Heterogeneous Computing

1
Heterogeneous Computing

• Instructor Dr. Najeh Jisrawi
• Prepared by Basim Al-Ajarmah
• Birzeit University
• Scientific Computing

2
Introduction

• . Definition of Heterogeneity.
• . How to build heterogeneous computing system.
• . PVM Example of build HC system
• . Heterogeneous computing advantages
• . Conclusion.

3
Heterogeneity of

• Architecture
• Data format,
• Computational speed,
• Machine load, and
• Network load,

4
Heterogeneous system
5
Architecture

• that include a wide range of architecture types
  such as 386/486 PC class machines,
  high-performance workstations, shared-memory
  multiprocessors.....

6
Data format

• data sent from one computer may be unreadable on
  the receiving computer
• Computational speed
• supercomputer doesn't sit idle waiting for the
  next data come from the workstation or 486 PC
  before continuing its task,

7
Machine load

• Networked computers may have several users on
  them running variety of tasks, which made the
  machine load vary dramatically
• Network load
• Time it takes the network to send all the
  massages over network it also varied.

8
How to build heterogeneous computing system

• Programming environment,
• Task profiling and Analytical benchmarking,
• Matching and Scheduling,
• Interconnection requirements

9
Programming environment

• Verity of tools that used to execute program
  over heterogeneous system like PVM, HeNCE, P4,
  Linda ..
• Task profiling and Analytical benchmarking
• decomposing program into subtask, and
  determining the suite machine for each subtask.

10
Matching and Scheduling

• decides the machine and when the subtask is to
  begin execution and
• So we need mathematical algorithms like OST
• Interconnection requirements
• network would require high bandwidth and low
  latency
• communication speed of the network will match
  the speed of the machines

11
PVM Example

• Grip all message routing in HC systems
• Decomposing the application into subtasks
• Automatically allocating them to the various
  machines.
• This powerfully tools exist in every HC computer
• It can convert data between two different
  architecture.
• It can arrange task across a network of
  incompatible computer architectures

12
(No Transcript)
13
Heterogeneous computing offers many advantages

• Costs are very low.
• Max the Performance by assigning each individual
  task to the appropriate architecture
• The virtual computer resources can grow in stages
  and take advantage
• Program development can be enhanced by using a
  familiar environment

14
Conclusion

• The heterogeneity idea is exists by default in
  networking system as in Figure 1
• There are different types of heterogeneity as
  architecture .
• To make parallel computing mechanism over
  heterogeneous system, you need to develop things
  as programming environment .