Write a 10-page (double space, 10 point font, 1 margins) survey of three or more papers in some sub-area of Operating Systems. Ideally, the surveyed papers should not have referenced or at least discussed each other. Also, the papers should not have been discussed in class. The best way to find them is to pick three papers in some recent OS conference or journal, as these papers are almost guaranteed to not reference each other. You can also look at ACM SOSP, Usenix HOT OS and other conferences, as well as the ACM TOCS journal.
You will be judged on the amount of integration in your survey. The least possible integration is having a separate section for each paper that does not reference any of the other papers. The maximum possible integration is writing a seamless survey in which the boundaries between the topics covered by the papers is not visible. In this case, we break the union of the topics covered by the papers into a set of issues and alternative approaches. To illustrate, consider IPC. It is possible to cover it by independently summarizing each paper on it in a different section that summarizes the specific system discussed by the paper (CSP, Ada and so on). A more integrative approach is to still have a section for each paper, but show the relationships between, for instance, the synchrony provided in CSP and Ada. The most integrative approach, the one I have tried to follow in the design of my class notes, is to break the area into a set of issues such as nature of ports, synchrony, buffering, message selectivity, reliability; and present and compare alternative approaches (presented in the surveyed papers) to addressing these issues. Time constraints may not allow you to take the extreme integration approach, but I do expect you to do better than independently summarizing each paper.
The amount of integration you can do is a function of the papers you survey. There are at least three kinds of integration possible:
1. Providing a single abstract description of multiple concrete features in different papers. For, example describing the addressable entities in different IPC systems as a port.
2. Presenting features in different papers as competing alternatives to a common issue and comparing these alternatives. For example remote assignment in CSP vs. remote procedure invocation in Ada.
3. Showing that some feature in a paper helps overcome some problem with a feature in another paper. For example the multiple receive construct makes synchronous receive more useful as otherwise a process cannot receive a message from one of a set of ports.
Therefore, try and pick papers in which one of these three integrations (or some other integration) is possible.
You will also be judged on the quality of the content and writing. You should try to convey as many ideas in the surveyed papers as possible and not shirk from describing the more complex concepts. You must use your own words in describing these concepts do not reproduce the text in the surveyed papers as your own, which, as most of you know, is considered plagiarism and is a honor code violation. If you must present text verbatim, put it in quotes and give the reference. The number of quotes in your paper is an inverse indication of its originality and the amount of work you did.
Try to have a good flow, where each paragraph you write builds on something you said before. A paper or a talk should be a mystery story where each part is answering some question raised before, ideally in the immediately preceding part! You should also not duplicate information except in the introduction, abstract, and conclusion sections. Moreover, do not give anything irrelevant to the immediate point you are making. Use the spell and grammatical checks before you submit the paper.
Naturally, I am not looking for something perfect only for something that can be reasonably done in two weeks. This is the reason why I have never before formally described what makes a good survey paper for 242. But now that an IP is a requirement for graduation, I thought it would help to give these principles of good writing. It will not hurt to try practicing the rules mentioned above in this mini-IP!