master host, Secondary Fail over host (optional)
• Execution hosts
• Admin host
• Submit hosts
• A system can act as more than one type of host.

14
Job Submission
15
Queues

• Cluster queues
• Job slots
• Resource reservation
• Parallel environment (MPI support)
• Policy based resource allocation
• Job checkpointing

16
QMON
17
Useful commands

• Qsub to submit job
• Qhost getting an overview of available
  resources
• Qacct job accounting information
• Qmod managing queue state
• Qmon graphical admin tool
• Qconf configuring queues

18
References

• By Jan Stafford, Editor, 30 Aug 2005
  SearchOpenSource.com http//searchenterpriselinux.
  techtarget.com/originalContent/0,289142,sid39_gci1
  120769,00.html
• California State University, LA cluster (1 head
  node with 10 clients) http//oscar.calstatela.ed
  u/
• Sun Grid Compute Utility Web Portal
  http//www.network.com