COMP 242 - Fall 97
Survey Paper (Due Thu Dec 4, End of Class - Absolute Deadline)
You should write a paper on some relevant topic in operating systems that has not been covered in class. Ideally, it should cover recent research - SOSP'95 and SOSP'97 are good sources for such research. To have sufficient depth, you should read 2-4 papers on the topic. (The variability comes because some papers are longer/more complex than others. 3 conference papers is probably what you should aim for.) You should then, in your own words, summarize the main points of the papers. If you do need to quote from the papers, explicitly indicate that you are doing so. Of course, you should try to keep these quotes to a minimum.
Assume the reader has taken comp 242 and an undergrad OS course, but has no other background, and wants to understand this topic. You should explain all terms you use that are unfamiliar to the reader. In other words, imagine that you are making class notes for the students of this class.
Keep the length to 10 double-spaces pages in 10pt font. You will be judged on the quality of writing, complexity of the issues you present, and the insight you show. A few suggestions for good writing: Have abstract, introduction, and conclusion sections in addition to the main body of the paper. Be very clear what the points of the paper are, and try not to say anything that is irrelevant to those points. Also try not to say the same thing twice, unless it is the abstract, introduction, conclusion, or some other form of overview. The complexity of the issues you present will, of course, depend on the complexity of the papers you pick. You will be judged on how much of that complexity you are able to express. In the insight criterion, I will be looking for some new way of looking at the topic that is not expressed in the papers you read. Of course, I dont expect may new insights in such a short time - but if you do gain some, do put them down in the paper. In particular, if you see a design space of solutions to the same problem in the different papers you read, you might want to explain the dimensions of this space. (An example of a design space was in our discussion of IPC, where we saw various issues/dimensions such as reliability, synchrony, and buffering and saw different ways to resolve each of these issues.) In general, a survey is more interesting if you compare the papers issue-by-issue rather than system-by-system.