hi there, now can you help me about parallization between two computers. whenever giving the following command the fluent is being divided into three parts rather going in one for each of two core & another processor in onother computer.

fluent 2ddp -t2 -cnf= hosts.txt- path \\AERO-PC\Fluent.Inc

the host computer is AERO-PC (core 2 duo) & the other computer is ASTRO-U2(PENTIUM-3).

hosts file has---- AERO-PC AERO-PC ASTRO-U2

i am using local direct LAN connection between this two computers.

i saved host file in different places in the fluent.inc directory.

help please.

if you any more details please ask.

hi , i did the parallization, but the speed of processing fell, although all the processors and cores were working 100%.

PLEASE HELP.

hi fellows, i successfully did parallalization betwn core2 duo and pentium 3. but after working for few no of times i am facing following problems, shown in the Fluent console.............

abort: Unable to connect to 'ASTRO:8676', sock error: generic socket failure, error stack:

MPIDU_Sock_post_connect(1200): unable to connect to ASTRO on port 8676, exhausted all endpoints (errno -1)

MPIDU_Sock_post_connect(1247): unable to connect to ASTRO on port 8676, No connection could be made because the target machine actively refused it. (errno 10061)

please help.thanks.

If you really are trying to do a distributed parallel run using a Core 2 Duo and a Pentium 3 (!), the best advice I can give you is to unplug the Pentium 3 and rather use it as a doorstop. By adding another computer to a parallel run, you introduce latencies and other delays. The gap in performance between the two machines is so big, that you are most likely to see a loss in speed, particularly if the problem has been partitioned equally.

first thing, i am not using it as a doorstop.

i am working on load distribution & load balance & grid pertition.

last thing , nothing is impossible.

thanks for wishing me goodluck.

OK, well good luck with your project then. But if the objective of your study is parallel computing methodology, is it sensible to be doing this with Fluent rather than a code that you have source code access to?