🚀 go-pugleaf

A new mailing list is being set up for people who are interested in
getting news & announcements and participating in discussions related
to the Distributed Job Manager (DJM).

The name of the new mailing list is djm@msc.edu.  To get on the mailing list,
please send a message to djm-request@msc.edu.

To forestall a flood of queries (like "what is DJM?"), I've included a 
a list of Frequently Asked Questions at the end of this message.

The purpose of the new list is to promote the exchange of bug and patch
information, answer questions, and offer suggestions for enhancements
of DJM.  General discussions about managing parallel machines are also 
welcome.


FAQ List for DJM

1. What is the Distributed Job Manager?
   ------------------------------------

The Distributed Job Manager (DJM) is a free drop-in replacement for the
Network Queuing System (NQS) for managing batch and interactive jobs on a
Thinking Machines Corporation CM-2 or CM-5 system.  It was originally 
developed by the Minnesota Supercomputer Center on behalf of the U.S. Army
High Performance Computing Research program.


2. What does DJM do?
   -----------------

Briefly, DJM manages the resources of a Connection Machine system.  It
allocates resources (memory, processors, time) to jobs based on a 
set of site-defined scheduling criteria.  It provides efficiency
and flexibility for system administrators, and provides ease-of-use 
and productivity for end users.

The major features of DJM include the following:

	o Upward compatible with NQS

	o Full support for interactive use

	o Load Balancing

	o Flexible controls and fault tolerance


3. What machines are supported?
   ----------------------------

DJM currently supports the Thinking Machines Corporation CM-2 and
CM-5.  The package is modular to promote easy adaptation to other MPP
architectures as well.  It should be especially useful on any MPP machine 
that can be divided into partitions or can be controlled by independent
front end computers (e.g. Intel iPSC, Cray MPP, MasPar, Ncube).


4. Who else is running it?
   -----------------------
 
Since its introduction in July of 1992, DJM has been installed at several
Connection Machine sites.  These include the Pittsburgh Supercomputing Center,
Los Alamos/ACL, NASA/NAS, ANU, and NCSA.


5. Ok, I'd like to try it out.  How do I get it?
   ---------------------------------------------

DJM is freely available software under the auspices of the GNU copyright.
The beta-test release is available from the CM LIGHTNING User Group archives.
Use anonymous ftp to ec.msc.edu and fetch /pub/LIGHTNING/djm_0.9.8.tar.Z.

You should test it extensively under `live' operating conditions for
several weeks, running it side-by-side with NQS, before you switch over.

All documentation is included.  See the file INSTALL for installation
instructions.  You need an ANSI C compiler, preferrably gcc version 2.1
or later.


6. What kind of support can I expect?
   ----------------------------------

As much as you paid for: NONE!   One of the reasons for creating the mailing
list is to promote exchange of bug and patch information.  If you include
source code patches, I will try to fold them in to the distribution release
as time allows.  Suggestions and patches for enhancements are also welcome.



DISCLAIMERS.  This project was funded in part by the U.S. Army Research
Office under contract no. DAAL03-89-C-0038.  There are NO WARRANTIES,
EXPRESS OR IMPLIED.  The package includes software developed by 
the University of California, Berkeley and its contributors.


--
Alan E. Klietz
Minnesota Supercomputer Center, Inc (MSCI).
Army High Performance Computing Research Center (AHPCRC)
1200 Washington Avenue South
Minneapolis, MN  55415
Ph: +1 612 626 1737	       Internet: alan@msc.edu