Purpose:

To encourage students to engage in experimental work that is either a masters-level project or the beginning of a PhD research investigation, and to help them learn how to share and discuss ideas about system technologies, abstractions, and designs in a constructive environment.

Paradigm:

A participant can choose to engage with other students who are working on an existing project (several are described below) or can propose an original project. Proposals must be approved. The general criteria are that the project must have an experimental focus (i.e., a student must propose to build a working prototype), and each project must have a concrete objective that can be accomplished in a semester. Thus, a student can work on a long-term research topic, but must define a short-term intermediate step that can be completed by the end of the semester. If multiple students propose a group project, the overall project must be commensurate with the group size, and each student must define a specific piece of the project for which they have responsibility.

Background Required:

Although a general background in operating systems and networking is desirable, several students have mentioned that their undergraduate institution did not have faculty expertise in systems areas. Consequently, we will find a way to accommodate any student who has the desire and dedication needed to be a ``builder'' (i.e., enjoys designing and implementing programs, especially in a systems language such as C), even if their background is weaker than that of more senior students.

A Few Possibilities:

Note: the list contains both large, open-ended projects as well as smaller, well-defined projects; think of the list as illustrating the type of projects that are appropriate. Furthermore, because we will focus on finding small, elegant designs and making code easy to understand, almost any topic can be considered (i.e., the topic need not be completely novel).


• Learn A New Programming Language. There are many interesting languages, such as Python, Perl, and awk, that can be useful for specific tasks. Choose a language and explore its features. Design and build a nontrivial system. Compare performance to a conventional language.
• New Router Architectures. This project investigates how to use parallel and distributed systems techniques to make the next generation of IP routers have high reliability. One particular aspect to be explored involves the use of "observers" in which some processors are dedicated to the task of observing the overall system and reporting potential problems.
• Device Driver #1: Ethernet. This project involves building a Xinu device driver for an Intel Ethernet NIC. Intel has promised to donate a set of PCs, and we will need to build a driver before the new machines can be used in the lab. A knowledge of synchronization primitives (e.g., semaphores) is desirable.
• Device Driver #2: USB Drive. This project involves building a Xinu device driver for an embedded platform with a USB memory stick. A knowledge of synchronization primitives (e.g., semaphores) is desirable.
• Device Driver #3: Wireless Network. This project involves building a Xinu device driver for an 802.11 wireless access point. A knowledge of synchronization primitives (e.g., semaphores) is desirable.
• Multicore Operating System Scheduling. Intel has promised to donate multicore PCs. This project investigates how to schedule processes/threads on a multicore platform. It requires one to study hardware facilities and invent scheduling techniques.
• Reconfigurable Testbed System. This project extends the Automated Reconfigurable Testbed software that has been used in the Xinu lab (and should be running again by Spring). There are two aspects: a better graphical user interface, and more powerful extension facilities. Neither requires an extensive knowledge of networking or operating systems.
• Operating System Testing. This project involves testing code for a small embedded operating system. Because the system runs on a relatively slow processor with limited I/O facilities, the challenge arises from finding ways to exercise the code without using extensive processing time or secondary storage.