[Next] [Previous] [Top]

Workstation Clusters

Cluster definition

• A cluster has been defined to be

"a collection of computers on a network that can function as a single computing resource through the use of additional system management software".

June 1994

BASIC SOFTWARE COMPONENTS

• LANGUAGES - two broad categories
  • conventional languages such as C and Fortran, which require additional functionality (usually from libraries) to support parallelism;

  • languages with syntax constructs directly supporting parallelism.

• LIBRARIES
  • support the development of parallel software within cluster environments, in general, may be used with conventional programming languages to provide support for parallelism.

• DEBUGGERS

• CLUSTER QUEUING SYSTEMS
  • a unified cluster providing distributed computing services is achieved through use of specialised system management

LANGUAGES

A cluster requires languages which support parallelism

• C, C++ , GCC & G++
  • C is the most common language used for systems programming on UNIX workstations.

• Fortran 77, G77, Fortran 90
  • 1978 - revised standard, Fortran 77 released

  • 1991 - Fortran 90 released, 3rd ANSI standard

• Source to source code restructuring
  • HPF - combines Fortran 90 with enhanced user defined data distribution

  • Fortran 77 to 90 Converter

  • Adaptor - Automatic DAta Parallelism TranslatOR, transforms data parallel programs to parallel programs with explicit message passing

LIBRARIES

A cluster requires message passing libraries for parallelism

• MPI
  • Message Passing Interface, proposed as a standard for writing message passing programs

• PVM and XPVM
  • Parallel Virtual Machine - software library system enabling a collection of heterogeneous computers to be used as a flexible, concurrent computational resource

• CHIMP
  • Common High-level Interface to Message Passing, a project from EPCC

• Linda
  • a parallel programming language which extends a sequential language such as C or Fortran

• P4
  • parallel programming library consists of macros and subroutines for writing shared-memory and message passing programs in both C and FORTRAN.

• PICL
  • Portable Instrumented Communications Library

  • a subroutine library used to develop parallel programs that are portable across distributed memory architectures

UTILITIES AND GUI LIBRARIES

• MOTIF and X11
  • It is useful to have a common interface on all machines in the cluster

• TCL and TK
  • PVM requires TCL and TK

  • TCL - the basic programming language

  • TK - ToolKit of widgets i.e. graphical objects similar to those of other GUI toolkits eg Xlib, Xview and Motif.

DEBUGGERS

• Xdbx
  • a graphical user interface for dbx (UNIX)

• Xmdb
  • a source compatible single processor implementation of PVM 3.x (Oak Ridge National Lab)

• HeNCE
  • Heterogeneous Network Computing Environment

  • integrated graphical tools for creating, compiling, executing, and analysing HeNCE programs

  • relies on the PVM system

• Paragraph
  • visualizing the behaviour and performance of parallel programs on message-passing parallel computers

CLUSTER QUEUING SYSTEMS

• DQS - Distributed Queuing System
  • batch environment containing different queues based on architecture and group

• CODINE -COmputing in DIstributed Networked Environments
  • optimal utilization of the compute resources in heterogeneous networked environments

• NQE Network Queuing Environment
  • UNIX batch environment consisting of NQS and NLB

• LSF
  • Load Sharing Facility distributes the workload around one or more large heterogeneous clusters of workstations

• OTHER QUEUING SYSTEMS
  • Balans

  • Load Balancer

  • NC Toolset

  • Task Broker

  • Condor etc