MathGroup Archive 2000

[Date Index] [Thread Index] [Author Index]

Search the Archive

About Parallel Computing

  • To: mathgroup at smc.vnet.net
  • Subject: [mg25829] About Parallel Computing
  • From: Makoto Nikaidoh <nikaidoh at mech.fukui-u.ac.jp>
  • Date: Sat, 28 Oct 2000 01:41:21 -0400 (EDT)
  • Sender: owner-wri-mathgroup at wolfram.com

Hello,
My name is Makoto Nikaidoh.
I am Japanese graduate student.

I want to Parallel Computing using Mathmatica Parallel Computingkit..
However, calculation time is determined by performance of the SlaveMachine.
Therefore I think that my Parallel Computing is failure.

Two PC is each machine A and machine B.
  machine A(CPU:Pentium$B-6(B 433MHz, Memory:128MB)
  machine B(CPU:Pentium$B-7(B 550MHz, Memory:128MB)
I use FrontEnd with machine A.

I executed next.
machine A:
  Needs["Parallel`Parallel`"]
  Needs["Parallel`VirtualShared`"]
  Needs["Parallel`Commands`"]
(Setting of kernel is always local)

MS-DOS prompt on machine B:
  math -mathlink -linkcreate -linkprotocol TCP

machine A:
  ConnectSlave[nnnn at xxx. yyy. zzz. ww]
  (nnnn:port, xxx. yyy. zzz. ww:IP address of machine B)
(I was not able to start up kernel of SlaveMachine automatically.)

machine A:
  TableForm[RemoteEvaluate[{$ProcessorID, $MachineName, $SystemID,
      $ProcessID, $Version}], TableHeadings -> {None, {"ID", "host", "OS",
           "process", "Mathematica Version"}}]
The kernel on machine B was slave of machine A.

I compared evaluation time of Do[i, {i, 1000000000}] with the next case.

case1   machine A(local Kernel)->2282.76 seconds
case2   machine B(local Kernel)->1770.79 seconds
case3   Master:machine A, Slave:machine B->1765.53 seconds
case4   Master:machine B, Slave:machine A->2252.88 seconds

Performance of two PC is each as follows.
  machine A(CPU:Pentium$B-6(B 433MHz, Memory:128MB)
  machine B(CPU:Pentium$B-7(B 550MHz, Memory:128MB)
Therefore I think that calculation time is determined by performance of the SlaveMachine.

I tried next in addition to RemoteEvaluate.
  j1=Queue[Do[i, {i, 1000000000}]]
  Wait[j1]
The result was the same as RemoteEvaluate almost,
but processor use rate was different.
  RemoteEvaluate:machine A, B->100%
$B!!(BQueue and Wait:machine A->about 10%, machine B->100%

Please teach a way of Parallel Computing.

***************************************************************
   Makoto Nikaidoh  nikaidoh at mech.fukui-u.ac.jp
    Graduate Student, Department of Mechanical Engineering
    Fukui University, 3-9-1 Bunkyo, Fukui, 910-8507, JAPAN
***************************************************************




  • Prev by Date: Re: Is it possible to plot a point a time on the same graph in Mathematica?
  • Next by Date: RE: Adding a new style environment
  • Previous by thread: A stupid but practical question
  • Next by thread: Question concerning Mathematica packages