From: Steve Coleman <Steve.Coleman@jhuapl.edu>
Newsgroups: comp.parallel.pvm
Subject: New user questions
Date: Tue, 29 Jun 1999 13:03:45 -0400
Organization: Johns Hopkins Applied Physics Laboratory
Message-Id: <3778FC71.86F455B8@jhuapl.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Xref: ukc comp.parallel.pvm:8526


Hi,

I have a couple of questions.

1) Is there a way to get pvm_spawn to dynamically pass environment
variables during an exec? I have seen mentioned something about
PVM_EXPORT being used for this. I guess that a process would have to
dynamically perform a setenv() in it's own environment with the required
values just prior to the pvm_spawn in order to export what it wants the
new process to receive. This sounds intrusive (or even dangerous) for
architecture dependent variables like LD_LIBRARY_PATH where the two
hosts have physically different path requirements. Is this correct?

2) Does the messaging/group system in PVM provide "reliable", ordered,
or virtually synchronous Quality of Service (QoS) delivery as in ISIS,
Totem, Horus, Ensemble, or Spread? What types of QoS does PVM
messaging/groups provide. I have seen in the documentation that the
order of messages with respect to the "sender" is maintained but have
not seen any indication that all peers see the same message ordering as
one another.

3) Is the pvmd itself fault tolerant in any way? Can the pvmd be
configured as a leader-follower replicated process?


-- 
Steve Coleman     <Steve.Coleman@jhuapl.edu>   http://www.jhuapl.edu/
<<--------->> Johns Hopkins Applied Physics Laboratory <<---------->>
Balt:443-778-6330 Fax:443-778-5597 Wash:240-228-6330 Fax:240-228-5597

