A beowulf for parallel instruction.
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Walter B. Ligon III walt at parl.ces.clemson.eduWed Nov 1 09:53:03 PST 2000
- Previous message: A beowulf for parallel instruction.
- Next message: A beowulf for parallel instruction.
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
-------- WARNING - WALT'S OPINION! NOONE'S OFFICIAL POLICY! I think this is a great idea - and I would suggest that you seriously consider the new Scyld Beowulf software as a platform. I have found that building parallel programs based on a data-parallel SPMD style works very will with this configuration and it is quite practical to teach a number of the classic parallel algorithms this way. I make a point of this because I find that once programmer start down a path, it forever dominates their destiny (wait, maybe they was something else ...) so if you are TEACHING parallel computing, it makes a lot of sense to teach a clean and well structured style. Unfortunately the past has been full of various systems-on-systems and master/slave setups and things that easily confuse the issues in parallel computing. I have found that a very simple and clean implementation based on N identical copies of a program operating as peers allows students to focus on parallel computing. Later there is time to learn to spawn tasks and set up signal handlers and do more sophisticated stuff. We recently constructed a cluster with 5 nodes (one front end and 4 compute nodes) with Celerons for under $4000. This makes an excellent system for students to develop their first parallel codes on. Walt > We are planning to provide a vehicle for instruction in > parallel computing. Given this rather vague objective, > what influence might this have on the selection of > hardware and software? The selection of non-bleeding > edge components is one obvious issue but there must be > others. Although this is slightly off-topic I'd > greatly appreciate comments from those who are running > machines that may cater to this area as part of their > workloads. > > _______________________________________________ > Beowulf mailing list > Beowulf at beowulf.org > http://www.beowulf.org/mailman/listinfo/beowulf -- Dr. Walter B. Ligon III Associate Professor ECE Department Clemson University
- Previous message: A beowulf for parallel instruction.
- Next message: A beowulf for parallel instruction.
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
