Newsgroups: comp.parallel.pvm
From: hans leichtl <hans@brc.uconn.edu>
Reply-To: hans@brc.uconn.edu
Subject: Re: libpvm [txxxxx]: pvmmctl() connect: Connection refused
Organization: University of Connecticut
Date: Fri, 31 Jan 1997 02:46:46 -0500
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <32F1A366.71F3@brc.uconn.edu>

sorry, me again.

it just hit me,

if i have small problem sizes, then the runtime is very short.
and doing socket programming i know that the system only completely
frees up sockets after about 4 min, before that they are in a time_wait
state.

the problem now is that the sockets don't clear out fast enough. which
is not true for big problem sizes, as the runs take substantially
longer.

hans