CS 6404 Project Guidelines

General Requirements:

• Groups of two or three. If you know whom you would like to work with, and have some idea what you want to work on, then get together and prepare a proposal and submit it. If you don't know who you would like to work with, let Dr. Ribbens know, along with any preferences you have on topics or types of topics.

• Each group needs to submit a proposal by February 26 (earlier would be smart for groups presenting the first week or two after Spring Break). The proposal is just a page or two describing what you think you are going to do. Most groups will want to discuss their ideas first with Dr. Ribbens before even submitting this proposal. Pre-proposal discussions can take place during office hours or via e-mail.

  As your work proceeds, if you find that things are changing direction in some way, you will need to discuss these changes with Dr. Ribbens as well.

• Each group needs to submit a final written report by the end of the semester (before exam week). This report summarizes your work and is generally expected to be in the 10-15 page range.

• Each group needs to make an oral presentation summarizing your work. Your primary goal is to educate the rest of the class about something that we have not had the opportunity to dig into as deeply as you have. You do not have to impress us with publishable results. We prefer to gain insight into the goals or approach or strengths/weaknesses of a particular software tool, for example, than to be blown away by some esoteric results computed with that tool.

  Each group will have two class periods to present your work. Class periods from March 16 through April 26 are available. Not everyone can present on the last day ...

• As part of the oral presentation, each group should prepare a 1-2 page handout summarizing your work. If you get this to Dr. Ribbens in advance, he can get sufficient copies to hand out in class. You must also submit an electronic copy of this handout: TeX, LaTeX, postscript, html, or plain text are acceptable. With your permision, Dr. Ribbens would like to post these brief summarizes on the class Web page.

Sample Project Ideas:

• Generic example. It is anticipated that the most common project will be ``Software system X on machine Y.'' That is, groups will investigate a system such as MPI, PVM, HPF, HPC++, Pablo, etc., etc. on a machine such as the Intel Paragon, IBM SP-2, a cluster of workstations, etc., etc.

  The goal is to learn enough about the system, and exercise it on some nontrivial examples, so that you can describe and illustrate it to the rest of the class. For those of you already working on a research application of some kind, it would be great if the examples you choose were related to that project. Dr. Ribbens can also suggest example problems to try.

• Something more exotic. By this is meant something that doesn't fit so neatly into the generic template, but would still be relevant to the focus of the course and of interest to the class. Examples might include:
  • A survey of object-oriented approaches to HPC, perhaps with particular emphasis on one system.
  • Using WWW technology to do HPC.
  • A survey of issues impacting HPC from operating systems, or software engineering, or HCI, or some other important contributing sub-discipline of computer science.
  • Problem solving environments that make HPC more usable for a particular application area.
  • Other novel approaches to HPC (e.g., a visual programming system such as AVS).
  • Software tools for designing HPC systems.

Questions:

• Do I have to choose one of the software systems mentioned above? No. In fact, through the proposal approval process, Dr. Ribbens will try to coordinate things so that we don't have redundancy among the projects. Visit some of the useful links, especially the parallel computing archive in the UK (slow at times) and the list of parallel computing sites at UMd for some ideas of what's out there. There are also some good links in the well organized site based around the book ``Designing and Building Parallel Programs'' by Ian Foster.

• Can I really get access to all this software and hardware? No, probably not. Dr. Ribbens will work with you as much as possible to provide access to a variety of software and hardware. And projects that actually get their hands on real software on real machines are definitely preferred. But if you are very interested in a system that we cannot get access to, it is still possible to do a project on it based purely on paper-ware. Of course, relatively more work will be expected of these groups because they won't be spending all that time trying to get code to run ...